From 7f97413de702e86bc099040597cdef13a9e5c685 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Mon, 15 Jan 2024 17:44:07 +0100
Subject: [PATCH 01/29] rm rxn with no solv set to False

---
 Makefile                 | 4 ++--
 README.md                | 2 +-
 orderly/clean/cleaner.py | 2 +-
 3 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/Makefile b/Makefile
index 038bd1dd..9de70f56 100644
--- a/Makefile
+++ b/Makefile
@@ -161,10 +161,10 @@ paper_3: paper_plot_uspto_no_trust_unfiltered_num_rxn_components paper_plot_uspt
 # 4. clean (filtered)
 
 paper_clean_uspto_no_trust_filtered: #requires: paper_extract_uspto_no_trust
-	python -m orderly.clean --output_path="data/orderly/uspto_no_trust/filtered/filtered_orderly_ord.parquet" --ord_extraction_path="data/orderly/uspto_no_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_no_trust/all_molecule_names.csv" --min_frequency_of_occurrence=0 --map_rare_molecules_to_other=True --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --remove_reactions_with_no_solvents=True --remove_reactions_with_no_agents=True --train_size=0.0
+	python -m orderly.clean --output_path="data/orderly/uspto_no_trust/filtered/filtered_orderly_ord.parquet" --ord_extraction_path="data/orderly/uspto_no_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_no_trust/all_molecule_names.csv" --min_frequency_of_occurrence=0 --map_rare_molecules_to_other=True --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --remove_reactions_with_no_solvents=False --remove_reactions_with_no_agents=True --train_size=0.0
 
 paper_clean_uspto_with_trust_filtered: #requires: paper_extract_uspto_with_trust
-	python -m orderly.clean --output_path="data/orderly/uspto_with_trust/filtered/filtered_orderly_ord.parquet" --ord_extraction_path="data/orderly/uspto_with_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_with_trust/all_molecule_names.csv" --min_frequency_of_occurrence=0 --map_rare_molecules_to_other=True --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=0 --num_cat=1 --num_reag=2 --consistent_yield=False --remove_reactions_with_no_solvents=True --remove_reactions_with_no_agents=True --train_size=0.0
+	python -m orderly.clean --output_path="data/orderly/uspto_with_trust/filtered/filtered_orderly_ord.parquet" --ord_extraction_path="data/orderly/uspto_with_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_with_trust/all_molecule_names.csv" --min_frequency_of_occurrence=0 --map_rare_molecules_to_other=True --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=0 --num_cat=1 --num_reag=2 --consistent_yield=False --remove_reactions_with_no_solvents=False --remove_reactions_with_no_agents=True --train_size=0.0
 
 paper_4: paper_clean_uspto_no_trust_filtered paper_clean_uspto_with_trust_filtered
 
diff --git a/README.md b/README.md
index 781dea34..a1e9558d 100644
--- a/README.md
+++ b/README.md
@@ -167,7 +167,7 @@ If you would like to extract all data in ORD (instead of just USPTO data) simply
 
 ### ORDerly-condition
 
-```python -m orderly.clean --output_path="../orderly_generated_datasets/orderly_condition.parquet" --ord_extraction_path="data/orderly/extracted_ords" --molecules_to_remove_path="data/orderly/all_molecule_names.csv" --min_frequency_of_occurrence=100 --map_rare_molecules_to_other=False --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --scramble=False --train_test_split_fraction=0.9 --remove_reactions_with_no_reactants=True --remove_reactions_with_no_products=True --remove_reactions_with_no_solvents=True --remove_reactions_with_no_agents=True```
+```python -m orderly.clean --output_path="../orderly_generated_datasets/orderly_condition.parquet" --ord_extraction_path="data/orderly/extracted_ords" --molecules_to_remove_path="data/orderly/all_molecule_names.csv" --min_frequency_of_occurrence=100 --map_rare_molecules_to_other=False --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --scramble=False --train_test_split_fraction=0.9 --remove_reactions_with_no_reactants=True --remove_reactions_with_no_products=True --remove_reactions_with_no_solvents=False --remove_reactions_with_no_agents=True```
 
 ### ORDerly-forward
 
diff --git a/orderly/clean/cleaner.py b/orderly/clean/cleaner.py
index 4e0468af..15bb7245 100644
--- a/orderly/clean/cleaner.py
+++ b/orderly/clean/cleaner.py
@@ -938,7 +938,7 @@ def get_matching_indices(
 @click.option(
     "--remove_reactions_with_no_solvents",
     type=bool,
-    default=True,
+    default=False,
     show_default=True,
     help="Remove reactions with no solvents",
 )

From 6c577e2c526c9da8e2d71cd12be7f536ee4434be Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Mon, 15 Jan 2024 17:46:17 +0100
Subject: [PATCH 02/29] rm rxn with no agents set to F

---
 Makefile                 | 4 ++--
 README.md                | 2 +-
 orderly/clean/cleaner.py | 2 +-
 3 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/Makefile b/Makefile
index 9de70f56..d5e3755c 100644
--- a/Makefile
+++ b/Makefile
@@ -161,10 +161,10 @@ paper_3: paper_plot_uspto_no_trust_unfiltered_num_rxn_components paper_plot_uspt
 # 4. clean (filtered)
 
 paper_clean_uspto_no_trust_filtered: #requires: paper_extract_uspto_no_trust
-	python -m orderly.clean --output_path="data/orderly/uspto_no_trust/filtered/filtered_orderly_ord.parquet" --ord_extraction_path="data/orderly/uspto_no_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_no_trust/all_molecule_names.csv" --min_frequency_of_occurrence=0 --map_rare_molecules_to_other=True --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --remove_reactions_with_no_solvents=False --remove_reactions_with_no_agents=True --train_size=0.0
+	python -m orderly.clean --output_path="data/orderly/uspto_no_trust/filtered/filtered_orderly_ord.parquet" --ord_extraction_path="data/orderly/uspto_no_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_no_trust/all_molecule_names.csv" --min_frequency_of_occurrence=0 --map_rare_molecules_to_other=True --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --remove_reactions_with_no_solvents=False --remove_reactions_with_no_agents=False --train_size=0.0
 
 paper_clean_uspto_with_trust_filtered: #requires: paper_extract_uspto_with_trust
-	python -m orderly.clean --output_path="data/orderly/uspto_with_trust/filtered/filtered_orderly_ord.parquet" --ord_extraction_path="data/orderly/uspto_with_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_with_trust/all_molecule_names.csv" --min_frequency_of_occurrence=0 --map_rare_molecules_to_other=True --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=0 --num_cat=1 --num_reag=2 --consistent_yield=False --remove_reactions_with_no_solvents=False --remove_reactions_with_no_agents=True --train_size=0.0
+	python -m orderly.clean --output_path="data/orderly/uspto_with_trust/filtered/filtered_orderly_ord.parquet" --ord_extraction_path="data/orderly/uspto_with_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_with_trust/all_molecule_names.csv" --min_frequency_of_occurrence=0 --map_rare_molecules_to_other=True --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=0 --num_cat=1 --num_reag=2 --consistent_yield=False --remove_reactions_with_no_solvents=False --remove_reactions_with_no_agents=False --train_size=0.0
 
 paper_4: paper_clean_uspto_no_trust_filtered paper_clean_uspto_with_trust_filtered
 
diff --git a/README.md b/README.md
index a1e9558d..4ac2ae0f 100644
--- a/README.md
+++ b/README.md
@@ -167,7 +167,7 @@ If you would like to extract all data in ORD (instead of just USPTO data) simply
 
 ### ORDerly-condition
 
-```python -m orderly.clean --output_path="../orderly_generated_datasets/orderly_condition.parquet" --ord_extraction_path="data/orderly/extracted_ords" --molecules_to_remove_path="data/orderly/all_molecule_names.csv" --min_frequency_of_occurrence=100 --map_rare_molecules_to_other=False --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --scramble=False --train_test_split_fraction=0.9 --remove_reactions_with_no_reactants=True --remove_reactions_with_no_products=True --remove_reactions_with_no_solvents=False --remove_reactions_with_no_agents=True```
+```python -m orderly.clean --output_path="../orderly_generated_datasets/orderly_condition.parquet" --ord_extraction_path="data/orderly/extracted_ords" --molecules_to_remove_path="data/orderly/all_molecule_names.csv" --min_frequency_of_occurrence=100 --map_rare_molecules_to_other=False --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --scramble=False --train_test_split_fraction=0.9 --remove_reactions_with_no_reactants=True --remove_reactions_with_no_products=True --remove_reactions_with_no_solvents=False --remove_reactions_with_no_agents=False```
 
 ### ORDerly-forward
 
diff --git a/orderly/clean/cleaner.py b/orderly/clean/cleaner.py
index 15bb7245..99e584d8 100644
--- a/orderly/clean/cleaner.py
+++ b/orderly/clean/cleaner.py
@@ -945,7 +945,7 @@ def get_matching_indices(
 @click.option(
     "--remove_reactions_with_no_agents",
     type=bool,
-    default=True,
+    default=False,
     show_default=True,
     help="Remove reactions with no agents (ie no reagents AND no catalysts). Does not consider solvents",
 )

From b11d0dceb39dbf21a1a34cbb71903a3c05080a1e Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 13:46:08 +0100
Subject: [PATCH 03/29] added orderly-cond-prelim

---
 Makefile | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/Makefile b/Makefile
index d5e3755c..f8708ad4 100644
--- a/Makefile
+++ b/Makefile
@@ -186,6 +186,10 @@ paper_5 : paper_plot_uspto_no_trust_filtered_min_frequency_of_occurrence_10_100
 
 
 # 6. clean (final)
+# NB: I changed this one, min_freq=0, train_size=1
+paper_gen_orderly_cond_prelim: #requires: paper_extract_uspto_no_trust
+	python -m orderly.clean --output_path="data/orderly/datasets_$(dataset_version)/orderly_cond_prelim.parquet" --ord_extraction_path="data/orderly/uspto_no_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_no_trust/all_molecule_names.csv" --min_frequency_of_occurrence=0 --map_rare_molecules_to_other=False --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --scramble=True --train_size=1
+
 paper_gen_uspto_no_trust_no_map: #requires: paper_extract_uspto_no_trust
 	python -m orderly.clean --output_path="data/orderly/datasets_$(dataset_version)/orderly_no_trust_no_map.parquet" --ord_extraction_path="data/orderly/uspto_no_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_no_trust/all_molecule_names.csv" --min_frequency_of_occurrence=100 --map_rare_molecules_to_other=False --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --scramble=True --train_size=0.9
 

From b22e6d515d00ca7d843744cc0d85d67e556c11f8 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 13:46:26 +0100
Subject: [PATCH 04/29] added test case for removing C when Pd present

---
 tests/test_extract.py | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/tests/test_extract.py b/tests/test_extract.py
index c69987ab..e369b28b 100644
--- a/tests/test_extract.py
+++ b/tests/test_extract.py
@@ -338,6 +338,21 @@ def test_rxn_string_and_is_mapped(
             [],
             False,
         ],
+        [ # Test that [C] is removed correctly when [Pd] is present
+            "ord_dataset-0b70410902ae4139bd5d334881938f69",
+            262,
+            None,
+            [
+                "O=C1C[C@@H](C(=O)N[C@@H](Cc2c[nH]cn2)C(=O)N2CCC[C@H]2C(=O)OCc2ccccc2)N1",
+            ],
+            [
+                "[Pd]",
+            ],
+            ["O=C1C[C@@H](C(=O)N[C@@H](Cc2c[nH]cn2)C(=O)N2CCC[C@H]2C(=O)O)N1"],
+            'C([O:8][C:9](=[O:32])[C@@H:10]1[CH2:14][CH2:13][CH2:12][N:11]1[C:15](=[O:31])[C@H:16]([CH2:25][C:26]1[N:30]=[CH:29][NH:28][CH:27]=1)[NH:17][C:18]([C@H:20]1[NH:23][C:22](=[O:24])[CH2:21]1)=[O:19])C1C=CC=CC=1>CO.[C].[Pd]>[NH:23]1[C@H:20]([C:18]([NH:17][C@H:16]([C:15]([N:11]2[CH2:12][CH2:13][CH2:14][C@H:10]2[C:9]([OH:32])=[O:8])=[O:31])[CH2:25][C:26]2[N:30]=[CH:29][NH:28][CH:27]=2)=[O:19])[CH2:21][C:22]1=[O:24]',
+            [],
+            False,
+        ],
         [
             "ord_dataset-0bb2e99daa66408fb8dbd6a0781d241c",
             0,

From d92c33e2c5eac5d8b3909120b0c2fd5d8263ef8a Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 13:46:59 +0100
Subject: [PATCH 05/29] added [Pd]/[C] -> [Pd] replacement

---
 orderly/extract/defaults.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/orderly/extract/defaults.py b/orderly/extract/defaults.py
index b619b742..2e310281 100644
--- a/orderly/extract/defaults.py
+++ b/orderly/extract/defaults.py
@@ -79,6 +79,7 @@ def get_molecule_replacements() -> Dict[MOLECULE_IDENTIFIER, SMILES]:
     molecule_replacements["[Pd on-carbon]"] = "[Pd]"
     molecule_replacements["[Pd].C"] = "[Pd]"
     molecule_replacements["[Pd]/C"] = "[Pd]"
+    molecule_replacements["[Pd]/[C]]"] = "[Pd]"
     molecule_replacements["[TEA]"] = "OCCN(CCO)CCO"
     molecule_replacements["[Ti-superoxide]"] = "O=[O-].[Ti]"
     molecule_replacements[

From 9f39b6e4417e5a9fd91a99955a2d414d03953262 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 13:47:18 +0100
Subject: [PATCH 06/29] delete [C] when transition metal present

---
 orderly/extract/extractor.py | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/orderly/extract/extractor.py b/orderly/extract/extractor.py
index ab9d4b45..ad3bd310 100644
--- a/orderly/extract/extractor.py
+++ b/orderly/extract/extractor.py
@@ -1007,6 +1007,20 @@ def move_unresolvable_names_to_end_of_list(
             catalysts, rxn_non_smiles_names_set
         )
 
+        # Add paladium on carbon exception: Delete carbon if Pd exists. Expand exception to other transition metals
+        def contains_transition_metal(agents):
+            for agent in agents:
+                if orderly.extract.defaults.has_transition_metal(agent):
+                    return True
+            return False
+
+        # Check if any agent contains a transition metal
+        if contains_transition_metal(agents):
+            # Remove "[C]" and "C" from agents
+            agents = [a for a in agents if a not in ["[C]", "C"]]
+            
+
+
         if _yields == []:
             _yields = [None] * len(products)
         yields = _yields

From 88613465efd9e0ace508639752c4800666ab5e53 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 13:47:31 +0100
Subject: [PATCH 07/29] change back to df.applymap

---
 orderly/clean/cleaner.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/orderly/clean/cleaner.py b/orderly/clean/cleaner.py
index 99e584d8..a61c5383 100644
--- a/orderly/clean/cleaner.py
+++ b/orderly/clean/cleaner.py
@@ -518,7 +518,7 @@ def _replace_None_with_NA(
             all_component_cols += component_columns
 
         sub_df = df[all_component_cols]
-        sub_df = sub_df.map(lambda x: pd.NA if x is None else x)
+        sub_df = sub_df.applymap(lambda x: pd.NA if x is None else x)
 
         df = df.drop(all_component_cols, axis=1)
         df = pd.concat([df, sub_df], axis=1)

From a880d9ce5bdcd843b2ac6a42e9394f0c6b538120 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 13:47:48 +0100
Subject: [PATCH 08/29] find [H][H] regex fix

---
 notebooks/orderly_benchmark_stats.ipynb | 216 ++++++++++++++++++++++--
 1 file changed, 201 insertions(+), 15 deletions(-)

diff --git a/notebooks/orderly_benchmark_stats.ipynb b/notebooks/orderly_benchmark_stats.ipynb
index bb64af05..0f1099b0 100644
--- a/notebooks/orderly_benchmark_stats.ipynb
+++ b/notebooks/orderly_benchmark_stats.ipynb
@@ -272,7 +272,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 1,
+   "execution_count": 2,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -281,15 +281,14 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 29,
+   "execution_count": 15,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "833112\n",
-      "1\n"
+      "Number of strings containing '[H][H]': 9738, that's 0.011688704519920491 % of the data. Len(df)=833112.\n"
      ]
     }
    ],
@@ -297,8 +296,48 @@
     "# path = \"../data/orderly/datasets/orderly_no_trust_no_map_train.parquet\"\n",
     "path = \"/Users/danielwigh/projects_local/ORDerly_project/orderly-benchmarks/orderly_forward_train.parquet\"\n",
     "df = pd.read_parquet(path)\n",
-    "print(len(df))\n",
-    "print(df['rxn_str'].str.contains('[H][H]').sum())\n"
+    "# contains_hh = df['rxn_str'].str.contains(r'\\[H\\]\\[H\\]', regex=True, na=False)\n",
+    "contains_hh = df['rxn_str'].str.contains('[H][H]', regex=False, na=False)\n",
+    "count = contains_hh.sum()\n",
+    "print(f\"Number of strings containing '[H][H]': {count}, that's {count/len(df)} % of the data. Len(df)={len(df)}.\")\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 23,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "1862\n"
+     ]
+    }
+   ],
+   "source": [
+    "contains_hh = df['agent_002'].str.contains('[H][H]', regex=False, na=False)\n",
+    "count = contains_hh.sum()\n",
+    "print(count)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 30,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "0\n"
+     ]
+    }
+   ],
+   "source": [
+    "contains_ = df['agent_002'].str.contains('.', regex=False, na=False)\n",
+    "count = contains_.sum()\n",
+    "print(count)"
    ]
   },
   {
@@ -345,24 +384,171 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 2,
    "metadata": {},
-   "outputs": [],
-   "source": []
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "833112\n",
+      "1\n"
+     ]
+    }
+   ],
+   "source": [
+    "# path = \"../data/orderly/datasets/orderly_no_trust_no_map_train.parquet\"\n",
+    "path = \"/Users/danielwigh/projects_local/ORDerly_project/orderly-benchmarks/orderly_forward_train.parquet\"\n",
+    "df = pd.read_parquet(path)\n"
+   ]
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 4,
    "metadata": {},
-   "outputs": [],
-   "source": []
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "833112\n",
+      "0\n"
+     ]
+    }
+   ],
+   "source": [
+    "print(len(df))\n",
+    "print(df['rxn_str'].str.contains('procedure_details').sum())"
+   ]
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 7,
    "metadata": {},
-   "outputs": [],
-   "source": []
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>agent_000</th>\n",
+       "      <th>agent_001</th>\n",
+       "      <th>agent_002</th>\n",
+       "      <th>date_of_experiment</th>\n",
+       "      <th>extracted_from_file</th>\n",
+       "      <th>grant_date</th>\n",
+       "      <th>is_mapped</th>\n",
+       "      <th>procedure_details</th>\n",
+       "      <th>product_000</th>\n",
+       "      <th>product_001</th>\n",
+       "      <th>...</th>\n",
+       "      <th>reactant_001</th>\n",
+       "      <th>reactant_002</th>\n",
+       "      <th>rxn_str</th>\n",
+       "      <th>rxn_time</th>\n",
+       "      <th>solvent_000</th>\n",
+       "      <th>solvent_001</th>\n",
+       "      <th>solvent_002</th>\n",
+       "      <th>temperature</th>\n",
+       "      <th>yield_000</th>\n",
+       "      <th>yield_001</th>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>index</th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "<p>0 rows × 21 columns</p>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "Empty DataFrame\n",
+       "Columns: [agent_000, agent_001, agent_002, date_of_experiment, extracted_from_file, grant_date, is_mapped, procedure_details, product_000, product_001, reactant_000, reactant_001, reactant_002, rxn_str, rxn_time, solvent_000, solvent_001, solvent_002, temperature, yield_000, yield_001]\n",
+       "Index: []\n",
+       "\n",
+       "[0 rows x 21 columns]"
+      ]
+     },
+     "execution_count": 7,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "hydrogenation_rxn = '[CH3:1][O:2][c:3]1[cH:4][cH:5][c:6]2[c:7]([cH:12]1)[CH:8]=[CH:9][N:10]2[CH3:11]>>[CH3:1][O:2][c:3]1[cH:4][cH:5][c:6]2[c:7]([cH:12]1)[CH2:8][CH2:9][N:10]2[CH3:11]'\n",
+    "df[df['rxn_str']==hydrogenation_rxn]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "index\n",
+       "127288    [C:1]1([C:7]2[CH:12]=[CH:11][C:10]([OH:13])=[C...\n",
+       "865425    N[C:2]1[N:7]=[C:6]([S:8][CH2:9][C:10]2[CH:11]=...\n",
+       "49540     [NH2:1][CH:2]1[CH2:11][CH:10]2[N:5]([CH2:6][CH...\n",
+       "327337    F[C:2]1([O:9][C:10]#[C:11][CH3:12])[CH:7]=[C:6...\n",
+       "703910    [ClH:1].O1CCOCC1.[OH:8][C:9]1[CH:10]=[C:11]([C...\n",
+       "                                ...                        \n",
+       "137770    [N+]([C:4]1[CH:5]=[C:6]([CH3:11])[N+:7]([O-:10...\n",
+       "377939    [CH3:1][CH2:2][O:3][C:4]1[N:12]([CH2:13][C:14]...\n",
+       "321854    C[O:2][C:3]1[C:17]2[C:12](=[CH:13][CH:14]=[CH:...\n",
+       "492935    [CH3:1][O:2][C:3]1[CH:8]=[CH:7][CH:6]=[CH:5][C...\n",
+       "109743    [OH-].[C:2]1([C:10]([O:12]CC)=[O:11])([C:5]([O...\n",
+       "Name: rxn_str, Length: 833112, dtype: object"
+      ]
+     },
+     "execution_count": 6,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df['rxn_str']"
+   ]
   },
   {
    "cell_type": "code",

From 7b4b0f545cb6adf64afe0ff157d0de052abd517d Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 13:48:45 +0100
Subject: [PATCH 09/29] changed paths

---
 notebooks/inspect_orderly_data.ipynb | 140 +++++++++++++++++++++++----
 1 file changed, 121 insertions(+), 19 deletions(-)

diff --git a/notebooks/inspect_orderly_data.ipynb b/notebooks/inspect_orderly_data.ipynb
index 06fa5bda..75de1c6d 100644
--- a/notebooks/inspect_orderly_data.ipynb
+++ b/notebooks/inspect_orderly_data.ipynb
@@ -10,7 +10,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 36,
+   "execution_count": 1,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -23,7 +23,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 10,
+   "execution_count": 2,
    "metadata": {},
    "outputs": [
     {
@@ -32,7 +32,7 @@
        "'[S:1](=[O:4])(=[O:3])=[O:2].[S:5](=[O:9])(=[O:8])([OH:7])[OH:6]>>[OH:8][S:5]([OH:9])(=[O:7])=[O:6].[O:2]=[S:1](=[O:4])=[O:3] |f:2.3|'"
       ]
      },
-     "execution_count": 10,
+     "execution_count": 2,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -53,7 +53,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 12,
+   "execution_count": 3,
    "metadata": {},
    "outputs": [
     {
@@ -62,7 +62,7 @@
        "'uspto-grants-1993_09'"
       ]
      },
-     "execution_count": 12,
+     "execution_count": 3,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -73,13 +73,13 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 37,
+   "execution_count": 13,
    "metadata": {},
    "outputs": [],
    "source": [
     "# Can also find ORD file just given the hash:\n",
     "def open_ord(ord_hash):\n",
-    "    folder_path = '/Users/dsw46/Projects_local/ORDerly/data/ord'\n",
+    "    folder_path = '/Users/danielwigh/projects_local/ORDerly_project/ORDerly/data/ord'\n",
     "    # look for files within that folder path or deeper for a file that contains the ord_hash\n",
     "    for root, dirs, files in os.walk(folder_path):\n",
     "        for file in files:\n",
@@ -93,25 +93,45 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 39,
+   "execution_count": 15,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
-       "'uspto-grants-2001_07'"
+       "'uspto-grants-1986_09'"
       ]
      },
-     "execution_count": 39,
+     "execution_count": 15,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "data = open_ord('ord_dataset-85c00026681b46f89ef8634d2b8618c3')\n",
+    "data = open_ord('ord_dataset-0b70410902ae4139bd5d334881938f69')\n",
     "data.name"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'/Users/danielwigh/projects_local/ORDerly_project/ORDerly/notebooks'"
+      ]
+     },
+     "execution_count": 8,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "pwd"
+   ]
+  },
   {
    "attachments": {},
    "cell_type": "markdown",
@@ -122,7 +142,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 14,
+   "execution_count": 17,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -131,34 +151,116 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 30,
+   "execution_count": 18,
    "metadata": {},
    "outputs": [],
    "source": [
-    "path = \"../data/orderly/uspto_no_trust/extracted_ords/uspto-grants-1993_09.parquet\"\n",
+    "path = \"../data/orderly/uspto_no_trust/extracted_ords/uspto-grants-1986_09.parquet\"\n",
     "df = pd.read_parquet(path)"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 31,
+   "execution_count": 20,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
-       "'[S:1](=[O:4])(=[O:3])=[O:2].[S:5](=[O:9])(=[O:8])([OH:7])[OH:6]>>[OH:8][S:5]([OH:9])(=[O:7])=[O:6].[O:2]=[S:1](=[O:4])=[O:3]'"
+       "262    C([O:8][C:9](=[O:32])[C@@H:10]1[CH2:14][CH2:13...\n",
+       "265    C(OC([N:11]1[CH2:15][CH2:14][CH2:13][C@H:12]1[...\n",
+       "273    C([O:8][C:9](=[O:35])[C@@H:10]1[CH2:14][CH2:13...\n",
+       "275    C([O:8][C:9](=[O:25])[C@H:10]([CH2:19][C:20]1[...\n",
+       "276    C(OC([N:11]1[CH2:15][CH2:14][CH2:13][C@H:12]1[...\n",
+       "277    C(OC([N:11]1[CH2:23][CH2:22][CH2:21][C@H:12]1[...\n",
+       "279    C(OC([N:11]1[CH2:15][CH2:14][CH2:13][C@H:12]1[...\n",
+       "281    C(OC([N:11]1[CH2:15][CH2:14][CH2:13][C@H:12]1[...\n",
+       "285    C(OC([N:11]1[CH2:15][CH2:14][CH2:13][C@H:12]1[...\n",
+       "368    [ClH:1].C(OC([NH:12][CH:13]([CH2:23][CH2:24][C...\n",
+       "377    [C:1]([O:4][CH2:5][CH2:6][CH2:7][NH:8]C(OCC1C=...\n",
+       "379    [ClH:1].[C:2]([O:5][CH2:6][CH2:7][CH2:8][NH:9]...\n",
+       "582    [CH3:1][C:2]1[O:3][C:4](=[O:14])[C:5](=[CH:7][...\n",
+       "Name: rxn_str, dtype: object"
       ]
      },
-     "execution_count": 31,
+     "execution_count": 20,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "df['rxn_str'][0]"
+    "# find df['rxn_str'] which contains [Pd] and [C]\n",
+    "\n",
+    "filtered_df = df[df['rxn_str'].str.contains('\\[Pd\\]') & df['rxn_str'].str.contains('\\[C\\]')]\n",
+    "filtered_df['rxn_str']\n"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": 22,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'C([O:8][C:9](=[O:32])[C@@H:10]1[CH2:14][CH2:13][CH2:12][N:11]1[C:15](=[O:31])[C@H:16]([CH2:25][C:26]1[N:30]=[CH:29][NH:28][CH:27]=1)[NH:17][C:18]([C@H:20]1[NH:23][C:22](=[O:24])[CH2:21]1)=[O:19])C1C=CC=CC=1>CO.[C].[Pd]>[NH:23]1[C@H:20]([C:18]([NH:17][C@H:16]([C:15]([N:11]2[CH2:12][CH2:13][CH2:14][C@H:10]2[C:9]([OH:32])=[O:8])=[O:31])[CH2:25][C:26]2[N:30]=[CH:29][NH:28][CH:27]=2)=[O:19])[CH2:21][C:22]1=[O:24]'"
+      ]
+     },
+     "execution_count": 22,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df.loc[262]['rxn_str']"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 23,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "rxn = 'C([O:8][C:9](=[O:32])[C@@H:10]1[CH2:14][CH2:13][CH2:12][N:11]1[C:15](=[O:31])[C@H:16]([CH2:25][C:26]1[N:30]=[CH:29][NH:28][CH:27]=1)[NH:17][C:18]([C@H:20]1[NH:23][C:22](=[O:24])[CH2:21]1)=[O:19])C1C=CC=CC=1>CO.[C].[Pd]>[NH:23]1[C@H:20]([C:18]([NH:17][C@H:16]([C:15]([N:11]2[CH2:12][CH2:13][CH2:14][C@H:10]2[C:9]([OH:32])=[O:8])=[O:31])[CH2:25][C:26]2[N:30]=[CH:29][NH:28][CH:27]=2)=[O:19])[CH2:21][C:22]1=[O:24]'\n",
+    "react,ag,prod=rxn.split('>')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 28,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'O=C1C[C@@H](C(=O)N[C@@H](Cc2c[nH]cn2)C(=O)N2CCC[C@H]2C(=O)O)N1'"
+      ]
+     },
+     "execution_count": 28,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df.loc[262]['product_000']\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "O=C1C[C@@H](C(=O)N[C@@H](Cc2c[nH]cn2)C(=O)N2CCC[C@H]2C(=O)O)N1"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  },
   {
    "attachments": {},
    "cell_type": "markdown",
@@ -1248,7 +1350,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.10.12"
+   "version": "3.10.13"
   },
   "orig_nbformat": 4
  },

From 1899235f81b5254fe705f5f286442f411c4c29e0 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 13:53:43 +0100
Subject: [PATCH 10/29] make black

---
 orderly/extract/extractor.py | 2 --
 tests/test_extract.py        | 4 ++--
 2 files changed, 2 insertions(+), 4 deletions(-)

diff --git a/orderly/extract/extractor.py b/orderly/extract/extractor.py
index ad3bd310..37768638 100644
--- a/orderly/extract/extractor.py
+++ b/orderly/extract/extractor.py
@@ -1018,8 +1018,6 @@ def contains_transition_metal(agents):
         if contains_transition_metal(agents):
             # Remove "[C]" and "C" from agents
             agents = [a for a in agents if a not in ["[C]", "C"]]
-            
-
 
         if _yields == []:
             _yields = [None] * len(products)
diff --git a/tests/test_extract.py b/tests/test_extract.py
index e369b28b..de8f6e65 100644
--- a/tests/test_extract.py
+++ b/tests/test_extract.py
@@ -338,7 +338,7 @@ def test_rxn_string_and_is_mapped(
             [],
             False,
         ],
-        [ # Test that [C] is removed correctly when [Pd] is present
+        [  # Test that [C] is removed correctly when [Pd] is present
             "ord_dataset-0b70410902ae4139bd5d334881938f69",
             262,
             None,
@@ -349,7 +349,7 @@ def test_rxn_string_and_is_mapped(
                 "[Pd]",
             ],
             ["O=C1C[C@@H](C(=O)N[C@@H](Cc2c[nH]cn2)C(=O)N2CCC[C@H]2C(=O)O)N1"],
-            'C([O:8][C:9](=[O:32])[C@@H:10]1[CH2:14][CH2:13][CH2:12][N:11]1[C:15](=[O:31])[C@H:16]([CH2:25][C:26]1[N:30]=[CH:29][NH:28][CH:27]=1)[NH:17][C:18]([C@H:20]1[NH:23][C:22](=[O:24])[CH2:21]1)=[O:19])C1C=CC=CC=1>CO.[C].[Pd]>[NH:23]1[C@H:20]([C:18]([NH:17][C@H:16]([C:15]([N:11]2[CH2:12][CH2:13][CH2:14][C@H:10]2[C:9]([OH:32])=[O:8])=[O:31])[CH2:25][C:26]2[N:30]=[CH:29][NH:28][CH:27]=2)=[O:19])[CH2:21][C:22]1=[O:24]',
+            "C([O:8][C:9](=[O:32])[C@@H:10]1[CH2:14][CH2:13][CH2:12][N:11]1[C:15](=[O:31])[C@H:16]([CH2:25][C:26]1[N:30]=[CH:29][NH:28][CH:27]=1)[NH:17][C:18]([C@H:20]1[NH:23][C:22](=[O:24])[CH2:21]1)=[O:19])C1C=CC=CC=1>CO.[C].[Pd]>[NH:23]1[C@H:20]([C:18]([NH:17][C@H:16]([C:15]([N:11]2[CH2:12][CH2:13][CH2:14][C@H:10]2[C:9]([OH:32])=[O:8])=[O:31])[CH2:25][C:26]2[N:30]=[CH:29][NH:28][CH:27]=2)=[O:19])[CH2:21][C:22]1=[O:24]",
             [],
             False,
         ],

From 957060aeb02610d179113059e72e3d37d5fce033 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 13:53:50 +0100
Subject: [PATCH 11/29] regen test data

---
 .../extract_config.json                       |   1 +
 .../uspto-grants-1976_01.parquet              | Bin 835375 -> 835352 bytes
 .../uspto-grants-1979_09.parquet              | Bin 319641 -> 319631 bytes
 .../uspto-grants-1980_08.parquet              | Bin 403081 -> 403061 bytes
 .../uspto-grants-1982_08.parquet              | Bin 580051 -> 580054 bytes
 .../uspto-grants-1986_09.parquet              | Bin 453402 -> 453362 bytes
 .../uspto-grants-1995_11.parquet              | Bin 678317 -> 678280 bytes
 .../uspto-grants-2001_07.parquet              | Bin 2067902 -> 2067818 bytes
 .../extract_config.json                       |   1 +
 9 files changed, 2 insertions(+)

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extract_config.json b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extract_config.json
index a01d232b..a9b26011 100644
--- a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extract_config.json
+++ b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extract_config.json
@@ -340,6 +340,7 @@
         "[Pd on-carbon]": "[Pd]",
         "[Pd].C": "[Pd]",
         "[Pd]/C": "[Pd]",
+        "[Pd]/[C]]": "[Pd]",
         "[TEA]": "OCCN(CCO)CCO",
         "[Ti-superoxide]": "O=[O-].[Ti]",
         "[[Pd].c1ccc(P(c2ccccc2)c2ccccc2)cc1]": "[Pd]c1ccc(P(c2ccccc2)c2ccccc2)cc1",
diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1976_01.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1976_01.parquet
index 20d01210209976cb7b5e63a5a5cd273619a21e52..90e2f09e9cc325aaaecde7a5ffc3cda563dace30 100644
GIT binary patch
delta 2723
zcmZXUdvFuS8Nk1_@9>><TFYnM*%FfckWZFuWJ~MdF%B}e4?hEvVS<xjObP<?Dh^QU
z36q&X0F%a{m~y~TGf6NlDS}{LfdIp!Or2&3v?Zi<ff5QSZo(wEm?@A@P`irrM<d^T
z-|p=EzHje#zsLS4*%l=SM%&E2U5QfP)G^YfObMh>5(n_tkMyBUf~*<*hQp1v!|Y13
z8Nf09Szz2jd@92f4ikS?&KMKp4zX<zE?)wU1gqzG;V2-BfEeM`ai}~CNC2Ysh)jk*
z8lh<tYHfhiJ|GRy836Gj&jf?=xm0jhXNVtIK&{$_PfrIz4UEvKgXFB?LAk-zd0TM%
z=$W7n&Bkwv90|L{4axMg=#&s9#h$X|R&l{-RB-bUo{w#+hra$msAG}1$_#kmZ}{wJ
zl#OilTgY#LPpg_Wn&d*a5yVg${x&_MS*@sRaR;r{+Lt+sBSl)aojV<dZXa~#0eb?!
zgIT^u=pF~vQwO|SfwRQ8+&r}iki@A@^+B=L=Cq}aGcaoKwGMt%>s%mKV0)HjX>U0T
zyc<&(5<YT?3Hz&Oh})8tr2Hxi{@9d1z>^M;w&N=q_@RkuR`K!C@*CQH=fs+&A_RB&
z#e;(Qquj{dvoFqiEUqyLVS4_yMM}q5$IG0Z!w-1~;VP%Y;bLDEnPefL20}HN0IBj5
z0RpnYuC2_qqYt&Wa%}<er4$I%_*`UJ?8xYt&>Zp-c{wNcWUy&xJus0!B1(CT>c>ty
zD=?m*ya8JWE&O><9T&8RxpqUJ+tuf;&V^uEk(Q~rxufY^kf_#fEu>7~o)<ZHjYq81
zwkoCRtE~O2tmGH5*rh%oQf!v);Q88s@=8K-rZ(F(?&(8$ok@rd5f@^|VlJ7;F(W)C
zi%~=Pr~x0B%wfOB8}@fR;KFV%_GCeuxY3V0C96y8cTGU|G&#@5hm`W!h<BAowNRdl
z4r*Fn3+mK{@|2VvrG@KCaV<B{AqHLeXPV+J&bXNESZ#-3$(aoHYUS@+p9h>@1VOo9
zTjG{bySC0<fKF+L+{K2)Pn+FEY9LsIf2rMdYv`J`*7NfuB@k5tK}E~RKf!hRZl*!A
zb&7U1f0Fg9YLgmL>CH_uGeN&m7zi`rHhG~T=Q*vSU<%r+Z7TRMA*oh-VN73be<SB5
zE%?vL)0|>s5Zo^;klVA`cL*c`_#Yt}&ed)RWPb_-|Jx(W1_+n!RNIawD3&TSga0d_
zNNOW^T#vOm-g-1y>-FZK?b;{a@he+r!67fS)<UZl#FsgID3@~wReY@u)UyZ6)|QcS
z;5R8pD_e*SqQ;oC83E>s2WYL~D=yBjs)&ac?6os~kNhMX{3wm9Ybiz+IzGkbK%M+;
zwly&iWiu^f@m3z5!dkj~r=Dvmc8iTUE1^#G<>0}hlEZwrh?LMQR^n-`MDw78R?n;1
z{Tp~7uK=E9y^~M7#<FwMq-IhPYXwq6+xn!|7eJa&GB+7U;j7Pz>6z2BflP&-D|m?^
zl_PeLDw||j(7=|x8{o+^&}syJ2FZe^4e+@W@WtYl_)nbo0feojUJOaJ&Kw)+*z3z?
zno@J{{2-3-_={3nkzbS|6X}Vaevs~9oa|-%YIJe^KNr_WS){Pmv*~5c>D5n!#yJI~
z{u8Eg4$tM>c3Br2z%ToFarF#vi$^MYQgEG&5L2}*x)a!)QO8)9$uTzC1;bL_MH5-u
zQIu0Ox#4!*K+{cGto4BV_QX5IPo6_L_}olfks{u8mV9Y%dI`v;Smv{*9BisKRFr}C
zYstQ>tf?jw2^WX`)?J*Z2`@F#oR}R8gi+*F;*qy%4L%!Mq0RG=89#NR`d;0Jd^Ib8
ztc%^~AKrW;sX^xo6-8RZ@;Q+dKmH5Oudpaf%1<Pa-H@$7sAL5k$^sjF|Ikh!-I^#}
z_5Xj{KXd-Owtv{*pORb|oe#s(J|Cb|TqWW=HdBp<7QGy?n7`|tes+|2+#-G^ilIXB
zSp^j`flwQpN|vK)q4I>b>*-+g_M$-sDU8w_otHQ)i|0iz0F_4<VApL2<1GJpOsaiw
z2q{C|l@OhTs63j4*fXyY7%OfNO{zQW{|Ww=5xn?6JmNEGgAb}|5WlPi#}~GQCULZK
z#Rdy0_&&{tlbw#A8$tPXwB1gnrrm!1SEguP!ljs4Tlaup>&WPFQ73VBO3g>PHMszd
zIK2t4Ef`!d&bnY6-kO%bMY}TIVf)I+mR9XTtX#DRvoe~4UpLXJ9Uio_z<5-!rZMn(
zV*pW2pqfaPp~_MnMKy^kN0q0VOx5r~V?bEC&~~{iQ68LX0r|j^@g)Ej8x09a2{b;g
zxAo6{BUtUb^+b6sp%!od4#k`7)XlpEr+Mf75#)gWOpF}ZoDlDGOkW)%$9hMQ_x0m3
z@_ye4(y8CwE#%9cS17rZ`Gdc_aYgq<gfZqDG4{g)tk;)Bgj}><-x3jA3D@@Pry@d@
z`P%-(c*8z@C?a^xeFsL6Z93T_xTI|-%j11+e626uBjidqzOIZ%{hRf7W2k>~RXn;J
z(f>|2lrKj{pl$ai>NoZXX1!>ykSfzvyzjM7YlaOL?iEZDUBw$b?$EpT3X1%=qc$Er
z+EF{~dp+h$SMdh7j_8T+3EAdbM@FFadMPFJ8mGS}<Rx71($~C44|lyQ9KXliTl&cu
zx_fIR()Ih_A$?A~*Efgt!hJ%n`J2Nd&^o<!pP-o6ogYCq>hDk=v{CQdNB`;aJ^ju;
zAv-pkdTFSyZ`4bB>6*^XbX|IK(_EXBY>~i{!LL{__s4Ty^!bRCAYaOqPcw28&24HC
QpIP$yp%0Gt3hk!<0d#rNp#T5?

delta 2758
zcmZXVdsGzH8Nj~_GdKgoWf^t`c3BqoG5dfluop!@7i9-FtTDT23~G>ibd{JFu_U!=
zZ19mlEk{9_*zu4iCulH&VyF*%B(cXNrM4z%qIr5+Pc$`&rg~Cyq9>`Yy_@()3p3w0
z_k8E~-I+W0d)#et@=BbH<XViKJ;~DfxpzpX@+FW`Bo5%sKWqLuvUa01ZGVGhKf9*?
zI-^?en+=R7f)C~xVlm<`usaQmC(5=!ta=4le=vFfP@DqfI*@c|_oM0pAOVQiAaXan
zn+{WI`vP#R6v%vN48W2FT$9=?zQHq*i2Q0Mc<NHcK@+IKF5EQ_hyz;Ep>Ye4Qji=W
z9+Z|dT)nRuugC@U-}CTOB1d8#acwI7JbGP-kuq;pi&<>SMMdXc!uMf|>ZP|o5bbCd
z+l+uOyo2}WA{(-Vd&u*^X9e%cB~Iu|2eBp#f1RDvq|U5w_C(B6wa@H|6-jNIxMMNs
zD}_Ecut)J0%<}8veU(r*_d>qraFkDPG0rUkByp-ky{2TLx^?q2IkGdcX!i49A!6mm
zMucSV23(ZJsF8IYd`io5AZB3u+eB@vqsX^DgCVhlr<j<(ZjsoPs!S@JY{IV@3P0va
z2T0H1lR5aJfoW3l?p%4bcG*!IdRTzSfM47ph;yBT-yC1O<c9bQgAk*;(%P(aOtC(}
z<#G5r4^ceXVKunec9BUH0%{-{$_L1luL}^6Z(6ipI`f(iWrDhHAMpd&2a1nZnWv@>
zR6rDoNE)q6#k|0%fsoxr+7dg24hv0DACVvDL@kHSI_`y9{-8L?&8U7X<*@?ejmT@U
z`7;xL5>)Fk?TRxmZPeo$_0%~bQf1R*#l!8&<|0HjZ`4A{3~sK-c|u;XOxvVXWVe|=
zYBLiW=XA)tMx?|jUB*kbzbmVgZdkR2uF7dIxw|GIHcDKGor1YkA}5@%wSZBhcr*<k
zk&H3F*BA45T;pOMAI>d+R&l)_cS+_P?H$()^iM6_UCP%e)k_iYs{Td`xmC17>vA`v
zUu$2vm5kns;wLKbo!o^Eai$ADsaZT_#c$ZG>+>K|{>L2lOwCx!0>G9Mh{z`h1k7J&
zGqIRlk`*M#*3Lu&;tK64PZ4@W8}yW=EuLodl&FD73D&ePJsNsfTkYL6DLW9)4n(rG
zjKZVby3%u5&}3G%_X}sEe`waC92C~37R^D=Yrie}Q_}Y#t>KQh@=i2xKGKXoNWIq~
z1|#5E_<+2sVAYEP83g>H=J!d`vuf*Tl41%P8GK7XgP9HBbzRY#d|@<08}QlDChdf8
z`WmeAG>OVzM6=uw-vH$sKy2gi9w+CCsQ5%ZsK<9#wO7$d^S#PQ&3s~kczQxw(*c%?
zL$o0AX&2{LRm4LxcIPpEuYBDGew4-4H<uw3eNJZWP%nRLGv9O}8`C@mZ{}e#-jpqW
z8s^qCmwCill#7LS{8dT$FyAL4C0cV!;%PO--B3>JtKIfdHxFbr;3o5BC+#n72k(`d
z$jn63kY=@ZPHI^Q(v0%OsgQ!l=82j4_t}8Vh5pm{!8BTRc_7s^%J9G}wrb~8p4^ov
zLVgh`fW~fEbKU{?uVwfzobMXM%+|0NmT1XYpR8xkwk<GZ+VQ*y-ofK@6|^|dRUiZD
zPyGEA`iAtoeT-j?FARUNFdS!*!kS;ruDUn7?xxVNtcZl)XBw9Aoc(i``TjKU%Z0qy
zx(GbtD>M5ua8O2w3AV+zgUoJ^=S*?sOo{ivPg30j*rjbLvDa75x>$dq@tiEyc)@eA
z_Hx<H`N)nB{tQ=Th?gAY=kppL1=5>HcJ_7aEyMA@5$lge5i|P%(^t~N65oFpgdcoP
z`%g(h!EA$p#L8lR^8n{<#19!DU|X7qO}fab#4B&srj=UIW7>jJlJcknh4*gJHkTTM
z`HJ*3%iDKk`0<Chu*#&YtUj7Vc7aWSX!);TUjbO)_C=3{tG`Q@F8TlWj5u;KaYpRF
zQ936zd}RgjD{!BMrdxajQK=q5P$51k;x8?Rke8Ocyu)PtvC;b36mh>v{8$vj#bQK3
z#Y`aDddomsQC&*S0j+yl#JRQP3WF3zX^JmN{z)EL5`P=0{O~pmd~S`~tKaIKo@?14
z8dOir{}{fQ4nF)HkN6zg_9Lnq!EM^q>BU@lHb)CmtkoWwZd$W4FTOG_v@-AP^M-hR
z(kUuIbITC;wT_&A7bya9QAX&vGvowlz?qG>z357;(j2SAo3jdg3L5)>JPl(6$c#iB
zewYotGk`t4s)JRwRbNPFD}n>G?+0JPtc;f7bq3mqZ$E^nH`;(#rfF=T3@J}n27Xf+
zKva{cCR1govQ$&3PNK?D<*BApO*>Q>5Ptq(%1BSLyzw4=<Yhssc<oJOkpz<jrW}6B
z{mWY)UtC&Bq$K%NzWjSVS$>w7K&C#^%K=Ol8aR-WXcS%V={H{%9LD#)n?TO$lXp>a
z_SvNIbH30YN{}!1P9W#??Fn-JtqJ6?{^>5EP(J)AC8zTLZwcLJbkm@4hq3$2gjq-R
zg@b|<9o5?h1y@qnh`w`BC@^-7B#&RxRsGDM;4@yGNXGQsA;Be$y;ePbPWNTKaY%4V
z-Ir^|qi442Jqh&8*5G*5Ii#PUFEn-zO+aIM(r&>mja{uBKdI{jy?VExNL?R<#-pK5
z{fPt`>b!G2+SRY`OQ2o-6HxD{{;%DF&Dc9S0UgzCdjuQZGj)&PPTsgVrq3A`vh=lk
z1cOYi@gI5N!fl5m)Ip-w1cz;U!?0jfZYR?xTC%*sLX+u*ZOg{H?mY0Lt6nrL7>qj)
zOyCFf=&+y|51gDpj_DhQX`qkk!^46*sq<U?qhZ087{4E;f#0}Ax9z1i9c%XrE_6wM
Xh??@H-sR(yN&e)UL)-TXs|^1I6gAGv

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1979_09.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1979_09.parquet
index b383f19dab7eb1185fdbf7fdfae8f9bb4947b7d7..32552c282686e4c47083ea84ca7e36d0986d6658 100644
GIT binary patch
delta 1045
zcmZY7Ur19?90%~<ndgRcbDO*OT-_YP-P-i3L9H+u&QdcpP=m-K5hC>VCPuJcN+qNG
zFPL=DdQj<gGE@@GK*<br>Ut@W5k1Hngc35u2c^#RrS5$=Tz-6h-{1M&%N<$d2N(I!
zl{6vVhe+5_NMZ&8<P=E-n7K-&4>)VOHLdrQ!u>H;H8xaPEspYHo%~CD%ipZ=mO;xS
zi|p_@!Dj|<4FuAmeuP&pK)n&UFP45ckOKX*TeMW1X>`?cHj~rH>B|)}CrbUkuYj^F
zb^Z!}or$DYN2E*K<I;ndcQbN?=fjA+v;MnD=e6V}+0(&{y_ELRbl`4CGMtyb8T}2l
z1@;BIbO?Ez8T;$3pY0%9d+a_QijEMeQJw&CRd<QiR+A^2Yf_5AYX?&46+vNB&Prv9
z3luvvx8^juq<pS6aELe;dcf&T-7ccHcBdP!-z)Iy`SL2@WO5(rmDioJ_(B#DJ34k!
zW@QlvW*kHPNZfbWKzu5f6am;A>>~ov0f6U#zF)yMM`##{Kdh|dpi<qjNq?_kC%<`H
z4HMM8%3z{d4U_av5++;J5T%0*glIqw`)%~oD$W!3+e(UJhW{3vqbQ6E`MGv=CPXiX
zu?vN0Qy6>DGVKjxNmw33>b^0$zy^@9agFQ_P_%~K!tOw(Iuxb8q(o=4)G|aLu3?Wb
zw5*YB8cfReN1HbD^EfkZS4$6-)>)=!N+VmefeG1~*GOWCzFKGRnpo0MjIJ_4F}kyk
zb9M7Uy10P_n!kZ9WIpItUp^W+cK11=6JrMPunUQMnd~1Jjc8i?;wPituu(!^T;m;d
zVFSBKM?xc0M<+>Bge9krG#;bg2tLhLz7@f_C{EiW*vTFf$tpxwl2wTAB&)<M`GTLB
F{{nak`f~sP

delta 1044
zcmZY7Ur19?7y$5d=DFha&TZa*ZgW}lw&~D7ZZ={#3(atWe}a-q3^H8ggBYX_5-qJz
z%<w_+)I*{KzA2fa4;7_NAKaWbhyt^hgyDnOsHHXvI@6cD_u+8(_<i4ZzH>PJbKKK8
z?!}pO{#zdumTZMWlo5cCDx?80c^0?6Mdqp2^hc>phD|f7FK07sWhZp54RMsacup<_
zEw@cBizgR&j6kje?HQnMi1QbNIvP5POTREe0X<d0n#zwgNY#|Za9B^#W!H_AXz%rW
z0?=%#@s@jQ3~ZWj*?x+;Q~GP*T4oL(AAmxb{eO^zFF8o|)I{cwlz=}2u#8?`*-&3e
z*V>)nxc#HvTVGvZePgx%2Kkf`d28696}Z@LtA_*b1f-t9D!Bs~nk8}rOS3N^o@Nb)
zvZ<@G3`n=SM$l0N>?i#a@LNICjhqV-{7l_#{|PAx-lAy3JxQP)ZTi~A0-5IS^I|tI
zLb7yZrOfgo>YTRqJ%(c3W&vWCtHcd}#YWx`2s!|Oa~n@CA&YHz5Q^I@$=I$?XKdo=
z63XQ^FR5V$XNL&PG^t@8*Ct`UMGZ0hkN_WRS3_?AzY3vzzBf?fjxq-ZjNqLRD&!(W
z4o)7#?l6*I5MK!+CtSsyVPxl5N1!@y8qb7Lp)ftFk*<Ec6Gjfct3OL!6vIW!L}IgM
zwT$7L%gD)(Eox*JKTpc;dyBT_>t4;MRV~9fD?)09Uut9vpC=-0z1B$20`7~DyY?(-
zD1qmQzyw~8pgf&2j$<o`$46F>Nl?Ze>cc0LeZQRa2}R>08+-1M1)UgW#I+&vix~1p
zMm995gYo@VA<~K;jB9)me_tW*x%gKj37oH>qh#bN1?9m%xJ^O1WP6f~g7{rB3gY!-
Mv^L9pKo5-n02q}2H2?qr

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1980_08.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1980_08.parquet
index c363934b6cfe0188dd9900baf2220a00311df796..da8960319c18bdc0bb8ed5c72f404cd36eb805a8 100644
GIT binary patch
delta 3088
zcmZWr4{#LK8UNm8-*S7o+q>JlT`p(H<#Jwf$0a8BkTV35k}MY!qIUsGEG9QLY;(cL
zoK1vCu_dH57^xG5WmrrBAvXLQR3IrR>J%+nh!oR^w8jooQCjWTn1W)hdiEVK9h;kZ
zJG<ZhzW2TF{oe2Ub|3qL<MXqQ!>i2pbBB@mg1=Dg<wX!#Q4g^12imLGGQ^>s=G!yo
z&bbfd(2p|X7h%E)*y-K&dLF5RT31=?G+gL?4TP2o&IQqTj)#XTirK{?kd;78;C+74
zBdXEn?0S*hpRz?lByeMp-NzgepDfEO{U|fEg2APgF_9ycDIfld38H0J5WYiy*Zl!*
zegg+2oC0u#yRhn3rn&&SSI{u2Hw%#mah8ZIS9K5<EPqOeig78IM`nYVO<fL}z!)u;
zeeOk_p?hGlpgY(ApReMEaFmHOw+(`Ln&Y5o^GT+Nw9N*x8StFN`w?DetP8{pE%hLu
z>4RwauYxBNm$|n$)I+rdRoehBuV_-gwdK{n#Ck<{)Jn*mAm(7M`5T%<oTjCL%)e!X
z$An7;kdKUC&elg7<vo9bph<$rk;zxX`kENwrZGM_%Yoc`u5o(08~vH-YP&tBi{(h*
zdXy6Pe9To6%yqAe&j$CXlxyJ$7xzKj3D%dv`F`)7y7zzLZ3;H(WjPXh#LyXli5!UU
z1L6dpHFMWs2_C}FLxi7c<x8x54<Fi%NI9G{0c7@Y|5}!eL0JsWZHD;m5T4z=L*(Lx
zUC{4@3rX+}eqRYD6vZbuCIT^{JVc-5s9z!cdCnN|CxR`15l-l{4ET8SwD?XCJ=OG_
zDDN)Hk=sqlZ?3gR%NwJCSaWUXiD_ic^#kjuY62`(y$wiWqU9X{i8R3kWz2-y@y?GG
zY(~Cqd_CuMn$P^htxM@>10*CJw>*&tRB+c_#k2hjfiF^D5R1Be_6Y-E19=c?5x>wz
z4``ClT_Tfm%7`orCmo~^z%&?h#<|L$g56~0KQUPuV|mmcMN1Qxc+vp<l#K$k()o$a
zP86RK^z+9omI(3t<h!%MZ{^5h=q@?40Uf;@u24**1d=9bGl6?-Z7nl*@ouPEgUGMJ
zf&gd^f8!@J(T9mISa*NE#E0({Ug6QNdLG(Uo7xE1Ky25Sv$-#X3e=rae)aD-CMr8i
z?DKrGLX?E!TO#`#6E!h~pJxRnv^JwZ0b(;<6ifU*L*nlkd(jRuH^t>`x7o#Fkqh4?
ztX8?{g<Vh2U}U-Hec^xrLFq|>8vZYk=O*u?6|;>bFO@nSWhS_HX~x)l8RR{v)KN$N
z9!uG2E6GoV^_F+EWA8Pd5lCsu-WUd)$K2i8wgAA~yh`5~3Gj-Yf7s4Mi`t5StV?Ni
z(@UTZPxtG$+Vzw`X_<O?dXu@-#z|0f<i|Nbw22?^+tqqUUXZwGt<b2}Vwl_ftO@Ze
z9nfSlO0z8d0%7JSRtREiexbkwqX#~vi8(veXB>rSr#k9z2J!}^8~IZem`p7#-G8Jd
z_gefNsy#l$Nv+vsUr#6Vz<a2=RkdZ+WB!d>m`1fVtH(Angv7m$?>4u~TKiX68WNo^
zqtavtggPJ@flvhS$?7|d;d0s@{3qRG$xaA$rjWJ>G<bE%b_lhDH;1VU_;ac@t9-WH
zwK0?5m`Mf^zOS<%Unv1vUUYkSEqXWAD^9ULZ)VAY)S=^UqM&|S1qHPD__gNj@+Ceh
zLtnJ)f(0+H`T)rHZXVA7<?}3n;LThw`=b6vdH(#GSdcc$NKm9AlcRD_ru~Gz^vUX5
z*>fHMo%<@RoR|bW>3wAN9zqMT2}B>QLAvhXJr#Ci|2*g~0K7Xnjkz?9X;=rM1hZOd
zjr1^M(sq$=w}20u^)w&hplDR+sueZ}wPc8Q$E^sgxb;8*Za_F*taf{)u47IR_8Tpg
zD+9A)Sq7rhwGVbLW1`Vb59&z8jT)s?z19x#qp<0hfIWL|Bh?U;i3p)BS$Iz8e#X1I
zdNGg7sve@k?yLvWm)iR-%F^q`npk5^$b=@k0C#znRHJT4l`u<7leQzJG0{cYsIo$`
z2R8M<gaTMeR>D{%cw@<LjaA<o<8K0Am%}vHwVg%rxv6dmwlL`BU_>}#$g$aXeU4h{
z4<0hbdm-tBG5S6G?L{*KW`2o5(DX)`R{nDpdi-$J4u<GZkG1*&LvBq`*;fa6%SBFI
zpOeRIG)QaIy*W9?UISSQS6&3~?}h)Uqd5!FZ1onGv*@`&jDtdhG3h|N94I+*b8H(y
zudg28#7UC|XGT8L7%dCWi0@ZdxN^~x>SHb!ldyLt)I+WZ?5Bp2c$ojLW(V#~eUKeP
za;{gW_IkxG(HY*DylT8LoL7zfNNU$fGydOr&Wwc;r@pG@_JLeiat%XAVht01Gr0&4
zF7i&lv?{m%R#a-Fs#1a}YHI>9MM<784V*CLm-1+xBP9hz2_~Bi=_8B2%4;(8zoMI<
z+A6s6Y=-cqf!2o9pt?LWn1#6q7pn6;B2v@_Pa|5r;i%`)qaH*r9lbK>#T-BC!7Dz-
z|2~YQOPb5XN}*G&*`hH^ge(Ww+}e^@bzeyd5p_~8FMT$Iq)|r}NT*uog#fa$>3<4{
zI+}X^R_y^3>&%%LM$$M>J+`h#Q(Y$dv}a6go;cB$28}<ZjZ?@z{!|(?rdfE_V;_4a
z4H?kxq=-21Y#MUynAXM9aMzBdK_l94cy^{Za*`sw{QqsCPy38#i|l=`rcK(S%{H?h
zv_)HHX7kZjZM~Uwi(5ZR^PBugJ7#7B(quYx<<l8j(8Ag@gN3!)sWp|I(^|2G{>q#_
z;hKt^8r4cHtX=z=g|*mEji%v;v^{jDbZ9&c+47-FTV$nEC+VUhwWfxCe7#n0WkoGu
zrJ)~R@1BaE)1p>3UpjX+4LScr-i^>t)2R}*rcV9Ej++kereafTszaaljg_8J-`+H2
zSeqrVZZxbd71#pZ@BwX7VC~+L8yY;c6-gsAsltX26i*GAeEYhq_NG9WOun6l_i3LB
zl;ggCrXg>9bz^3^jkSnxe3gb@?9rCnSdWh0AGNXhy04$mp0}~?)P0QF=%n_gjm@LC
lY&%<k&T9+oR3Yd0mZXVnyS6q3ZQqv$Nq;_m*v`Id{VyY?16}|C

delta 3067
zcmZuz4RjM#7QQdd44E`DNhUMXBpZ@WW7@>F(mWswg;bicwfrS4l`c}M)@*}C3{qT{
zgKk4{LHq-CRt6(hTLhI;{shayB8tbKMcG{h^sI`y;Ln1`tp&B7b=k0QKs;`<=iE6j
zFW<XA@80{}nQsp}@BOoL#~oH-#||XE9;y&mb0UbeXaZP2Ta}B&T~AoY6ifhwdPBcT
z9f6!1B5nvg`kIVq!(?xDY6MjgI0$@Ol<`60o&LJ#Yp$YaihvgYW+AZc7t2(O!=3p9
zZrArB#uCUBDDn%=cu-N41tC-znP;TN)y@?eT&thq&a#lGK8vWIo2IpwGn1Maz;!x6
z)yNewzp=gyy5<ox`ae^UA2GJLqQp9|4<@S&ja`67QRC7cK>so}EAu-6Hv*c&S`twu
zDBm8r4Q5&lXw<j$Sr(YBRKgf<oxdK$j~E6jAKq)M#Pi1j{vA-`XZ9kh(=xTZ#oRUp
zltcGGqV9d(Q%F_Ik2KAIdI@6dfV#P6mOA7pnX#J=h;qV?@wFg&DW>&15>1N4r-8!#
z$pZKCADBVedF(UM6mM2~jzE}|Aind=!5ULT3uZ<bgNn<E<esk>6S<b@kWIZs@ZNgC
zgv&2lsgfsAOf9CEuBEB5ARm!jHm;mW1tH}I`+jgA{J5v?WJh3bxY?vA@yH$KCFL;S
zh14^E-N0Som?3DSc2V0Q&iU<J$j)_gk*5%j!e3Z`!j;U~M)j=VH1+0y_cMVfR8Dke
zUSllrP&(XpobNNa%+&VU3#msz^w*POj0Mq0?k~GY5n5}7M6@|k-qKpTq;CWsf3YeZ
zvAIA=v15QFV<Ot&k?>p?AlYEiBI=1V&6E|b<6?}v(0b^UY|wk<Zb)+m=H_$>;Hj5R
zjt~!LrvT?w`^3tOOBgT%WyT9(GU6sUNb*l|#f=J%lG`c@ztf2;0N72l+&3__cY(m#
zxer;p(GpFB5@=TX7>ApHR2$0B0{4dw0VjjWaEN=xW{cxcP?_$5ke$IZp{wf9o#^?`
zqBXNvTm?B6=CdFlT{hV`e&$mUTZ-`0(DN#Z;B)SDA^J4^6)k5=O<dg^emjRAU%9w@
zvVISy2|F%2bI9*ToT@67*8h<)CKR_UObjYXoJ=Q!jY_CF-O!TcD(#SrEGx*SL2M;1
zDT#a1oc;$z1<(dOLC@q>K0zEMGIi7V+tdqEMe#!;jf%4JB%kFWEOqf@6Z#JMf5`fg
zno4zQkw;9}s`uG(sjeT0b}XU-L)?&Ey{pJ=yiVY5QL{zkO{)YGNr5y){h(;BwaURr
z(6IA%?`((o3HOj%?<@)XeWa6#0sAPJ(7KgHT#FOtvKFb_#!cj}{L~I%%3gW}Zwx1%
z`-(*2PO6VNE3R2nOhO@<PB$5l;%SaGjIL}^{1LASxepwaa$|r<dJMF+b-hhE@=9tO
zX(%qjNV7fF-xT4A5a_F)trogw82Ojnpx)$KDQww=#6IUwhtfOOUmQxDdq0W_1^`eN
z<d#BYDGbDcic@>*t1NX9Lh8_7`K^I=pxSe_&|M2nUPG=EBApQM8mE?rys`UKgGb3^
z3b{-n-h`+dm#EaaDj=ND$I3GFo=)%==@oNme2so3stYGkmb7;nVGOs}>WMZ6Nq+?s
z)qk~7H^)8!e6`*(M0#1c?=S!#D4e7O6QO2h(xir#Fd236uz1p4kx;@4nE~X#pdzcU
zdB%BiEaPUGz@5MD!Gv7XkQgMqA}%-mvW8W?Bfa=xk(+OWAZ0ZXkNif_k|cI(91v+M
z5T|21;#L(|nw1ti)%u)F%|H|msc#ocnJ-x2do8xw1?8h#TxM)FEPAr5(U?dywHt6)
zk2FH!T?DTzg49cJ-~B-OdxE$g!U`54veHG3U(#y~JXJr9qpD+zNpmil0eFqB*UiF%
zYHlG!7DC`AeSS9S>?<IwN+**^y|u$JJG2ix0@Nd`R>D9Ms3fVv(ORJXmK(CfhAgSK
zpsoa_&OeM&6Z9bljyf<Y;W+=S*;^oVe~a3tZ`sJER_kw@qi&#H`p9AgD;Kr!YfKh}
zkapVy6ZM<eV@7O1bB?KLZ`D=5HA@T0jBPC~4l1E!l5lM_qJ!jDa&#h;#EDRK#*Amd
zxxEm0ga1N3?42^If`xEz2o9`()CS$tCFyWOn_9orocac+FyClNUGc*^PrceFn!gSh
zn-kSzN2b=OH~5OtJ!;11Gqwp!+SFd(t&Rh`k@zb2({_DwS8;KTAye%JeyL<>XlZW9
zJzy9n(;vwC40U(s1}&GwJ!s+H@EV(^=9U|Nlx@omZ7-N&4P$yo`m#-iTp1iKgSxkJ
zacWaMP;_charQqj&O-Q*3}384c}p^x+r#$mVN1ty<4~tlZ<=H>oV(uqf2HJlbM{5U
zKp9YF=e8E`LuNt~sZlNTj|^U0yye>BbG#v!K}32pi2D4a+Zigl<t9r?LV=HSr|8a8
zbSk5^21?Wpe>v*7yW4+Hw;z$mK%N5f825Dhsawuu-rs?w!H^52Q_+bL0GEe+Y=Efq
zxb)~ek7#j;;Bqpp^V-sP+*nnGMT4}OlaB8~(z{NgbadWurDvC;Jgw1}v`+6v(f~)a
z$eGJ(=F863t<~1Dbcwib?T^@HXhl|goj}5hY##KzcAlmE!uK2VkT<o_9PJn1+?t1U
z@6m4Mh`H`PdC)nnhoi3)&%I2L)!hFiqhI?wM^_5{2lJHnX|k2}qkUSPl`cj5wFOpM
z7WbddJGcIf)@!B9#r0?MpzNS_%1Q@>?BK}DiMf3FLG5QYf(-8S4M)DzuG5h(kK`f8
zwFe0z9e+O$+4reW+i9l-?SzfC3uHIk=xz<!=~8L;a(Os1^tF6ZHSj|ZtT>3IRUT4J
z0t`QUeyz5{PK(<8cG@PKUt2O9@6)yuFGAmjJfvScV<)xg-<*e>(0n{CqZ3*!PnVIU
zohP4oC!QNM-1g?<+D4wfLfCvf59-%GAWB02Kl6|QO>hu%1OLuLR;|+F4%%-Z>tY98
zYIyn~?QsV!>+1osN2j&V9CQg;Oack(yjCTU>Yv|SmDfNoYbhOi`I$UO`u&~_0{x!-
Ezd20&U;qFB

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1982_08.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1982_08.parquet
index ce0dc1f702b6993f14177e511bd5823c328906a6..21677838771165edba7d8a30ef5a54fa1c2149f1 100644
GIT binary patch
delta 1414
zcmZXTZ%i9y9LIl8>9tS6RyOa>Qn3TtLjPQ8dm0u9jR_ruXhO}17^!BmB8iP5!51W3
zVrbBf$w(Yt;-f?>5aw7V8(6~BZAkxsX8K}*gb>pg@<KzFhN|I3-HXQO+}?PX+&7nR
zKHu-<_q%&;CuQ4B*_M4a_RrfI?sL1q70euPIt~MrzL0)<ich^}Gib-QEx>%*cLO?j
z5a+-#^zJKG(FbKC#Ej6Fh4Ki*M!?jJ2Y!Dd7<M_o8nl$BAvO)=h<wdQVEt<Z?yF&K
z$T@J4e~GJ(23*mAu}Uir&pi*~@g$xb8PcN*hg}Q0{-<H68_E_qw1C0l=tzE9)sOsa
zIM^EuH<<pe-0#;0!}T3(RE!LFzhV8W0f$_}Ci628NW}DIjA9s}l^gpYUIr#O`mI*<
zgJ=Z4RLg{t14s<PkP(<y%b+_Cy7aQt>a0W8rGBR!wMt`7-dU{QwBl$qVCku1p42+K
zMn5(JGTF^yRNgmXgh>uDHdCK;*ZKOfq8oZF^L4B%>I~B<ur1|w4IsvJ`vq-8`pC^|
zS6i$Z>AE{ww<AK2>4PS4Wp>3w4mr4AH>~^TszaejulIuR;>(iYc@ABZ-t;tk&iXrD
zruDX#<r9uiPVZ^UrPfTTwU2JYkFM!C8)py?kV7orZ11=!edmdo<0mI8(A8p>&(5@!
zi_G~VV;BT)+t~YpFd+ztN<&parKMu1bX59Vf`E0>vom)z{Fc22_`NCA-U3ilOAi2W
z8UScaknj)K#wJ!&q(Cmpr0|o99Fe;+Im)ZZBsr4FWKl&HiD!e7#kfW}!3H@ilZ|N=
znI)geWOiOfO60ywO5do+F43p4o7>G&QZWB-)kV^k#$I+Yrz$Oxi8QvOC9;&pjcAqZ
zq;WmFdZ|J=>I}gdEU+^ls>l@S&frFND&<jjw#j$~JGkw8UL_enAa^p@&W|6c(7k`W
zB#^}{sm$VIEOp9mlK8T?iBJCGQ<A+u{Ew>fEUxCMQw9=Kfk%Pw<p6ccKw<WA;7ATo
zrwkmdJPtJHa5YDrG7!&@-W+{}@r(*RAeSgX4@feHo6r@qox=|H%C%F<M-)iaCMAV^
z70HuRGRa@=RCnI{ud|>c>tszP>uV~KC&f+r>GOo$qVqf??OV8k-p<HwflSD5fh4Ir
xxqR4J%T?BJP-8Znd;6_-2G98YBB$XCX8wK!f13VYG@(3n{KH%LNBr~ge*q*7Yl{E?

delta 1378
zcmZY7Z)g)|90%~<lisyYjmf&~F2>WvHh<bnlbGg&4m*R8Xz>p;pwKzf4Z6&}D6B8G
zA+7^gH`~qBH|3;Ha5KzUTj^$J4ODTmHfd6aD?=1RH^#P97bk=iN<k*&*mK(}bMSmQ
z9zOh@=l4AK<Wu$sDf=~_om=`|$B#NZe96KCU(I8H(gpd?SInt5?X}D&2?31EHYO5!
z`n2;31kGUB;XL~Oeyfs&L>8Rk-~=cqq4y*hZMgfFje)RB96ur`BhWhnN<<yW2>kF3
zg67L1CM0(En|JedQNJtdH|QCu@1h8jDS<B@>#0G5%kA;%qnn^-A1DHp1*jFAM}m{G
z9<^}cz~MmHW-M+fbTEN%Ly(I~k-mLzS^u)(kgLzQaSQkpZ`UXo^<spkrwSkq0~`2V
z&zMVF+0bM+k_Mp10PGve8r#3drpM&h#d>r>?h+lyE1we0;-iL~6-T3fA*g3xV#N02
zLk2)bi$#ipVt|AJVmxB)#sl)K*!A+GX6O(`>p53c4AcGBZ7Gkd8(B>!>fxa2V}WT8
z8vhU!0nw2Z5QwHLt=1X&+vaHf9SJ&&Cz`;Oxhs`<<m7*yvlc$83xy(wy@Q?udt{$`
z2l`Tuy0^PKTlTt)*V|gJZF7zuT4I#a&P-`%Cq0+>_)HCN=WC_EkyBDG?FwF!Z@VLw
z4_}K_q01}nJ_p;Ttg!tnZ0)<?ZTsLepXYO*2hr$gs%RJ*mZqAf=7!IMtL0Zey{$9P
zIW~iN=`7mS3b1)AeGb6u0H85W4$Wda7oXNhk(^Pb_@hS3WKNaxZyGsGo~d$rMI%YF
zeU3_UKv&tpEpkkiTO%4dPrg#+d{QH;<bf)y*EMpFY)Io~{@y&5lI4Y?CdvLZ_HxOr
zrf8hR(%6B<$#-eoh^9zBjT^YBFRChAJxBBz?BULx(Z~=vn8A(QP|96dvq%Or*vT*6
z^;XLI6Ed5@4sQL4hO*>Y1~+lpe>5^iq<Jc1i@wU`?vwZCX$$v@EtL`{6RO0|`YWYK
z3aS*p*2ps1lEsbu^0Y>tr%7j)-p}*2h8~k)D(Eq}n#D~hMjmFdlZ#F4(Qa$Z(K{&?
zG?E`Cr*fDhT{&FG<%jobD+m6o9Mj0nng1()Rx5ALXgp6=bMznQ$)*L`+Zx%ufNk{i
kjyg+_m^w?4tLki$JY2vH{A8J~@!}IQFW=DL!9SV)1qgCxv;Y7A

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1986_09.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1986_09.parquet
index 820357f6e63c8a0fe8ff0419f4857f8e407b5d77..3e30ed0f9136b13ea14f9c7b5d5740d5c49e984d 100644
GIT binary patch
delta 2599
zcmZvddvFuy5y1D8@6I|~I!R~UVac*A*?JmeTW1i2Al25@itV%#1I4tp#iPPXa2!o|
zISqu^Y0ayp=5cY4DjBF9n^F(BOd#M8hcOHg+GYq38EiTMC&B$EjOZku(j-On6xtcE
zI<q_Xee>JhZ};}wyZH5jFGmZGKCRJqokL8@HH#TmF~AUv9AMkej)gQ2WbLU*+2Ev%
zYX`6^&ttXW6XyV(#P}dicnWj5We`^aWs|EzF?TGsC*D$l;)@|^2i#s`!GbG2Ps#dM
zDHg7sS?QSppf=aMJ`W*v%pHlu^xu&Fjk20%#Ck~Ru0sX!_kr=pD*zvs`)g<0bc*`Z
z2yg$V?-T}kXBI+eO9UKSR>=8Q=h~+UO~8L$=GOpT+HzRoGl9Pv>d!%3jV{^Y^m-Iu
z1r1A~fdwg>xE-u3jGVf9bHfIxSwJ_*d$LRsccd{^8;->G{~-%svg8iM<;D*HX9r-a
zx7BTB=gJk{kTX=>u%iwm3FZY-#Z?1tK$G<b;G963HogQL2T3L1O6QJVzskIgBIRec
zW6ltHG~Nk#i*Fp`7624cZ;}|#yIL(FXT;%(aj{4x{XO9fV`@**=1tsc;G`J)e48+?
z7|;d<A9V2aGL<40iMTxknsb{WIp23SFZw<qB^$h^`nXgWUgW<5RsyST^_d9W3~FVL
z@w{5cauLwYz;l}0V8Hw%2=%|JE$6+wyPSPU0*A7(a3xih%c^pa`qTEmVj-t~k6eF1
zuFS@Gr}3#A!)(5scX;_Gd@&Z{snaxdfmZMi-YG5Rd{q^dpcRhLY$wXY96bIql%sq@
zZP>xPogvQniyZQ|z$3eXck-2z4{tS4BU)<|p0yoPY1Fvc?Dd>M9twa)Q|_cfxXoJ8
zlz`6|^65($Wj0rkE^|B>oC)Lul9*qv<^m9vUe3R6HDpDX0;dOj)vY2N_fi8IZgCnh
z71sImYr?lP+RS1;<_}Xn3e4d${#gx9>ODE|ls{KTV{R9+P=jyp%}3srx8&SU(#J4g
zw~#<xmSc_n`eM+$%`F1_Xy86#R+R0}LPN<3RzJaV9bmRm4i+J-aWRz;KGGRor_$bw
zn$RK!@E@Y(Tvd7kkeA+75Ul`I5;y|*ujRF_Fx`}|e_mkqq1ou)c;RS|(k6F>d1Uac
z*QIC}qydZ+r&E;hJa3zC#s3ISW+2c7R7pDRbwXUff$Fd+m(s)9vYj?vS74V=SWt#`
z3O^|LKuuZHjSFjMh8t)LUhgY33pztLdPVq?p$KgiJ~kNhj-N-&m-@%Y`Q|%?0huv1
zrJ$!2sRZZ}@oynzhyBu|j;DsQ2ZyrV31Ll<LHoBHZYeM<5Olp7tDUVstnrlyJw^HY
z6ggi7pgM9>9&++VjE+7Fm!yuofAFDjt*A_OM=OnPR0>z0`U3C&LQtEGs*P+kOEyrZ
z<Z1wau5^yzHO*?b$y#i(fpI*w89kDln^E#U;P<I+6;@XleFmCeYuk=$P-?rfZM!o1
zdr0YFzaFge)V=J<dw6}@B9s~h&nVz){C<*K1$YU1LmHAIP_O_%r3`XMz)3?PE>{Pf
z5qE&@%T0e!=KsB*x<StHts2^v%la9rC$B4!hYsOO=&yxuOm3XA7GsS<nc0H0!eVo!
z%BwO`n*LEN6r#Q|8%u?5^StmeS|N3jrwlZN*g{(E;2l9HU2gf5#u|gT$;1cUQbT#S
zmu{k!j5^5kp#t6a$tt-c<SgW=O|-I%T!zKm=3UA>R86fZ##Q-^d{ugeO8bS<;!DJ!
z0eJd`kSez2AF$*1&OB)I!fB1GC)wYijzqY15o)HephSm6p`pZzYJ^uyp3(075HY>#
z$J>1&B3$v<kEKd}Ec<C{M{cH^*pa(sE5^6RM)WB&_DFov;xargfy;3IO@;LGTx?TL
z^cJaeMz^5U2>z=P9N+etM|j6pioC)nwg~mQ!xVCb8~N6mm0?yP1nowwh25_cUb4H<
zykGX(xAoeQq%ujfB*i4fPxjggasVGXhuA~n(JKTyci%b#qtG#+%O_Vp*Bo2FyrP0*
zWbCkty>%Y3Zx!f(-M4Om2S8_#{!_rnq=ws$cZm<K5W2z>A0qZ6wPbR<3r)A_5G$_|
zB}~VG3^H+8eBmll%uF24KzE14UI}UM4rQRd;+PaqyZ2lMIVh^GN#x+i8Kh4Pq)Fe+
z|3fxkBTCu6Nr?=ro+vIkDSmK`n59jo*^_m5iVviCs8cj22%GHUZm}*wn6wvn%cdLl
ziJKF|ET(Ty1{q0;rxS#g8A+O^JB^Kt?omQ3Dn<!4Gd5m4onM_wa|?TQDg(tI7K;;U
z-}X^L!%Dm9z7vC%M+T>);96-n-Qb{jYm~4t2ai~%ld~7a+oJ>{mR*<9I(spLcZ@ya
z&!>6ESOy>YT>Q;-!t5FOJOg!g&M`1K-`zg>+MmakuV4LQ^P|gmcrJTEBVDX6@y>O^
zE=zQa${U1*Np#Pdp5E|TF(4uB@YxKsTYT*XVb$*bEQ9P5Po~MfuQEud_>VN{?C?xK
zC@wn32n&jf!7(YLzx+;od5kcppYbumX1)8a=YOS;udHrvT_p{Y`@gl|Y;TrS%iVXy
eNy!v-iHdQ;%yzve9r5JdW0P;3<j08{n*RVDq^KhR

delta 2625
zcmZvddvFui6^GB2ch_25yOP%0#d;u1mi02WvDTna7i+K=JH}xwCImB7bqEMVV_da3
z;I<xuF%2P5hDSEA8Fj{q0aHpbF{S0vIG(z|P-YTpu#6wmQgBVk(8e^Wd9^|574#2c
zb!P9Zj=p=2&OPUMuD%>J$F7;XzpXbMIEVP(6&3J3S{`_cR|8yGaWbfXtZHq(rMgKG
zA<2Ip#7fv=2eAmi&738w>rb5l^eaNDB`IjEZ34#NPX|_|)&--!XmoEg_!yfgL#+Ws
z?8U&TB5+O+wJZGVptDG`sA)!te;R<!Zf;W{q>K8(;i&06T`^x(ITHzf8un!&v)KAu
zh%mrcL>wTRfb3CMRL^q#Ez|ps=mhe9;0*$@`b-eqS_q!4`D&@r+eA|Os*3kj6?!1^
z8=5r%8&s4-4bJ`wBpC2B(7PVRHbLD|s1qQz7-DX4zGD4KTmE|82B@6Nxzuf+*uuVW
zC|X?~j<z;*5aMPamn<1)zseAEAjSaj<=Ym#DIm3`Ea(lEZ(l&r6u*|%iXk8^=vIvd
zL@&^`mLGyB!UO{(<K2EhqqT3lIneha;SJK=u{VHh2#gToT!0*QHBCrTVPjs<8}<}N
z#b~&MqcaR8QPCUr$w&BXbVlyujn8?snrJxe^HXT{UwK*#Bx_23Kky4$%muQt{5SYG
zWKP8ua8g8hW57lkr?E3Kth;5sU}?QXgNV|J`VuxHQy8I<?q&xO^qN9y)2nJGoshiN
zv<%BEX{O{UlfFwXCV~?C7RT=4G?GX1;*V-+-aNLQ(|CfjyeNzCNaT-TCd#%{*Lx(N
zHz-;cX3+aVLT-!Xl}fOW<g~Cm49*C=;A+z9<zvEgs8=2r&IH&Y&37gZq6bE58Bw$X
z{g|doA8GLyxX<Ln>vd_7rB7PD>Us`Ima*&g;^HR~D0Y@|)$-`HX8Ve4qN1Md&=8(N
z>uDW$*|Tbp*osUeNB9a+9{b~i>#|X)f1O%f_{8&30}X7yng~_Y<bwX7xE#ptst1UF
zrKmj(ok^H;1Lj=2i}eTy5uwFwyL{YeRh1(7vQaQBGe?R6IjXvV{25VQDAZ(|MEQ|1
zIGe39lG6t}7?--RUP2cCN@Ie9;~apM6^$B)BuOs9N&XVN{L!jiz&esCmdP7U8`v5b
zvy|&G6xF$myQ-GSUz>_hom^!8NH^@zg%(xMsIU98gRBYMw#ol7??aNj*OG%4$tNsU
z!>$X6-(r4tN-wUIPdNCK4=u&2R(4Xu#`Snq+xtLlhu+IjyY|WStxgKR9as00#NqVg
zhtqw3(5rjfAW;mPib4Dd(szpd|7L^iXD?-lOMw@F;^ydx>RG0kKA`8wjsOp04UoCS
z!dLyv0ZR_ItMngqsTM%n^+$Ek%SzTS^^q6gG9J?p4}UHn$tlubHN=V`;s&xNF+z5Z
zL{=$n%Lsu#WrE6h;Cn(OO?A}5;A@3klkbo@TS03-{bE1uymxe%Ol(9?pV5tI;sKBk
z*ezOO!Gg&1u;VSX)s@)5G;d%c??b`_?Iv)l6BFsTCh+f!s*aDU$aNLvwD>JVI?y|K
zRB}Kg008@Ui`rA=<vzENYF!l`t}5<eX7WNQ9Wa;I;mbqMan4L(CeQB5+SQkZj*v^Z
zhw@dMFRk+~$ZC5Y%90E1CEB@KE2HlmB!WTqhTS?%{(*f?{jWI<)`7&DIS2|CPUn$4
zwO-DWcatMRwWQ4^)%vg`Bwra9;uv06D@j4KaXvjy?Fo9POKc0r6wxmeQJ;M`lZDFJ
zmvc#Nc1W6+oUi;tGLd_k`nd%p?z}vl>k@am$-8}zo6Dd#QP45bS*Ht!#n-}YwpCck
z^vwh@aXW4Fc3NZ%oMiJPsw`;QFBdzUNR;aw&u4A9i1@?0XGPR^vfHs)<%w<2OsqlA
zJ~3>zr!};vZEYi@#^{Hp#3pFiq<ryvO0Y7a3;Q*=c7?*yjK~OW^hW-T5-$-uULweC
zV?KF@D?fb>JFyDLRdQQiwmjgfWH)+j!NU5G)HtK0UPzZ+ZY!}`XgA2!ZXYW7X^VSX
ziyPsk!Yd6g0x$A(i<_d4T0TC9gz+LH2>+Zv*AHN{;J+#0Rk*3Hxbof=%DDK<1tc6Z
z;~nwWrta9h8(Dabkq2Woy<*Mk==$Zw#U;E-=+O$tE+U~{hrRHgsc!vB=2gnU_rI4y
zy1NwVDwWH3ccq}7vq}p_hMu!2XiVwG@eE_<Q^<Ygc9PuxQwo_>tOJ-#j{YCAWPr*S
zCU0TVqy0ujdOub64NwJ!-cM7!;>sY7hvLek0m`L%cdufPQ#Ql9dsR~pKdCH<Qw73g
zD%o?_rd)|rS<2BkrRT9V)%D1z(jTYt_#>maQ_+EYNn}u_4^leAfqN<Zj#55Ix%oR^
zr;zRsl-CC-C*S=+-qfJ?yOb{mDPH*)2j%Z~Ij8csFD1E?fBRAj>K}gUI`0~G?H^9z
zd&ZU3*C@MT&v**DuEa1w*Oi`Ylv{OZzp`ftAM8;w(4qaar{=Kp+!OBfJ(b<z6G({X
zVA<_Fm$LQ7*wavJhj5K<jHU3;l-)y=Q~2y(DI~txuk0Pdo%u`B6<brYxTBz9Di7UJ
yT*El$SWF2HQ}*P$X&9@a=adpl+W7quw(-%%u_;_Xi{pG#VZv4Uw0VRY(*GBnC9W|5

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1995_11.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1995_11.parquet
index fdf5e4944e9c0b0a9b6b6f696797063db5fc760f..a9951d4bfeff3a1ebeacbaf79c3662904bc08fb3 100644
GIT binary patch
delta 3864
zcmb_eYjjgp7Tzb#3CV49o8IQ8N!lb$)4cl5Esau}LUQ}!l~g(uqy#CIFm^#jM(Yc~
zg1jn%aBwaZ=(IQ#s8cCco&p2Il3@^ViNJDLE(CcD=+Z$J;>bXp6Yz(&;}2)d+WVf|
z^L=~goW1wiCp~@c#eMGH9mbSnT{8aTp*j4q6drg5uLS5K=(;G?uTf~p1><`~@{!rz
zN1*u`Xf^{sVay5$lt5B<zyQKfh^QflT7WpcuBIt0PLDVM9LP~Vk6cLA+PiV0ToCIE
z<zfv=Lh|C@04V&+HNmRE*<k|{c^bepV-7K<SuE4+bNxrg@L+FJR^^84H*3Y|q8Jog
zEYPqb)t^Mhr&TK#Ce#J9!%LQs-n4=^sd+!Sm}W)A1e?lbNJU1OM%qf(D}@rkW4gZq
zT!6k1Dq`#73-Bv2{MA>HpEpK)WT(l6#**_UXX@^FJlg++wOKEUF4650Ib#BEfI(aG
z*gTS94x*#vNwYiQMj{TB{v8K~Ahv{bnLE)bQe}A#X~|zKHl!d|ET78O8OdMM>(OxH
zwieBe9+0;jsND)!XSHBe^9f*|H~D|hjx&KTqt6gI3I0L;4DMh;LW9uY;Rnle1;GpK
zw~5z$^1v*?0PG8Df02epvnu>ZnKfHCS+_AN7d;*#_`v<OkPJ~Zw5x#~tVFlW>p9_B
znAaqW?uO5IgRljnWd6`3;}tbbJzjq%dI)(Bq39YJm9N&w?8(jNi8@1{vQ2@FsHW$N
zt}DU2PWj%y62#oFt<g(LGAxtUa(dAx*3~>7)@AD;WQbSNS60O$Q|CuYw>bT>zDR?=
z;l5O=x`Y>icL+KV)L<Bf8SVjP?E`?*aU<T9wJ-x~^8THGljK#osv2C#*KlK*tVXgY
z!!mdN!7y83<tA!$D^0ARfrjDZ>DC?1iTK?tbK-^R7hXrL!sC$S#!s?6@$<4`b@r7S
zuo_a>6KahpiXIm`*KBMxCseq_P}5#D?(yFkavUe41Cf=0bGZX-Boj^;fh{!SLm;p(
zvFv7!z0GLcG|zD%?gFW^8B7fZt}G#)Ccehw7Hb4PE3z8cr<EY0=h#hTg)Ju~g|$Z;
z^>X1MXj6lGwuY98FUZ~?r)-0#jnk?o;B?e3yb}umwxJDq9(k;Sb$Mqv>_Btyn+d_k
zvxV1taVn}6rpH*O4FNZRhD+0MEv)sX{X3ddI_<Z1j>FmB1%s?-JDbM5A1{>O->=<L
zx|}?jSv=Q}hBHwvHkU?F+hP!$uxB`QUi1&2ve><+8ib{Ag05hu`Ige8Ja|~OMJWaa
zc2kzVDqKi|739`tB@~K+dsd5+Q<%8fxwy>ut$xHJxqXFoJ6>!&orH>~E<ji~4LIL=
z0IS%wPQenZvqoA+>W!5{@C)eWXcIIx!B=&F>+r_B@yI@T*8aT6Q^1c#HVEnV88!rX
z<kN(FqA!G0n4pTcR8G=h_l~P)NB6LeX^rOw;peRiP_uIq*mqS+c0kjXWu@M@yw8-v
zEWlNM%w1OFGH?k?^_@k5#~_dd2QxWuip%4oyP$r9qQLP9a>31>jh}FhT(=t>g)B7T
zg~GEK7xAUFDgXmFS;cm#(MWDK%SstY4j%8PB;S$#XkTJckp0@Mo0fZd1iLE@g2qVv
zm`NavjHN*JEH08)Kh<L|E-A##*6YF$@G@~Y)A}jo`0wgN`ScXH^0|o)y4CIJ7xqKQ
zdS2Pk;L_d6g$A}n@P0tva?BoFn~0gl0@cE0K+k0?vj-hV6NOa3t!y_-0~?o2150a(
z{UBLhH?q78g6+?cuv3q|A<a%#;x@0QuDU8a%zk1;sBt^#ai$_S`I~brY9bbw5PBjV
zrw4uq!b>sV>;)y)U@(54BA9@^7|+$HCM|!+{s;4m#=tnaFe*kt&0LY*xYmO-Qx<|y
z9GhD5-JE0qBP%eD)1<?-LBGv_%~Kvx3NOY`!*L-6=>C`L#*hi_LYbzJEO)yn{<fgj
z6KiPm@B3g=IW5hHE9hoh%Q(3vV@X+48qlhx-B#6Q9-h&}f=~wbGg;~B?3{RQ$X6%M
z3bR$_+}(U4_;hcS3jumI)zNX;fpJ8ewK9Kk1zR#$H!O!cm7Ejw1ZL42<=%(hNPm1&
z#xHgJ_<B0|I%yGgeP3j4kXJ=dD!G$N)pW>h-v>$gxaHt{+H`Gk*v97sb^x|C2cY5(
zbXThGA2obDt0fV)wp4gk5LgGiMF>Lq_KLU3*?fxz*}IWlm<z0^k&8C=TcRp>Xmmhh
z{RmvfvreSW6m9o9BW(lhhqNIW<?(D>ymyJB?G-5BQ?WR%trSi;Ve0tk7IpU)wY|G#
zv(N*LI%qzF@DK=^wPZy>KGTt+>qt2_4_fb%qXly#Vg1U7ZMyd}p@!X;X?!^iLMpY~
z7rLBPey-Wb>Sk!LJ6bJYP_3#KhRHzBUSd?Ig?0dUX?iBTI_`Bw)G7}K@!94$6lnzD
zjpt4BkQ@D3>8fyJxxB&C=p$P_CaovcY`3tJtsB4YnvxW01=aB49CF^XO-~+#5ba%U
zV-UADKKcp&_9uMQmEPteTZA&rVUw07Tb%A*;Ex`0b{!#ig)zuTstY41nd~gIsnaSr
zkDA@jMYra(ZYBRH9G1G;*;WDgwF}{B2vniAW{5hW(HW~dZC_xpMWH>`E8{=oew@{=
z)ikSyzGcO`kalWobhhTc1nHMajCT?vb*@t6O1-NTJjVz3_bT=mK;#<6Bck^Bx9p^^
zWE6USrH^OUBu4LvaSgHfW-Cvblq0jl+&uDz&y9M?abHz=bVlm`DJN#6l5lCdCQ%tp
zR2I|cayYph?5(7;G?Vm}X392Ta;9`5I?`7-bUh--pG7qORH}*Dhp($Z!Db{S#Qp<C
zZAm{T6KqL+WBqy=zWFu{R9{QwHRMXi^i0$R>$Ahr*X%zhwXfMpYq>G+30d@Kq?XRt
z{vchoGEi*XB_|)2TPKc)i{6ti`urSzMQQsS$io$S5~;|X>xqy4deTN+ZBqOadF&Ul
z{}=L&O^W$5D;LeIL{w!|<EYB1VyX(NgQzNZ&a6~5e^+y>OQxGQ%mBJ8jU`@sd8N^Y
z2fU2t*>SZSrSS;{9t?U_{j{k~Pd!>tAn-EXu@v2jH)XmV4r)a?{mt5?A%4cDZU6AU
zAT#}p(R8a<rdyg!joYscpma-{{EXRfLqp;2I|JZ#(z%$2_CD$%Mp^f&tpCq%miPeu
z`OWJGARW>uitrsh1CXnGrI!PYTX%IYg^s2CBnrFsNrwVV4!>*P0JBp!CArAhWOQ_k
zj2(4IA(63~I-&!Z^S+n;5K0LRWs;>1k<svUvj5`uUXk9Tp0s<fc=|&fQj{Wm$I-$4
zk;9jy2g{gD?cqxUprz8xGRDa-y*U86eOLNZ8ROt@-}UzQcYB{?EoXS?P8nU~?R^jS
z$Io4j;dcJq)dA2ucki3dEN3*@ckT|rmr1M2nGAl}vLXGQp1df%U(Te<==fzh<A^&J
z4M_(pXsFy3j7CS_`Wu`&P<@Z!@Z2xSRV!q=GkLToDA1qV{m=VcX)Sfe)3^SFTZ^Ov
z6-*|7Yth5~p-Zb{kXBMxQqxPT2jE9GOF|_r0y>VXWE`knnp4SS(4n)EamqfvDIKk3
htg-RS7`j=~RxwuH=3CUyPYX@A;e(wctC%TC{{edCZ$<zB

delta 3991
zcmb_e4|G%28UJqjZkxO&x9Mx%Ymzoe(=>leY4Zw^KTT=!+E6GYg(@grWu>!@9#$q}
z8Jl1Q2Lg`2_dsUGkb?BsY6Yo^ZVD99xruC3<Ty5D9m8qosmNeEPJQlH+&Qk@cD8fO
z`Mz_{&G-Ahckl1}?sxCT0e}C1|M1h+ygfY{@x2)n#MAjAh?$}mU@QH(^@2^?P}&li
z!8I!}KgbIqxedgd<w=nHK(q>$(B}(mU8}jEg0%XY`W6}In0!jmpHLzX;~Y*_TFW@4
z3FRVg+RXr&kxj;U{iM<s3*Y{20NXcKncIAgzBaiHAx`#NbX;3doesIvot(aT&eE3F
zR(DtA<a`m_o!a`cme&0yWtpPHm9-XV+aX4>*F6p560D=Q70%M$t8b2%wrskO_7zrS
z>_|RCFBCe_B>HJ#o#rDgjoBKCxQ5b>a=`QZt^=+>e@?Fhu>wB;v$kwO5z*Q{n?7mt
zp#;5P^A-$b<GGRL&ZJpUe2U+v2qwK~fvnz%#77<WIJ!hT?SB1xIXGJV8yuaANQfS=
zccU}3&hcGD=y8V&0lnn7s99*Hr;1vULOssOUFT15kq;P;i+TYccUrJ6`7)4BTVy+F
zutB)E-(pxSzYfliB9icwUe8zwiW9jqS@r|@bI!X#F1kXt09kH`_>Ij?o%JoZO>sJ;
zo27$SIVBjpUk-!+vYZF10rnbz__XM+_8o$}3f3*rsL#Uiv-Az8wd-fCzfxa7p%;+)
zpoYoFgBo{U@;o>DBIxpPRWyMXNFc93*oD5m_K18i6I)SJ|9L5`4{g`J_?})VYjJf1
zGGogD-J(0H%an_m-_}f!Z-TS|4&JYAT-idl31%g%G&e16ksJ~vEZJH%C7&3Ot-JPh
ztzTJoSwGo`uM0z3U5|V>h#r}J(*Twl%nPO31vddM!X5aTz{+B5ocO*T=W_LBx+YvH
zHVu9ViNNjjiDF0Bn!y%Q;uIQ<(gqvx7@=*}t&AW~M-KjJK*$+dHna<EmX|`VAKyi~
zv)2Vco&h5Uzy`=CCk;kLQG!15ExYv|yFTVu5}k()xIc1r>Z>>pooL?)xJ-DGw3omc
zD-f?8zW_2>Nyr|7jPOQ3D~Y0n@P>=k_q10s@MGw>^G!~1p(++OC@085i!g;QcbU<T
z=tkFs{A}V@JItIseUu~rY9lMDaGbnNUvW+9YB1>*;UctGeiGJRx6jq1JrhegELXFd
zH%>eyU=Lb_59;GfOXXdCxBx}uZ;euA+Y|Kz7zN%aEGNzp_g`4@+24_Z?s>R0)RpBt
z+x_+VFJ{YiH@56{RIjlv(4ySZ1$G$O40yA9oBrmdrIR}7>XH(aMY~H(9Ty_AmL&Z@
zX$1KPaE1}}P5TwCO*?o(*Q-_HGT9X{*SAzM+Ty~dfWA_Z{VSGxg=>Yv*UE6M^|E>P
zTF$-0`4rx0eIXZ3yJJ1V@*9B5ohPu4Z1T#E(Yov~z#WEz5$bT~7xzqTUkTz|<bs6s
zsAYeJh{DE1SP3VP4pS6+{BIt5t2@-yOy>5JTMIkh%EI4sW`cpN(vzojo1TWw-mTT4
zjER5H$}0fZN3d|wfNQ}gf6v@KIr=q-=E7i!P?GNp`q+|L+_)kOIeOM*$l2^gmrUXs
ztV3s*Zo@iJK)UXIoU#n>7s?29;tldDj3<lL3v|@r$ujOUpavmH$ogW$K``_xE%6lj
z`*SA8$wj+#OWEbwq^A(#)~{yIza8X;(M0H0;mKU%%6|8>EdlxsPeH+K58LK}qM>7u
zaQ<G~*5;Fj%b<+}<j~LQ0nh3=jX9X_sL*ZL3hcm+W^&MTDn~A0nIaDp#^5IcV~|yk
zaF#$rT`k1--X*4+0sk->LFq80z2N=|J0&4|2DwC+d3~i1g^bOO^`$N37fysa4uo?5
zinZ16Xz+oEd94`|p~W04a#RM2)kTkcZ$V$DqEEhQX%Q}pvI@I*bPZ%kE3{awALh$8
zAeXX*Cf)68rn?`qf5#eK$jNg?At))7o2{GZJ-&SOOS;>)-~6Zr+n3y?mDh}#Ew9S?
zfRoN2NHN;zuhghAUGMj!1@w^L5?xU%aK_@hYC8*o)oD8;t;enucM_0m!F?`ZEh6`1
zn-bw>Wkm~#+RL65b091|P$x%$R2tN}(r6v!13Sx;F%p<0O`IT{&6^MpMpwY+KX!-x
z@KKGp+w;W>?ewnHU5+eiU?{Mkn@e8PlE+-E9}nLRP`0-Ubi1r;_`x;!`$<~{c8L?v
z;n+)%j6$pe`Xag;M-0CjxH|{;uB{2_Ai4)cha88xL$Sx`E9DO3M<sn9mB?K{!ba8(
z$OrWA<<sX@8J(|#&wAF23?+)|v$n_@0C|PA0;|fCPx#P%ndt|h?kBO08R=>`<ApnJ
zRSy{Y4j9~hYafyOp+kb?IfSP|+)n7aigG@kFQxO}S_hl|Mt@q-RkYB&v%w|3SOQJt
zXo>ZnLP+QgTsUz#Q1@2SN~AmKqM&txwL}ixXa;CA7mwrf$qdxq0U%n>+qi@u9oPC|
ze$258xZ@_eFK9EB7T6sQvdDSxgPtY1?VCY2Ynq4tI{1Wn%M?hkD>*$M3D0G#zY~xA
zPE?DF(#13_*BVdQOks9qF7vOCs3*NWC+UbhADL*RvK<-dBb6?L6cd65a!gQrCv5Je
zZ&ub6?DVE%fVU5|s0q*|(n(Og(BTEJ5;9OZ$&s1&@?nj5MfhSGA5P^K^VWf`Y0sb;
zM7FaOst)b{B}s<wHH3N%bY7KN8i|82h3^aGAFF`&JsLb)b!9*9q7PNgDO#0hW)Bh8
z_7wFc#AN?VX#Hh1m2*R^DViN#t$ppYH<EBK-4^y|J#HPTg7$Q(Rpj&NYvJ<jzUNKq
zabPm-cxuR&H7(=qZ9rqyMLC0+>OW<0(7l=NsJ7F6)j3S{^iXw<CYVW2RDT7X9k9*#
zG0GCI=(O?HQaV&qMz7S^nd$;PS9Fj`igUH>Ri##D{U^D-GHYOJ#H_*Z9npx_j+Q={
zxgB*y4N!#HVQEgVr<J<M{-w+QMfS_`=#o`AYrHhx_-m1zHAiHtU=OV(zv4@<$5xZS
zEyx~(5LCN$|BEmNhV|MwJyhrP?8EB+AoPhyWk$3ngO<n4v^G}K^;@m_e@c|$=W$<>
z-1%h%`fqV&&*ScEn0bH0OvF^fG=nL}6f@0an#ELmv|*+$`C;KpJsN3ujRmB)JE}qe
zmO|D}LDaBBc|J~EsNt=)laFeo6CQ^6xDSnmJeM*>c$ew9U;i6RLxi_>T|KOk(s|7J
zxp&7w&V7suMZHA?i&@ut=N+9Z7gtnF7B$l8eCgo<jiefx8^etK=GiTpac-d-Zo8S=
z-cdJ(`OHoKh3#BTT^U{L)gRw5eZexE@<w@=INU$Za^(5c;wbNzjy%syr}O`Xo^{7k
zd!zgWaow?TZhNn%F0zOkHgXE@M#HIq!aGgF>Nw=-)MACjIQ{83%Z5Lt_Ay^+!=Hj<
z85sV>q*S-U=cPsz-YBuzSittnp98$LywSvFV*%$<4Yj;iJa=uJ<&>Jbzn1rir_|6`
zr&GsLCu@1Pc<R`cv8G4fO?}Kvrbpf#XHrv>>v)ez{bZcw&D8QbzF2&7>(sH~_WUu`
zQ^yx+*!VJY*6jW?b)}9kk#>KY7z^Dy*qHjDj&Tu=v8}S%SirSZYmCu&?fr3<6C3B<
za6QCaB{m!D`uKyN1Kwc(k<G>ecJ`(WGg+S5D9_|Q=$+JUGx=gRte(kxHT$op4$kDA
dqvOk?)~l)a*^~{r^?aV@{F$SkdVWdnKLDKdi6Q_1

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-2001_07.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-2001_07.parquet
index d263b9333338a202d2fc89e30b4107f14c0e15a5..1a9f4dc4dedba9e45420020010220a424803b7fc 100644
GIT binary patch
delta 6727
zcmZWt30zah);|+-Avc5}xo|HKATfbJ0+NyoL_~>_l*l3pVnt<9FcK@asVw4FrMOnD
zOC4<#6-BGI*4p~eT3f4CTeZGcUw!Ufp7zyNt)-jwbxpsC_I<xT;dc)APR@VMojG&n
zoHKK1{hV7Jb6%b3<c}T?<?kFt{rFQQJn$jB44~`Z-A_6#5|g(En8Z()0h5PnQKPh~
z=)yN4-VI<nHZ}yjIS^2TRSkf3qsd2w{YGyY`LENvpd$+kl==VmW`KDO!#~8cVh%7u
z>v~ChalEstEWO6)H1T~J7Rtn0I2{g#@}`@Tay_)#LYT~^9f;A}rh%ca&IbvlDw)Sv
zhWC+iMKL*Nk+1cAVAn{~Mbo+Vc!txRunz#nbHxzmJpiH_nA!HP<<3Mvo0)!yUECxZ
z03GbAB(D@jqaEZ#kujzjx3P8=@L8%_q(@)#cg211UL!Y|gciH=cZccZs#x<yP(`U^
z_IR*Okb?Zk@5A3I(&6(mF%nIB$WVo*U%?1M@KA-)kzSMN5^KO44xnV#<DtY~oQ~4S
zyTvom2V`(bf3%NGE77L=Bk&l7)MFI3s&Gam9xOL3fx>W;tT4TX8K`g-W~KFHB5@gX
zghTESa;&7JO&ko)F?dcSa*8l>I4ns($Y1n>I1E~IK}-vNu;M46=5<~L#;$w&Xf~ir
zVm2%(ff*K&Lc*cG9*Q2Xw=H6A36MG~Oe_c*`ni9Hnv=o0Kq9t)AzYVeO#-tG#0&s|
zdzX!JSR8o{hyEaLlR8bdu4sOgm<NW7YPWz5th!5%+Q5K00lyP1bs4z<u~N|?6voPo
zo=mKbka{xF1gU89$zm#6Kx_{Vg3mDYY^xOe1_#XNdJ_T0wx4HCNsSIckb{d2Ro{#S
zOZZg~1*IA?nFz&@!kVn`><U~Q8&l;NQj=wL;7qlwuqG2lXe?@3jYYKj5gx6=|A=PM
z5mr~L8!2_T&d0ykPbpUU3Tv*_ZA%9So~~3_h9STQ<ubR)VLq&u4k^?eAkC#Y=or~m
zn%MS*%+$pq%MQrP?+5N9`;{Kilj%_pj7M)6JTj9b&-9yaBI0BY(RAcSQKq6MGs{wW
zM6zIk)K%xYz=#JV#djIiK+zbSCZm^@q=nZUN!DMJ=g*?<Rh|cRMchiIr~xD_;6)+0
zv`pT$tE#S!91v3gGX?Gb))T;|SXnr-gH*d>g^Vb4hk3Jt1<?ssBcR(%i8(xuEOo`A
z(`1`V7r9jdmLqGW#&lsnukeW`kMKOX>@p>+P7qUr!!Sr&OL)4N^mT4^II|)!6jPVf
zx{Fnw=+;sw{vFWg3U{6(FSm`%bU#cT^{{&)+D=l+^og@M(F$EXQjZ01(n5`GODNYJ
z#;p&N8SPEj9|dk)OCBrhr_}J=07dSjDdK#<XJfg~$O~n)V^8smbu+badOMW=2B0dP
zfR(J&<PbLNz{&E$uW}|9F{hhB{4u5cDezW6*O2@GdXi~Br2_9*C|?G3ZG5xv29Lju
z_B!FT8tf;)U@$dfcN(rD-r+{{2U#)v68ne&83c&Y-zHPaqi|G$<}z!GPjC(djY3aW
zmFGxe63F}I+30;D8DX~tS7rd(S@-eCgALCz<xhi(XI$trhDD=zwzw5EV@$|m7%Q#x
znPa5PKooCUn)-E$@C)lKhy06V{)p@dyVBMzah8Jk2QWp?{zy)aNC<JF7`@(1z8aC*
z7N6|}#N_KnQI`u-NPQOkYgm-UVRV^{T)Yg=mC1Y_3kQifEC+?rg=6BFc%%dmz8WU;
zi4NSLwXTX3r@`<GA|G^-TU&YjLKr7c!ar)H?sTq-73QeaZ7ORwbe%~)AvbRbF)P5D
z42FmChpgCknnqP!<+iQYrx*7Jq+?Rs-PXP!y6(B#W87LfGZKyMpz$DudLOO`qmCFf
zN}_$HL`)36F3A94EAFAYPf&_8g1;2tg2Ol<&;1Qirvh!AQoWKBe&E0*6eVM`I{$m$
zo=CA6m|t3#pwRqk{CsZf8loPV6u&hOTw9K$Q&}hU8S@2=V9a8jlxIPW6G`RBl;o>n
z;tc?edA+3k3YeDo9tJ#HGxA^cDnn$v$z^dkJYqBciR>6T58Wkc6*}|=DXmCGhe%6>
zIpRW&T?G4DXc`O!a<hf(smMjYklPiRXc6f<Y6Y4`UL7?ET_9hN(y?V)`)puHtCvKN
z*0w!k0i%F%vXDpvg>XGh>JdA?M}{^oLKZ7$SK=$#@h*?W=b?+nG8~i673ImQModz-
zIqhdu;6g9skCnQo1pBbc-H8eZ4Uk-cC`?J>3>jvUh>s`Q2a94X3WP&vIOxKdb4P#b
zAD|Y`g)d6sv<WHa_yFHfwjaX_BEW<tdh+gQLwsCmLA}E0HW}+j;b}VVn*wm|Y3_Ig
z+MDnd5;Ddpe_i8E0-;dl)RUYs`4UqenKh=c?M^aOFCKt@HnMzsITy%bV>>PaGq!aC
zqGqqi0AAqP#aZ~#6x0c>?^bo~0Jrc|oWkR=m^3%C=;r3CF~nVa2k+&<!SPu%jN>=s
zAP;X(1`!V#IN%6BktXYHK}TShh>^j#jo*ynV){uPj)wijUdi>HZ+!Plu;A6X)Q~ua
z<vubRJy{-_159p*JBvnymP$LiLiSfC>p7RKs;ovodN7z`<v2zh8xNXT7EE(>P+0?f
zzGfWxqtb`UiLXi*wZ;vOT$`3Tl&_{km<9zaZRuodRT}z~oT-YjeVZlAbeOm-JI(X6
z`)$9$e^Cd_Vg62g2*VY#0FR|Q+|DqB9EPdevqDTO?3Z0QIvVd|xvQ+bmKdrV!uz`d
z|0P?0vY~oF{EZ@T=eC6K#~3y{&fZVOXm!KA69I1=h|iGksteE>(!WL@dbB)nwA_A_
z)YfE&UMde<Do?#cHq?}&mE=~90ewW2W5=S`$P;5tXbV|4HUU+W{bS!m7fJKDG3YV!
z!8i@tNp6qpE7P>dj1CjtMWSl;vMW_IIEY&&5?gIGnnBjr7WHi$hh5{Oo-FQ}5m644
zC)<;SZ%0ULMEs@--;NjFCepgW?2m;s*i0_BpC-fV79t1vXI+c*L<8PPO2!|d^+vd7
zFZvfb>bam+*D|rGi_-Zrnquk%hC?k()*hr9q~XcmCOi@`WCETY7c+voezVV`-r3F9
z>k#^?yT3OC$>ZAbN6%4rsTfLzPMm`FlGi8NQNQkACmN9IwHFw6Us?Vvs*uIoS-hd!
zGHEcPmTGKB)qJ}y^@^Rk*4taDQ(d$31Gt+@)T8pit^dxf8B7i|Oh@k$adNVJ!d^xM
zSKS`&2{L~2eBt*s@DP48532EseY)*9;Wc66P$7Z*HhC6WMW#QtK-zSW+d<Id$C92s
zj7HY4pj%pD#y1Dz53G3BBCyoW#H(JfpNS(kQ{(I=Hy*E$k6NKSl#jub+n*HrtY|c8
z^UXm=NWf=8^N8_@RH3lbj)1qPJ309#X<-fi{ACo18_A<j*y_fwl`=)C&qtWIhlowV
zO~BOYoT_28^5Qj40lIcFTI@(z{3xw2=}uAAsAOV7@QoQ^{|;zmr9GD90iVdT4WG!o
z4S;i%sp%@dR21nJWmx;$L_Q_1?LsQYGZV~1CA_LAj*p;T{}y0dvea$2R?3+ujc%}L
z4r(<xXG^Jm-lt0D2J)hR$Ne0y%A!ewDYuuRxdWtlMT~t<yQ+?Vgi%RLCt908Yzy{1
z^-UjfNN_yYP043j&F3t)XOQVb<|WOm2|TUKnA7bVCefCmWGb)EE7yXTkk_YZi{1-I
zmAOCn5z~V`bvv)0m%2>J)(r1M!E(P-4TfdhJe{`*>b}CWSgwV|JIzeCqOhi_(3DyC
zqnDPHYUtV9q~1&VHtOQm%JHo*?wc^N-#r^WWw%O#fvTX;)|XT__H7H=x7>acbUDeB
z)cIea$MG(CFrMWFO&Oejx8bT79URIVUqZ!7@O}VHhOUWWbs<cn`N<?J1@44~IAF?j
zFVK>fk7ih0$Ig5`<j6S93failXzn{NSZiQF&G=W5co=A~w$ci@I#66iDfxT)P$H`m
z1zd2Snp|L=Au;@F&X&_^4rTEA<dMYG<Z6qn74~kxb0hI%W?81k$Q_(7wTSp_Fa9<K
zzxx(sX=We7TgT&$1eFXjty{vt#tc`FjRXs~cpM%Sg~t}sN@|1}|Di_$^3nGqBUm*4
zD9sycadU94!g@IkyYxkqDWl6vU(72*8FNz-Ck_Nv^h1V1dxMe@+P{CXNqsO8n`E@G
zuh>Pxo1@#Zl=d-DF$}~p!Srv5qN~L}rb@$)YIWa>7d^Ix2q&iGPFjg+I@Gzi>+#@L
zfAvak#t3EtJkql?OuaPB`;R{tI6EiM2n~(!^Ge~SiTf-qznfc(v0oSE@z9D!#LuMQ
zU*+nRDd58&ZPstTz&QrBdA9*}9g?|RdF((<MILTE<WnSBll$bG@%|XG9`H$-`aiFM
zFIRlj8V3j`4+$3^hKdQuUeHI(3Qo0Pip%L1lYyCt4GO~)=Z1Fbf$GUIzCE#Zc43JT
z91H50+PdpJ%S_t+Nw_!<n4$K`()(Vb&{@R8N*pmDKLbBY%q^*HMlB<_#w|kYa`Rw2
zt+}_GqU^NHaJvjI=D#I=CAIC2eDRHj3t;XOVZQ@-`bW%3+xyuJZ@Ur2@EWX2Et}5F
zv44-)4zU@EBcWh9_k61ETi@Q{)H@5da+#cJ4O-y*B$b}L?z@y~hFAf)v!igzFtBSx
zUC?Ju32X@Zpnf%74!SG&i~%uqqed>Q8%lm^DGoJ9=D)6w7fI37L}Vk6)BpA4*{K;h
z^)WsVnD1io$)H{M$9*M$*N5TRG5Ep(B*X7SfI~PGty#me@lumJR^XCkSkB@T<mai$
zBdVR8v)z^i2CHiX)X&3na>|dV@%`Be<rQ#bBX6|nq2oDVQZ$iDYd8|sh<<_#i^bbi
zdf}j`1-6^Ipk)r1!7{(#=4rYg@IOdXYf@X!3}9x6`Jk%KJ|?O__lfur6w@`x#aF_M
zO+ZWA0(%84RDtLYj#@Q{IxyS8H0Iw_CH1HvABmiVDMe>!0Y1^ZPs;GTM#s@Kl0tuo
z6T3n6{6ZeM#xSl+TTu+LeHpp)a0K7Kk_EJ_d=XSOJrAk`>wZajA?SvkF70EwD2@nP
z7}Y!XAp7sNs{L$rHfN{NXA)RH2bF_5hT+K!dtp|34gC2>Z#OQT2<3N8`|DN5?e1CA
zX2rDy&V%(lbhv;&sk%&WEPgi9s6!{RJPAY#@XbpEz~7`|)4SZQRH(vlpAgFP@XI+Y
zlYDf4XFg({syi+|1%VGD)gMHPj|9)#OlSTwylslKs>Hetyu*R-Pjhj5g*W@&WyJuL
zL1S3o;gO6?b3GA}xg=JCk_pY^-jA);ghcNm*xe84frtj(9yPG)Q_h&cK4{nn!m0yY
zsZwTi+l44SUZcd*3TVZIAK%Fuaj_E|)*QV(Caqu;ULL+v4dN5Qi<;IHpvO@<9eX(x
z49CfD3;!gp)zQ+cs^#O!q}3^HgC|p`YRX4DUQ~Je2M=NVCx%htt}Z_A^k@iVL1z{)
z+v>8!MlgK{Ws^GaB^6Z|4Bjr?<e7<JZ=}bq4gtK?!P#*?D7dNF)PVOZ<AV9NeSJh)
zmUlU=@rE$_!-x^=IwY0_N5o`QIDtg0!DtlGtg%PlJsry5F8=EZz1iKo=0n8Y*Trnh
zkD;RoEJJ@=M#lL~Xb)NJ*P&+elE1%%QIOC5dUS$>u1$yv*9ZPWuZ8PHg@m;4){u&|
zLijRTkJD<@6Id7FTSsQBeF+UEhIM{g&%L#7aJVcj@Lv*Sgzd1k;VZV_Le7>4&VcpI
zpP7Q@3!4kLA-GA*rZ-sIc?eigj|G`51!;k~66ah=;4+-OOz($qj!c<X<w~m&+<n_e
zmOq06op81j0!#j+IrQZ!EV&9QPoOE`0qSW=U^)70d!MuIeMr~I^oM*?VRySF@Pmv}
zD*e9{?k)vp4*o{Lu(B~}nSpo(rS0*GhW@GEFBy`vzJ!~rni<}ox-vE6S!(s|+ySy?
z{iw9Q@ja8J3npt0M06AZ=7lJa#nbHZ1qP)*KywF?=nbP|U&#xwk{(u4#9n{qigsYn
zAP|26g9kgw(;Mc>cew-iNT6lz#^F&LMg;ys<r_wb6A@j44D#Z}v$FTfWNup2KSAbo
z_Cx*1tDPlEw=eJ)s&JFAO$O<O0(%6>+Vr<j=kS1YIOil^Z<-|E_hjH6ar?S!HxEa+
zXO5z(EUhN_yE(CP^;UAYYh&cs3j_a43%~AuVoL*3URxgc3sqbrH@6;4t6XN@wOLW+
zNw47=mT}!WZ|nc>Jnq=jP<A#G1>S<5w}5?%eE2d?u0LO!@c9c(4+G!XM9)tj{+R^u
zCB5e7dGuhhzI*((yGYu7mg}BZy<<`J4n$c(xesM2WrlJH<xt8p$}D9$<uJ+$%HfnT
z<p|13$|}kn<w(j=lzGYm<!H(=lw&EYDaTQcr<}0AdWYt-_q097Lxs)9P%H>1*GUFh
z0b=8690XoMgJSZHkNq``+SrkEPlgH?rcum|kNZ4`Ido1M&nseih>c<ECf55}#${w=
z@)F@xiLm)hsPJ3~?M5*Vc00Gp|My0%o<FyVc~G?KvR~3fMXN6N60KhD9}pC+ULNvb
zuN#N_Q-Y!!hkA)t_4s$(Q`pl>)Unb3X_M9@bZiVAWIa{#KeJ!i=}&0Zrt>R1sqB9Z
zbLa2=an0I934P9Q)*2*pZ~9+q*6R4VH)Rh(m*4VVqgsUJw|a@b{c4TBwnZD|&uP)B
z1lmFmcJ5i|AJ?Ka2t5lsd&#z(3(9!^D=k`O^p<nI0q5p#y088gMF_OjOGQt&UpZBq
z65Z3?TlU4FZGP2Ms^2qJtBR(rUcIk;`Qm-V=Bac9wABl7<I|n@5g!K;wABl7ZsSY$
z5izY4L0i2LE6@Cic!VMZ+UkWk9`Mg>)g}tZ1HEOd7y7$fwR#DCUToE-NX~xx+%#=+
z^x04E^ip$b%UwU3rj0UM?tAZ{8zI8crNNy+CB4wk?EcR^w13$&I&j+RjriqH#L0Vz
zUwR`BzWUQW9VuD&zxdAiP~r6q`avq_rS8tv`(VFqI{hGR^+LR{^5^?Qnn4i)ZS_KY
zeAvHfI(1k2JU*RT<Cz8iZ>DRLf=_-1ZGXGlZ<|4_MW5q??NxjIOJ`_x(X00UN_9Q3
aJ)Yj@KNb`|y|1@Wc=eP0zs}Gei~K)=aD6lY

delta 6835
zcmZX24_s4K`}lL_F6;t3Y!~hY25b)4*gz(`z(f>en=JmwCa9nwDh)@XEd@m>mGEy`
zsddyLDoXjkH?7D&Y5uo-v-j03wX$AQ%l>F#nZ2*2t?!xE?^m(U^W1yReV%)F&h!5{
z*E{AOT08fZdIx{|L@<BhIO@lrD&c_-;$;B)uXfLNm?fqwgQeAW%QRr(FD(Wp7geFj
z((0lM-voJn1mnSRLEy;-zX~iW0IV%17Yh3gSvTtA)Bk{uOej!T|Kmvq({hII&$D7S
zFhcWMN!!pkNA>Wu8iT{g_i1=iCQg9Ua?q7Gej_PQfaYvsEsAd4#UdmXjb%H**q9*k
zND<Q{ElHj9$=QD)Z}Y!^T_sHujpy3p7*2cAHV_#1MNO#Z4G>kp%&~nH;z$Ivk?Dun
zC5@sE(3@Q&71j1BsWpgY(zM}XGKg|Oncy;Y$wYNv!Z^_rJgbY#i<BsNrHp)3#6>@f
zTUeVC_)O&llz_hD|10i>cP4WaiKN(N-4?11QO1})1ZAXBW{ZQoIw^!4{k7~wkrtnk
zi4ka0e_b`2brB;7!qzZ_J*_6uDb|2T4xnJxVk>#RI1TB@k>VC~lH`;ONZX}gShS4F
zX;MHQO!3j5#9Y`sRc0&99C!+N{9HI56DD;Vg!M|CLOM${D2E&>DQQ(*40nvcbLGe(
z!owwCQzCf^zp&_vcnWNpK({Biijjdm^3|dM&H>=~CoqGxFNlK!53903TP7Nyq6FlN
zTjU~rLk<(IQ1qxZ&&}q=LrP_+SP)pz>chb%Jy>cb;&jjjYvV0RV9EorH61|Uy4Xm&
z*`8~+C$wX!!<aWSim#;0*R50O1dLeKoDkFsI{Y<{JEEjc0~a16743pLMrLql;JaAr
z&Omik-`-1671Kd%yT9CYUC(BV*f$_DKF5;?Fs|)Y=9JW67jCiOWP??QW56u0zYL=A
zWej96;bHlOHJS1(+&3n=+MZvNX|Q8fB`d7SK;i0Om8`}rT6_p+D)E*m7Ij_IPSB2$
z+MT=aydzVH)!xFIE4iClu;Y=6F!L}3xIIMXGTKe6RMPxH^>R{EnjO4GkQwYo-e+nS
zV&W>6BYaF|>|&95E96@Hf$JzyxJ7q{TcsptOCw886lH|fWMrDFj!G6TlsapjCm8V!
zN%0qqauAhx!0HmHQOj$NMkX8#u|7d<tvnZMi?}HYsx>4m<f*<oC|B;4H%YpaFYaPE
z#8!-ks{)7vp}7;p)PQhw26b2@y3HJ@U_o?%#Q^AAro{BQs;J5o1aH+ZSt4_L%@z#e
zxBz=W54VFXcWNVA!@zvBU1~@Z2p{GZjc%cn9CsSy8{);3!0L3;=44(>4-DsJ%bDf=
zM2r&&>nc_biE1u|;$H!M5$4LZ=jK$CI@eGXOt!n~(LACYo{%_(6D`oyBXyhc6B?+=
zdm)(X4&~N{$_%zf+!P5eJeX7r@2B`Bk{cM7lb0+m0K7AX+ee-sJ|SfbzeM}E22O8*
z@_PWv(mJeQrAE6jTMG`B7w(2IF^D<c1me3St}H%kHIM%l<#E7i71&OKPG@Ywx>Rf?
z!^;fl6Vg)lQEIvl=>&*QIA*L>dTs`yM8RiZ!#`kdnnLw`ocd!n?@pX!5U9V_lG)|i
zlG|})Pk9#FM!qe#<*WV~%VeNUwHHS1X?U6`p9e~wMvG6WvsO}Podc|(?r%h9-Dqi*
z*ThK~9f~t9OgWw`+-4o+U_DEwjK~TftjK#=;wT02H(-jMyGdRd5g(L~q7&|<kq<_s
zw0@iE0mN9fmDK#0Fsav!e+rE>+YL^mffHnSwoK-Ao4E@K_>%!BlxmELW#W(m-1uUs
z%q!Y)oyPJ^gg6t*E{J@<WUjC0@e85c<wSf_Ep?@FqgY{<Qq`igbVJt}{m~H9>L61X
zSoEM9ir;0$*3&dJYpXWDc&)K`03a=s(&n=C1<`rm{HAcvhcKhixQ-m4k(0^A!$YZI
z2Fy3n_P9h$3=Eg1gRma=C=@D%LTm{<DZmAX@tu&Idx%;SXlfOzrJV2`2Tnnw#|D-4
zmiMIyu^5<NnwO$r>sUNJr<q39QAu%H>fA5*(kQJH`i%Kf90JS|t(0d$h2x2BRI>hJ
zsQ5L2+O$?uei4jIy+Z-dRFC>>qEZ(TXLOqFcDLAsZ;|Dr=A-XPM1>ZuAXyc9w2zFf
zFopjk+a`jo9U6x~L5L}stf|OBx5>GR4Ae$smCMmXq`h)5dXJp1)Uw4I+Z<p>^91r+
zrKWX}84Lo(dPO1)8p3C(Qn%Q73+Y<9a9K<Uy8`dYigUWnUN==5%W%K-<21Rl>JgJv
zE{E-m5}fE|{Ha3wq`(cU+Ljn*r?HY7A_@&joF?7$Y$86AXd5VsF~~26PB~~pnRCZZ
z5Aaio=fIbxaN3B-`q5Z&PfK1Nt&6)ctYBi8!DTc|ti;o_+*<|Uc(tixB(yc+9i(Tp
zA>=i+Ckcc?r6Z2?9b=XJHki1_6t?#0VeFED_(ucFx0Q2uve}r94}lriT!*NzE7E}%
zcy>u9emEI*g7b&xyS9Q$crrH3?KT_LS2L+Eb7L{YUO9nZ=fTc}XVO5955z()-lzu=
z=L{Nnl&_~Qu{ofAcBvQ<h&K2@G{^Om+U*T{MV4=Nu;8f<IN)om0J6BbhPS=~GhUWM
zg^8nC?!3X^&UDk9V077CncSyjY?Te2BO9vp3164Ws)yGkln()8ObF)0adDuIVZk_0
z3sp71Th(L8S5;m#j8s%>BPYASo|C6x4p~)nK{KIXMP4#_x;nM;qs(hF6xAd0^K0-4
zS(x3O;l-IrQlp!@o+-<)8#$AWE^cPOEno9r)&g^wzugwZaK$XZl$02kBNQRKZpN0(
zAmeh|E+>wP!n;}SGHbIC`Pc@z-0A-V^5kT}*n#LOIW#s(>KlaLAs>$|K+B1yCLy@J
z+}~bqYbS*@S;0HY{X5H3c9I1(rN~3xuhF5G$eo&T=oL~q&WP5LwsG-DBpb#ZKrfT(
z@niaWMl(U44PbeaVe<3p1)5>=s$l6V5c2~WTG&L+jqfY#X$mAMJe%AbpCCI|O+$zH
zLM_ox7>itF{)D2wlgH!2@ltmtw`fG9-RRD8XX0<er8Ocxpv2$C2`k9A6Na!i3u!2t
zSZ?binYD{hE;&#;U3#<uH<65okD{l^PY=J2J|;Wd7f>W=tBXeAq@!+RcupOj6B|8(
zdO?%dt=iQ6Yn>LMYuy@85DMYj@T;q-4OLi4|N5zD0ohn@L;CK|>vc%EXdA=s9&UXi
zkl*kY7EkGpnKT4Zd5vjEQD0w^a?wWZZS{-P(yrL}f!yayRh7BmO8CK~9!R<xW}(AG
zHCZ2$`8p$lv-VZ4fD}(&ApCC?^v4HsVJsfMTe}5+w_B)xNQfd|Onw4AOeRcOD7C%G
zb&+qT97na}<wv7KAK5qYaV+0Rx4n(L_h>~(*>de6D+XhZnWTFy$VBSBbI}3vrPqk&
z5XECDg00hrfVZVNIOzsyVGX|YG783%$?(VW>e|}zghX06+eQOZ0woO%t@3Q~Nnl24
zbCh`%YH$e9)k!1ifcPb2X@N@pMOmYiiSdCEGXh<EkocG80-wmU4QFMZnSf1-6s?jk
z6-DZ&3~PIx+<Yvy^;Aj_&(xV7lJLr+SU#K@`$4dFNL@CIGlYp$^Fu^a;F~(fxOE)w
zRqDC^ytr_yND~qL%S@Uqm~vYwnwKWU^P_E>+myBZNJc3wt+zDN$}YfUS`s40Ko4D(
z{93&JZC3pe%WWQPJj=YK-V`rJ_66p2o0>_?OHwdZ*JhS$z_XupPSq40l%uMgpZkbu
zfgqsWaqS(c)2L5L_w)~R3!N&^J<HA1dK#hj27Zj?9$|5(iOC8ptf?+EX4L-Vp@pRi
zdbTyHwvnJ|+SX@6@TE}h^H8zheF;5jvq%ELs-Q40ax}w_J0KP7?-NEf-=<kf+rhY{
z!zMKx1ChoWIdrvARm7a2;?;`!N|p68s>dB6fn2u4))&zPCa>3R)t!6L3CL6TpNEPC
z;5iCRx^@Ch^9PyMX{K>WmRt}wK|?Gs!?mkv@ykasEOxPvZwWeDs-7bowTbF;uVATx
zft3$$LgHbdq1-}i>0*DsY6`dLRe>xD<afdYXmo<5MWVZFN(`Z;A4=yF^!Ct#cWRr(
zG{gwCgO>_G`xu)GT1b3jcI&U>h1KhDT?BTTWEpM)*SSDy7V!%n{C+a-ItZESF^BNN
zhw<!qr3^AGD?%ZUDO2S~fSIEXua3m|g|zC*QsFBJNI+gXPBnm8-4dx@rxG^?#3WEv
zM+64y3>meIfbv%aGp5EMP8<Zv*8aLe+h_$NwB6d?sOn6_Mj7o3Dwf4Exwf^4)}u@!
znh>z-E~AQ_otL1nje&|`AdU&t5avj#cz%j96&Gu?x5PKxd5aKcl5<K{;1B`#adIc)
zz@<95D~B7w)WOILEuo&{0YiCT1zFktxzI2dT7O<4yldo6q*}YVnHaZdBi(LV6p7fI
zjPHi18k50`4{l5teSy;r_G|{6a!BTM=CXs-*}1s%khdq%qVHoh;Wg3XM8N-)sXo{P
z-kiusEU|#FK2t6Zg^D_4E9mo}11gy6G`K`PF!fj$rkm=R+m=I3SydYCB{8)&VTu9l
z3u~DPwbyu-ne_OFa<MreFXkcJWa)!~NNW~fR^T79t?4vohD}dt4bd=y)4CX~$;pK+
z$$+0SM%rkt;&SS?TMs^W16LGVc7$)=>xc$st?3@NeFr%46Gor6HH+c%&O|c2`X@z7
z>MUKU?FM33i%n1*0R@Y>))ehG-sNR9ffuafRGe}ZTIkr9B4+(=eaaTG0&?miwc3?n
z6SV<nG{&<*=-(5cr5d3=#-|UAt{pzAwKku8Fugc9Il{Ux;eTe5G$RoWAcgcgj!c=6
zo;@tu>ju+JEY1$>lYi7(0(fO8ZivP&y@6!-U^v)?y;16AEE^{^x?%(_NrwBd_y{>S
zLqB4GgVVO<C4tW3903#O<1yLgU8(#4wywMaj;`koc?r<58kl5tn8G4Qq6*Q!;G{9Q
zQ<-q%fT#hso7$y$E|<<SzhLJ~?T>gbDQ!;D&_Yx7gNin{)WVb%SqDWWXg?H-pqN%|
zIr!h?nPzqC=gOoW`uMIGCSgj^8S2jUrd?8o=ha${rkiB?AXe-K<)*a5%5X6WxQU_M
zY)!>bi0R9OIQ|jN52#`RZOCbZ@+GwLSIR7#CF1L#Ow{I|9@NM9t~eqfQ#8@B8QI>R
zpj^q0%?h*8=rsu}Ux3nXq}HHoPNzM5S`GaEqem=npA6>Tp84nVme0G3XFd_zDxL@B
zBM+R9nRiJmr6(>wmOlbNYeX85<*+OXL^JSBG&JBZQ*ht6xGO18jrX1uigNL$Y?jf#
zvDcD~=s82c8lS8^NE3<wM1<-@ggEm4sY0jrIlOMFw7SHy89Zgc53sP%R_e(Lm;y5p
z8PXY6+qW!&k*PmTL<#iBh6b@egM;Rhsf~AZWIQSHd|<IeCVCdb&VE4KFY2y6DqvNo
za-;pbp<y?TCvR{=6*7a%CWI#7XB605Kua-Pw}Ulcs{`zo{t345)PhP}$L>&p_*md5
zshJ+lr~`|)l|w-pjz1!tB8rt!(hrr(#*oyN$)aU4wX{iAv~|1EGa%6N@&D0PiZ6EY
zv8P9aKNC7Lf!SP}DNY09Sr|U41D_!aR`!jWk|jn3HquEOlLdLZ`_R6P>2joz11pn|
zlU!bDi@0<;m~Sunvr2Y#XRSJmV)5Q6nm#%w1C4@#ZhV84qfN1YF?ww=iS`*2Uy`1#
z44if`b8WXJ9l797{{L_(8FVwc9wl|Y0g~TB$$DP`I!ccC;v<6-{C|Qy!3oqpiMCNq
z!rKM;;&58yvf%cUa4#WZ`%B15F17p6dD6UQh+LBD{{sXGIkqORm5Wb?63d(r{~55H
z0gn@y+5R=jjcbxUvjY<1a&V(Qi=M4*;o<bEfRHToY)<E&Epg12_%{PJg$^e@mBOM-
zkz4Idtr3d)wz<lmK^?;bJF(FHEd3qk(!pg|dKr{%|CIROA@Yw=&y;wMj;s1us``-H
z9cktsC&-;UpD_D>lu<~b*QepXPg6<ZZ^9T>HYPR0FN9qX!cHG%8q6sk7mzPHO1Rm|
z$I$?4=2TtpQ+aRUx`}mdWvVK!XR>tRWc3^29Yuh{g-EyA-Q@QAEopy)c@}bf?dX{8
zxqepC!%B+SYtLMa8Pr3MyM6(k8|RbEb@M_ty8QP+*hr47D^si(;r|nqujtmU4??oz
z!(}eo6ig!hJNqF$8QWQ+uzUS~f(kp?+o_YjS72kw<<8fF^UM7CWn4a4vtd%m_SydX
zz-{ji+gOJ1*K@<Fho{!)Z_JGeQLQJFyVgg1waEWhSoBqQ><bM@@yRm(pP=Fsvg*Y*
z<3~Pg+OjdM+MQOz)ji8~Ydy{X&*wNR@ot(Qe`#CeP~bZoX*+f{AU;}t;Exj2KhFnO
z&Z=&_`M*fI^?h#ZeA(8;vaN`sgkm3xQi=@4Ad0~hWfWP8ArwO?hEbGL#1z9RDkv%`
zaug#dMpEP{3KXL#MpKNTsG=B4F^*#VUfEXl)pvq6z7;IYJK_6&swQgSssHqe<HKTj
zh>2!v>nD1rk55k*d5LhUMEL$>u<-rN7!WAWU<*J@9F3d6OK1?)??3O$pQedXKXoQp
zI95V=_Meyjo#)<0-^^*61peMe=I^Ba7kxVer2Q9rk=|VHyA~k5xjg9aZ>>1ui)*B$
z6-RoJ_TTi42$1&Q>_s}(>08>UF$%{zg9lqqmHcHJ`!@LAY1E|g`!-P8Uw63f-@ec$
zO`?R})0;Fp$&Ty3$|jAL-*H{`ch)^Oe9tz~uk5+ei`4#|?@W_MAJzWdD)jf)RzKD0
zd#_0o>FYaPqllujUL2>+|ISfOIRrZE#c^+;uWh<U7j<vphF-MQyEpiD&Cu|M^auLf
zw2u?+EDfT{qLg0T%Z_Zm&+QAJp;1QBSuc)x-)w)tF>D51FP-(`SaEsB1CFOChd^h&
zIPP`&cFxe~g?pVZ^`h;2FF+GkH`90bz1N%G^P5lKtVtAle(OzpbFpuHvnD}8@AI2A
z$$ftL?6yxiOB4CW-p}m`67C#FbnleX>z{AE_}}~7zF(W^-s!9t$LqWP_kd&IOuBbE
z>&<cZcaG`zIqvr6_~O+&55BQ0z(HrdzOnA=?;O`Dhd^h&IIceDQ_Rxngsac}+>5s0
zsIP?5B=p`mOOq^leUWeNEKO42el##|f5mrw7L^gb;}$yKzuPyUMWaRgeb2RMBKqiG
d+2ebR&ZA!0bC-Vmubv|8`(*Fd7R~X9{{zK1b^ZVV

diff --git a/orderly/data/test_data/extracted_ord_test_data_trust_labelling/extract_config.json b/orderly/data/test_data/extracted_ord_test_data_trust_labelling/extract_config.json
index da2336a2..51c41860 100644
--- a/orderly/data/test_data/extracted_ord_test_data_trust_labelling/extract_config.json
+++ b/orderly/data/test_data/extracted_ord_test_data_trust_labelling/extract_config.json
@@ -340,6 +340,7 @@
         "[Pd on-carbon]": "[Pd]",
         "[Pd].C": "[Pd]",
         "[Pd]/C": "[Pd]",
+        "[Pd]/[C]]": "[Pd]",
         "[TEA]": "OCCN(CCO)CCO",
         "[Ti-superoxide]": "O=[O-].[Ti]",
         "[[Pd].c1ccc(P(c2ccccc2)c2ccccc2)cc1]": "[Pd]c1ccc(P(c2ccccc2)c2ccccc2)cc1",

From 79028d7415833f338e5c2350bc76989ac8e32a53 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 14:00:48 +0100
Subject: [PATCH 12/29] fix strict mypy

---
 orderly/extract/extractor.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/orderly/extract/extractor.py b/orderly/extract/extractor.py
index 37768638..f383c7f3 100644
--- a/orderly/extract/extractor.py
+++ b/orderly/extract/extractor.py
@@ -1008,7 +1008,7 @@ def move_unresolvable_names_to_end_of_list(
         )
 
         # Add paladium on carbon exception: Delete carbon if Pd exists. Expand exception to other transition metals
-        def contains_transition_metal(agents):
+        def contains_transition_metal(agents: AGENTS) -> bool:
             for agent in agents:
                 if orderly.extract.defaults.has_transition_metal(agent):
                     return True

From 0af60735cd80da5e7290c30d8d5f7e89eae2646c Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 14:20:05 +0100
Subject: [PATCH 13/29] test passes

---
 tests/test_extract.py | 27 ++++++++++++++++++++++++++-
 1 file changed, 26 insertions(+), 1 deletion(-)

diff --git a/tests/test_extract.py b/tests/test_extract.py
index de8f6e65..54a8c4e7 100644
--- a/tests/test_extract.py
+++ b/tests/test_extract.py
@@ -345,7 +345,8 @@ def test_rxn_string_and_is_mapped(
             [
                 "O=C1C[C@@H](C(=O)N[C@@H](Cc2c[nH]cn2)C(=O)N2CCC[C@H]2C(=O)OCc2ccccc2)N1",
             ],
-            [
+            [   "CO",
+                "[C]",
                 "[Pd]",
             ],
             ["O=C1C[C@@H](C(=O)N[C@@H](Cc2c[nH]cn2)C(=O)N2CCC[C@H]2C(=O)O)N1"],
@@ -1101,6 +1102,30 @@ def test_match_yield_with_product(
             [],
             True,
         ],
+        [  # Test that [C] is removed correctly when [Pd] is present
+            "ord_dataset-0b70410902ae4139bd5d334881938f69",
+            262,
+            {},
+            False,
+            [
+                "O=C1C[C@@H](C(=O)N[C@@H](Cc2c[nH]cn2)C(=O)N2CCC[C@H]2C(=O)OCc2ccccc2)N1",
+            ],
+            [
+                "[Pd]",
+            ],
+            [],
+            ["CO"],
+            [],
+            ["O=C1C[C@@H](C(=O)N[C@@H](Cc2c[nH]cn2)C(=O)N2CCC[C@H]2C(=O)O)N1"],
+            [99.7],
+            None,
+            None,
+            "C([O:8][C:9](=[O:32])[C@@H:10]1[CH2:14][CH2:13][CH2:12][N:11]1[C:15](=[O:31])[C@H:16]([CH2:25][C:26]1[N:30]=[CH:29][NH:28][CH:27]=1)[NH:17][C:18]([C@H:20]1[NH:23][C:22](=[O:24])[CH2:21]1)=[O:19])C1C=CC=CC=1>CO.[C].[Pd]>[NH:23]1[C@H:20]([C:18]([NH:17][C@H:16]([C:15]([N:11]2[CH2:12][CH2:13][CH2:14][C@H:10]2[C:9]([OH:32])=[O:8])=[O:31])[CH2:25][C:26]2[N:30]=[CH:29][NH:28][CH:27]=2)=[O:19])[CH2:21][C:22]1=[O:24]",
+            'In 150 ml of methanol was dissolved 782 mg of Nα -[(S)-2-azetidinone-4-carbonyl]-L-histidyl-L-proline benzyl ester (12) and the compound (12) was hydrogenated for 2 hours at room temperature using 156 mg of 10% palladium-carbon as a catalyst. When the catalyst was filtered off and the filtrate was concentrated, 620 mg of Nα -[(S)-2-azetidinone-4-carbonyl]-L-histidyl-L-proline (13) was obtained.',
+            None,
+            [],
+            True,
+        ],
         [
             "ord_dataset-85c00026681b46f89ef8634d2b8618c3",
             3948,

From e0d32c19b9b33ddf8f2f6d4cf44640c327661992 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 16:18:16 +0100
Subject: [PATCH 14/29] added prelim orderly-cond with 100 min freq

---
 Makefile | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/Makefile b/Makefile
index f8708ad4..7676e0d2 100644
--- a/Makefile
+++ b/Makefile
@@ -187,8 +187,8 @@ paper_5 : paper_plot_uspto_no_trust_filtered_min_frequency_of_occurrence_10_100
 
 # 6. clean (final)
 # NB: I changed this one, min_freq=0, train_size=1
-paper_gen_orderly_cond_prelim: #requires: paper_extract_uspto_no_trust
-	python -m orderly.clean --output_path="data/orderly/datasets_$(dataset_version)/orderly_cond_prelim.parquet" --ord_extraction_path="data/orderly/uspto_no_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_no_trust/all_molecule_names.csv" --min_frequency_of_occurrence=0 --map_rare_molecules_to_other=False --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --scramble=True --train_size=1
+paper_gen_orderly_cond_prelim_100: #requires: paper_extract_uspto_no_trust
+	python -m orderly.clean --output_path="data/orderly/datasets_$(dataset_version)/orderly_cond_prelim_100.parquet" --ord_extraction_path="data/orderly/uspto_no_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_no_trust/all_molecule_names.csv" --min_frequency_of_occurrence=100 --map_rare_molecules_to_other=False --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --scramble=True --train_size=1
 
 paper_gen_uspto_no_trust_no_map: #requires: paper_extract_uspto_no_trust
 	python -m orderly.clean --output_path="data/orderly/datasets_$(dataset_version)/orderly_no_trust_no_map.parquet" --ord_extraction_path="data/orderly/uspto_no_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_no_trust/all_molecule_names.csv" --min_frequency_of_occurrence=100 --map_rare_molecules_to_other=False --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --scramble=True --train_size=0.9

From d0172d0a1460d439bc9852c251d6c91c7208597a Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 16:19:05 +0100
Subject: [PATCH 15/29] test case [H][H] reactant

---
 tests/test_extract.py | 26 ++++++++++++++++++++++++++
 1 file changed, 26 insertions(+)

diff --git a/tests/test_extract.py b/tests/test_extract.py
index 54a8c4e7..87ac9f5b 100644
--- a/tests/test_extract.py
+++ b/tests/test_extract.py
@@ -1126,6 +1126,32 @@ def test_match_yield_with_product(
             [],
             True,
         ],
+        [  # Test that [H][H] is correctly classified as a reactant
+            "ord_dataset-0b70410902ae4139bd5d334881938f69",
+            277,
+            {},
+            False,
+            [
+                'O=C([C@@H]1CCCN1C(=O)OCc1ccccc1)N1CCOCC1',
+                "[H][H]"
+            ],
+            [
+                "[Pd]",
+            ],
+            [],
+            ['CCO',],
+            [],
+            ['O=C([C@@H]1CCCN1)N1CCOCC1'],
+            [98.7],
+            None,
+            None,
+            "C(OC([N:11]1[CH2:23][CH2:22][CH2:21][C@H:12]1[C:13]([N:15]1[CH2:20][CH2:19][O:18][CH2:17][CH2:16]1)=[O:14])=O)C1C=CC=CC=1.[H][H]>C(O)C.[C].[Pd]>[NH:11]1[CH2:23][CH2:22][CH2:21][C@H:12]1[C:13]([N:15]1[CH2:20][CH2:19][O:18][CH2:17][CH2:16]1)=[O:14]",
+            'In 100 ml of ethanol was suspended 4.9 g of compound (83) and after adding thereto 250 mg of 10% palladium carbon, the mixture was stirred for 4 hours in a hydrogen stream. After filtering off 10% palladiumcarbon, ethanol was distilled off under reduced pressure from the filtrate to provide 2.8 g of crude N-(L-prolyl)morpholine (84).'
+,
+            None,
+            [],
+            True,
+        ],
         [
             "ord_dataset-85c00026681b46f89ef8634d2b8618c3",
             3948,

From 714eeb7b7bf435047769bc8a062cb49c4baf751f Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 16:19:35 +0100
Subject: [PATCH 16/29] inspect reactions to find test cases

---
 notebooks/inspect_orderly_data.ipynb    |  87 ++++---
 notebooks/orderly_benchmark_stats.ipynb | 332 +++++++++++++++++++++++-
 2 files changed, 378 insertions(+), 41 deletions(-)

diff --git a/notebooks/inspect_orderly_data.ipynb b/notebooks/inspect_orderly_data.ipynb
index 75de1c6d..1722e945 100644
--- a/notebooks/inspect_orderly_data.ipynb
+++ b/notebooks/inspect_orderly_data.ipynb
@@ -73,7 +73,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 13,
+   "execution_count": 4,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -93,7 +93,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 15,
+   "execution_count": 5,
    "metadata": {},
    "outputs": [
     {
@@ -102,7 +102,7 @@
        "'uspto-grants-1986_09'"
       ]
      },
-     "execution_count": 15,
+     "execution_count": 5,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -114,7 +114,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 8,
+   "execution_count": 6,
    "metadata": {},
    "outputs": [
     {
@@ -123,7 +123,7 @@
        "'/Users/danielwigh/projects_local/ORDerly_project/ORDerly/notebooks'"
       ]
      },
-     "execution_count": 8,
+     "execution_count": 6,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -142,7 +142,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 17,
+   "execution_count": 7,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -151,7 +151,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 18,
+   "execution_count": 8,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -161,29 +161,19 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 20,
+   "execution_count": 9,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
-       "262    C([O:8][C:9](=[O:32])[C@@H:10]1[CH2:14][CH2:13...\n",
-       "265    C(OC([N:11]1[CH2:15][CH2:14][CH2:13][C@H:12]1[...\n",
-       "273    C([O:8][C:9](=[O:35])[C@@H:10]1[CH2:14][CH2:13...\n",
-       "275    C([O:8][C:9](=[O:25])[C@H:10]([CH2:19][C:20]1[...\n",
-       "276    C(OC([N:11]1[CH2:15][CH2:14][CH2:13][C@H:12]1[...\n",
        "277    C(OC([N:11]1[CH2:23][CH2:22][CH2:21][C@H:12]1[...\n",
        "279    C(OC([N:11]1[CH2:15][CH2:14][CH2:13][C@H:12]1[...\n",
-       "281    C(OC([N:11]1[CH2:15][CH2:14][CH2:13][C@H:12]1[...\n",
-       "285    C(OC([N:11]1[CH2:15][CH2:14][CH2:13][C@H:12]1[...\n",
-       "368    [ClH:1].C(OC([NH:12][CH:13]([CH2:23][CH2:24][C...\n",
-       "377    [C:1]([O:4][CH2:5][CH2:6][CH2:7][NH:8]C(OCC1C=...\n",
-       "379    [ClH:1].[C:2]([O:5][CH2:6][CH2:7][CH2:8][NH:9]...\n",
        "582    [CH3:1][C:2]1[O:3][C:4](=[O:14])[C:5](=[CH:7][...\n",
        "Name: rxn_str, dtype: object"
       ]
      },
-     "execution_count": 20,
+     "execution_count": 9,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -191,75 +181,98 @@
    "source": [
     "# find df['rxn_str'] which contains [Pd] and [C]\n",
     "\n",
-    "filtered_df = df[df['rxn_str'].str.contains('\\[Pd\\]') & df['rxn_str'].str.contains('\\[C\\]')]\n",
+    "filtered_df = df[df['rxn_str'].str.contains('[H][H]', regex=False) & df['rxn_str'].str.contains('\\[C\\]')]\n",
     "filtered_df['rxn_str']\n"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 22,
+   "execution_count": 10,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
-       "'C([O:8][C:9](=[O:32])[C@@H:10]1[CH2:14][CH2:13][CH2:12][N:11]1[C:15](=[O:31])[C@H:16]([CH2:25][C:26]1[N:30]=[CH:29][NH:28][CH:27]=1)[NH:17][C:18]([C@H:20]1[NH:23][C:22](=[O:24])[CH2:21]1)=[O:19])C1C=CC=CC=1>CO.[C].[Pd]>[NH:23]1[C@H:20]([C:18]([NH:17][C@H:16]([C:15]([N:11]2[CH2:12][CH2:13][CH2:14][C@H:10]2[C:9]([OH:32])=[O:8])=[O:31])[CH2:25][C:26]2[N:30]=[CH:29][NH:28][CH:27]=2)=[O:19])[CH2:21][C:22]1=[O:24]'"
+       "'C(OC([N:11]1[CH2:23][CH2:22][CH2:21][C@H:12]1[C:13]([N:15]1[CH2:20][CH2:19][O:18][CH2:17][CH2:16]1)=[O:14])=O)C1C=CC=CC=1.[H][H]>C(O)C.[C].[Pd]>[NH:11]1[CH2:23][CH2:22][CH2:21][C@H:12]1[C:13]([N:15]1[CH2:20][CH2:19][O:18][CH2:17][CH2:16]1)=[O:14]'"
       ]
      },
-     "execution_count": 22,
+     "execution_count": 10,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "df.loc[262]['rxn_str']"
+    "df.loc[277]['rxn_str']"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 23,
+   "execution_count": 11,
    "metadata": {},
    "outputs": [],
    "source": [
-    "rxn = 'C([O:8][C:9](=[O:32])[C@@H:10]1[CH2:14][CH2:13][CH2:12][N:11]1[C:15](=[O:31])[C@H:16]([CH2:25][C:26]1[N:30]=[CH:29][NH:28][CH:27]=1)[NH:17][C:18]([C@H:20]1[NH:23][C:22](=[O:24])[CH2:21]1)=[O:19])C1C=CC=CC=1>CO.[C].[Pd]>[NH:23]1[C@H:20]([C:18]([NH:17][C@H:16]([C:15]([N:11]2[CH2:12][CH2:13][CH2:14][C@H:10]2[C:9]([OH:32])=[O:8])=[O:31])[CH2:25][C:26]2[N:30]=[CH:29][NH:28][CH:27]=2)=[O:19])[CH2:21][C:22]1=[O:24]'\n",
-    "react,ag,prod=rxn.split('>')"
+    "'C(OC([N:11]1[CH2:23][CH2:22][CH2:21][C@H:12]1[C:13]([N:15]1[CH2:20][CH2:19][O:18][CH2:17][CH2:16]1)=[O:14])=O)C1C=CC=CC=1.[H][H]>C(O)C.[C].[Pd]>[NH:11]1[CH2:23][CH2:22][CH2:21][C@H:12]1[C:13]([N:15]1[CH2:20][CH2:19][O:18][CH2:17][CH2:16]1)=[O:14]'\n"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 28,
+   "execution_count": 19,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
-       "'O=C1C[C@@H](C(=O)N[C@@H](Cc2c[nH]cn2)C(=O)N2CCC[C@H]2C(=O)O)N1'"
+       "98.7"
       ]
      },
-     "execution_count": 28,
+     "execution_count": 19,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "df.loc[262]['product_000']\n"
+    "df.loc[277]['yield_000']"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 18,
    "metadata": {},
-   "outputs": [],
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "99.7"
+      ]
+     },
+     "execution_count": 18,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
    "source": [
-    "O=C1C[C@@H](C(=O)N[C@@H](Cc2c[nH]cn2)C(=O)N2CCC[C@H]2C(=O)O)N1"
+    "df.loc[262]['yield_000']\n"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 13,
    "metadata": {},
-   "outputs": [],
-   "source": []
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'CO.[C].[Pd]'"
+      ]
+     },
+     "execution_count": 13,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "ag"
+   ]
   },
   {
    "attachments": {},
diff --git a/notebooks/orderly_benchmark_stats.ipynb b/notebooks/orderly_benchmark_stats.ipynb
index 0f1099b0..64ea38e7 100644
--- a/notebooks/orderly_benchmark_stats.ipynb
+++ b/notebooks/orderly_benchmark_stats.ipynb
@@ -20,7 +20,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 35,
+   "execution_count": 88,
    "metadata": {},
    "outputs": [
     {
@@ -34,7 +34,7 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 40/40 [00:21<00:00,  1.84it/s]\n"
+      "100%|██████████| 40/40 [00:23<00:00,  1.69it/s]\n"
      ]
     }
    ],
@@ -57,7 +57,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 38,
+   "execution_count": 90,
    "metadata": {},
    "outputs": [
     {
@@ -66,7 +66,7 @@
        "82850"
       ]
      },
-     "execution_count": 38,
+     "execution_count": 90,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -76,6 +76,57 @@
     "# So around 10% of reactions contain a transition metal"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": 94,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "46778 1007 5\n"
+     ]
+    }
+   ],
+   "source": [
+    "molecule = \"Pd\"\n",
+    "count0=df['agent_000'].str.contains(molecule).sum()\n",
+    "count1=df['agent_001'].str.contains(molecule).sum()\n",
+    "count2=df['agent_002'].str.contains(molecule).sum()\n",
+    "print(count0, count1, count2)\n",
+    "count_total = count0+count1+count2"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 97,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "0 0 0\n"
+     ]
+    }
+   ],
+   "source": [
+    "molecule = \"[Pd2+]\"\n",
+    "count0=df['agent_000'].str.contains(molecule, regex=False).sum()\n",
+    "count1=df['agent_001'].str.contains(molecule, regex=False).sum()\n",
+    "count2=df['agent_002'].str.contains(molecule, regex=False).sum()\n",
+    "print(count0, count1, count2)\n",
+    "count_total = count0+count1+count2"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  },
   {
    "cell_type": "code",
    "execution_count": 39,
@@ -194,6 +245,88 @@
     "trans_metal"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Transition metal on Carbon"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 82,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Index(['agent_000', 'agent_001', 'agent_002', 'date_of_experiment',\n",
+       "       'extracted_from_file', 'grant_date', 'is_mapped', 'procedure_details',\n",
+       "       'product_000', 'product_001', 'reactant_000', 'reactant_001',\n",
+       "       'reactant_002', 'rxn_str', 'rxn_time', 'solvent_000', 'solvent_001',\n",
+       "       'solvent_002', 'temperature', 'yield_000', 'yield_001'],\n",
+       "      dtype='object')"
+      ]
+     },
+     "execution_count": 82,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "path = \"/Users/danielwigh/projects_local/ORDerly_project/orderly-benchmarks/orderly_forward_train.parquet\"\n",
+    "df = pd.read_parquet(path)\n",
+    "df.columns"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 83,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "81 2445 223\n",
+      "There are 2749 reactions with a carbon agent. This represents 0.00329967639405026 of all rxn\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Check if there are any agents == \"[C]\"\n",
+    "count0=df['agent_000'].str.contains(\"\\[C\\]\").sum()\n",
+    "count1=df['agent_001'].str.contains(\"\\[C\\]\").sum()\n",
+    "count2=df['agent_002'].str.contains(\"\\[C\\]\").sum()\n",
+    "print(count0, count1, count2)\n",
+    "count_total = count0+count1+count2\n",
+    "print(f\"There are {count_total} reactions with a carbon agent. This represents {count_total/len(df)} of all rxn\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 84,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "632 7244 1862\n",
+      "There are 9738 reactions with a hydrogen reactant (mislabelled). This represents 0.011688704519920491 of all rxn\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Check if there are any agents == \"[H][H]\"\n",
+    "count0=df['agent_000'].str.contains(\"\\[H\\]\\[H\\]\").sum()\n",
+    "count1=df['agent_001'].str.contains(\"\\[H\\]\\[H\\]\").sum()\n",
+    "count2=df['agent_002'].str.contains(\"\\[H\\]\\[H\\]\").sum()\n",
+    "print(count0, count1, count2)\n",
+    "count_total = count0+count1+count2\n",
+    "print(f\"There are {count_total} reactions with a hydrogen reactant (mislabelled). This represents {count_total/len(df)} of all rxn\")"
+   ]
+  },
   {
    "attachments": {},
    "cell_type": "markdown",
@@ -550,6 +683,188 @@
     "df['rxn_str']"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# ORDerly-cond-prelim "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 76,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import pandas as pd\n",
+    "from tqdm import tqdm"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 77,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "path = '/Users/danielwigh/projects_local/ORDerly_project/ORDerly/data/orderly/datasets_v6/orderly_cond_trial.parquet'\n",
+    "df = pd.read_parquet(path)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 78,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "753196"
+      ]
+     },
+     "execution_count": 78,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "len(df)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 79,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "100%|██████████| 40/40 [00:20<00:00,  2.00it/s]\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Assume reactions only have max 1 transition metal\n",
+    "transition_metals = [\n",
+    "    'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn',\n",
+    "    'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd',\n",
+    "    'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg',\n",
+    "    'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn'\n",
+    "]\n",
+    "trans_metal = {}\n",
+    "for tm in tqdm(transition_metals):\n",
+    "    count = df['rxn_str'].str.contains(tm).sum()\n",
+    "    trans_metal[tm] = count"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 80,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "77310"
+      ]
+     },
+     "execution_count": 80,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "sum(trans_metal.values())\n",
+    "# So around 10% of reactions contain a transition metal"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 81,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "45689"
+      ]
+     },
+     "execution_count": 81,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "trans_metal[\"Pd\"]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 75,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "['[CH2:1]([O:3][C:4](=[O:36])[C:5]1[CH:10]=[CH:9][C:8]([N:11]2[CH:15]=[C:14]([C:16]3[CH:21]=[CH:20][CH:19]=[CH:18][C:17]=3[O:22]CC3C=CC=CC=3)[C:13]([C:30]#[N:31])=[CH:12]2)=[CH:7][C:6]=1[O:32][CH2:33][O:34][CH3:35])[CH3:2].C(OCC)(=O)C>[C].[Pd].CO>[CH2:1]([O:3][C:4](=[O:36])[C:5]1[CH:10]=[CH:9][C:8]([N:11]2[CH:15]=[C:14]([C:16]3[CH:21]=[CH:20][CH:19]=[CH:18][C:17]=3[OH:22])[C:13]([C:30]#[N:31])=[CH:12]2)=[CH:7][C:6]=1[O:32][CH2:33][O:34][CH3:35])[CH3:2]']\n"
+     ]
+    }
+   ],
+   "source": [
+    "df3 = df[df['reactant_000']=='CCOC(=O)c1ccc(-n2cc(C#N)c(-c3ccccc3OCc3ccccc3)c2)cc1OCOC']\n",
+    "print(df3['rxn_str'].values)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "['[CH2:1]([O:3][C:4](=[O:36])[C:5]1[CH:10]=[CH:9][C:8]([N:11]2[CH:15]=[C:14]([C:16]3[CH:21]=[CH:20][CH:19]=[CH:18][C:17]=3[O:22]CC3C=CC=CC=3)[C:13]([C:30]#[N:31])=[CH:12]2)=[CH:7][C:6]=1[O:32][CH2:33][O:34][CH3:35])[CH3:2].C(OCC)(=O)C>[C].[Pd].CO>[CH2:1]([O:3][C:4](=[O:36])[C:5]1[CH:10]=[CH:9][C:8]([N:11]2[CH:15]=[C:14]([C:16]3[CH:21]=[CH:20][CH:19]=[CH:18][C:17]=3[OH:22])[C:13]([C:30]#[N:31])=[CH:12]2)=[CH:7][C:6]=1[O:32][CH2:33][O:34][CH3:35])[CH3:2]']"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 70,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df2 = df[df['rxn_str'].str.contains(\"O=O\", regex=False, na=False)]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 71,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "339       [Cl-:1].[K+].O=O.[C:5]1(=[O:12])[O:11][C:9](=[...\n",
+       "881       [CH2:1]=[CH:2][CH2:3][CH2:4][CH2:5][CH2:6][CH2...\n",
+       "937       [C:1]([C:5]1[CH:10]=[CH:9][CH:8]=[CH:7][CH:6]=...\n",
+       "1015      [C:1]1([OH:7])[CH:6]=[CH:5][CH:4]=[CH:3][CH:2]...\n",
+       "2850      [CH2:1]([N:3]([CH2:34][CH3:35])[CH2:4][CH2:5][...\n",
+       "                                ...                        \n",
+       "730862    [CH2:1]([C@@H:4]1[S:9](=[O:11])(=[O:10])[C:8](...\n",
+       "730882    [CH2:1]([CH:4]1[C@@:9]([CH2:20][F:21])([C:10]2...\n",
+       "732855    [O:1]=[O+][O-].[Si:4]([O:11][CH:12]([C:14]1[O:...\n",
+       "735590    O=O.[CH2:3]([N:10]1[CH2:14][C:13]([C:15]2[CH:2...\n",
+       "737280    O=O.[C:3]([C:7]1[CH:8]=[C:9](O)[C:10](=[CH:12]...\n",
+       "Name: rxn_str, Length: 348, dtype: object"
+      ]
+     },
+     "execution_count": 71,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df2['rxn_str']"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -563,6 +878,15 @@
    "metadata": {},
    "outputs": [],
    "source": []
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "'ClC1C=CC=C(C(OO)=[O:9])C=1.[CH3:12][S:13][CH2:14][C:15]1[C:24]([C:25]([O:27][CH3:28])=[O:26])=[N+:23]([O-:29])[C:22]2[C:17](=[CH:18][CH:19]=[CH:20][CH:21]=2)[N+:16]=1[O-:30].C(OCC)C>C(Cl)(Cl)Cl>[CH3:12][S:13]([CH2:14][C:15]1[C:24]([C:25]([O:27][CH3:28])=[O:26])=[N+:23]([O-:29])[C:22]2[C:17](=[CH:18][CH:19]=[CH:20][CH:21]=2)[N+:16]=1[O-:30])=[O:9]'"
+   ]
   }
  ],
  "metadata": {

From a25b5af635a5cd2f71d4a5c8788559cbd9bed34b Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Wed, 17 Jan 2024 16:20:08 +0100
Subject: [PATCH 17/29] make black

---
 tests/test_extract.py | 19 +++++++++----------
 1 file changed, 9 insertions(+), 10 deletions(-)

diff --git a/tests/test_extract.py b/tests/test_extract.py
index 87ac9f5b..fcd6cb21 100644
--- a/tests/test_extract.py
+++ b/tests/test_extract.py
@@ -345,7 +345,8 @@ def test_rxn_string_and_is_mapped(
             [
                 "O=C1C[C@@H](C(=O)N[C@@H](Cc2c[nH]cn2)C(=O)N2CCC[C@H]2C(=O)OCc2ccccc2)N1",
             ],
-            [   "CO",
+            [
+                "CO",
                 "[C]",
                 "[Pd]",
             ],
@@ -1121,7 +1122,7 @@ def test_match_yield_with_product(
             None,
             None,
             "C([O:8][C:9](=[O:32])[C@@H:10]1[CH2:14][CH2:13][CH2:12][N:11]1[C:15](=[O:31])[C@H:16]([CH2:25][C:26]1[N:30]=[CH:29][NH:28][CH:27]=1)[NH:17][C:18]([C@H:20]1[NH:23][C:22](=[O:24])[CH2:21]1)=[O:19])C1C=CC=CC=1>CO.[C].[Pd]>[NH:23]1[C@H:20]([C:18]([NH:17][C@H:16]([C:15]([N:11]2[CH2:12][CH2:13][CH2:14][C@H:10]2[C:9]([OH:32])=[O:8])=[O:31])[CH2:25][C:26]2[N:30]=[CH:29][NH:28][CH:27]=2)=[O:19])[CH2:21][C:22]1=[O:24]",
-            'In 150 ml of methanol was dissolved 782 mg of Nα -[(S)-2-azetidinone-4-carbonyl]-L-histidyl-L-proline benzyl ester (12) and the compound (12) was hydrogenated for 2 hours at room temperature using 156 mg of 10% palladium-carbon as a catalyst. When the catalyst was filtered off and the filtrate was concentrated, 620 mg of Nα -[(S)-2-azetidinone-4-carbonyl]-L-histidyl-L-proline (13) was obtained.',
+            "In 150 ml of methanol was dissolved 782 mg of Nα -[(S)-2-azetidinone-4-carbonyl]-L-histidyl-L-proline benzyl ester (12) and the compound (12) was hydrogenated for 2 hours at room temperature using 156 mg of 10% palladium-carbon as a catalyst. When the catalyst was filtered off and the filtrate was concentrated, 620 mg of Nα -[(S)-2-azetidinone-4-carbonyl]-L-histidyl-L-proline (13) was obtained.",
             None,
             [],
             True,
@@ -1131,23 +1132,21 @@ def test_match_yield_with_product(
             277,
             {},
             False,
-            [
-                'O=C([C@@H]1CCCN1C(=O)OCc1ccccc1)N1CCOCC1',
-                "[H][H]"
-            ],
+            ["O=C([C@@H]1CCCN1C(=O)OCc1ccccc1)N1CCOCC1", "[H][H]"],
             [
                 "[Pd]",
             ],
             [],
-            ['CCO',],
+            [
+                "CCO",
+            ],
             [],
-            ['O=C([C@@H]1CCCN1)N1CCOCC1'],
+            ["O=C([C@@H]1CCCN1)N1CCOCC1"],
             [98.7],
             None,
             None,
             "C(OC([N:11]1[CH2:23][CH2:22][CH2:21][C@H:12]1[C:13]([N:15]1[CH2:20][CH2:19][O:18][CH2:17][CH2:16]1)=[O:14])=O)C1C=CC=CC=1.[H][H]>C(O)C.[C].[Pd]>[NH:11]1[CH2:23][CH2:22][CH2:21][C@H:12]1[C:13]([N:15]1[CH2:20][CH2:19][O:18][CH2:17][CH2:16]1)=[O:14]",
-            'In 100 ml of ethanol was suspended 4.9 g of compound (83) and after adding thereto 250 mg of 10% palladium carbon, the mixture was stirred for 4 hours in a hydrogen stream. After filtering off 10% palladiumcarbon, ethanol was distilled off under reduced pressure from the filtrate to provide 2.8 g of crude N-(L-prolyl)morpholine (84).'
-,
+            "In 100 ml of ethanol was suspended 4.9 g of compound (83) and after adding thereto 250 mg of 10% palladium carbon, the mixture was stirred for 4 hours in a hydrogen stream. After filtering off 10% palladiumcarbon, ethanol was distilled off under reduced pressure from the filtrate to provide 2.8 g of crude N-(L-prolyl)morpholine (84).",
             None,
             [],
             True,

From b80ccb6efe2d1192b088f37af39b34a6c54b5a4d Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Sat, 20 Jan 2024 19:02:44 +0000
Subject: [PATCH 18/29] H2 always reactant, never an agent

---
 orderly/extract/extractor.py | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/orderly/extract/extractor.py b/orderly/extract/extractor.py
index f383c7f3..721f723b 100644
--- a/orderly/extract/extractor.py
+++ b/orderly/extract/extractor.py
@@ -253,7 +253,7 @@ def extract_info_from_rxn_str(
                         and (  # any(generator)
                             r_clean not in products_from_rxn_without_mapping
                         )
-                    ) or (r_clean == "[H][H]"):
+                    ):
                         reactants.append(r_clean)
                     else:
                         cleaned_agents.append(r_clean)
@@ -285,9 +285,17 @@ def extract_info_from_rxn_str(
             cleaned_agents = [
                 a for a in cleaned_agents if a not in reactants and a not in products
             ]
+            # We also add an exception for [H][H], this should always be a reactant
+            # Move [H][H] from agents to reactants
+            if "[H][H]" in cleaned_agents:
+                cleaned_agents.remove("[H][H]")
+                reactants.append("[H][H]")
         else:
             reactants = reactants_from_rxn_without_mapping
             products = products_from_rxn_without_mapping
+            if "[H][H]" in cleaned_agents:
+                cleaned_agents.remove("[H][H]")
+                reactants.append("[H][H]")
 
         return (
             sorted(list(set(reactants))),

From a326340c8ca8707cc2ff9ef548180f44aa76791e Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Sat, 20 Jan 2024 19:11:23 +0000
Subject: [PATCH 19/29] del C if charcoal in procedure details

---
 orderly/extract/extractor.py | 45 +++++++++++++++++++++++++-----------
 1 file changed, 32 insertions(+), 13 deletions(-)

diff --git a/orderly/extract/extractor.py b/orderly/extract/extractor.py
index 721f723b..1acd3761 100644
--- a/orderly/extract/extractor.py
+++ b/orderly/extract/extractor.py
@@ -473,11 +473,15 @@ def rxn_time_extractor(rxn: ord_reaction_pb2.Reaction) -> Optional[RXN_TIME]:
         else:
             return None  # no time found
 
-    @staticmethod
+    @sta
+    #NB this needs to be before the removal of [C] and Cticmethod
     def procedure_details_extractor(
         rxn: ord_reaction_pb2.Reaction,
-    ) -> str:  # TODO check does it return empty string or none
-        procedure_details = rxn.notes.procedure_details
+    ) ->
+    #NB this needs to be before the removal of [C] and C str:  # TODO check does it re
+    #NB this needs to be before the removal of [C] and Cturn empty string or none
+        procedure_d
+        #NB this needs to be before the removal of [C] and Cetails = rxn.notes.procedure_details
         return str(procedure_details)
 
     @staticmethod
@@ -1014,6 +1018,12 @@ def move_unresolvable_names_to_end_of_list(
         catalysts, _ = move_unresolvable_names_to_end_of_list(
             catalysts, rxn_non_smiles_names_set
         )
+        
+        #NB this needs to be before the removal of [C] and C
+        procedure_details = OrdExtractor.procedure_details_extractor(rxn)
+        date_of_experiment = OrdExtractor.date_of_experiment_extractor(rxn)
+        rxn_time = OrdExtractor.rxn_time_extractor(rxn)
+        temperature = OrdExtractor.temperature_extractor(rxn)
 
         # Add paladium on carbon exception: Delete carbon if Pd exists. Expand exception to other transition metals
         def contains_transition_metal(agents: AGENTS) -> bool:
@@ -1021,9 +1031,14 @@ def contains_transition_metal(agents: AGENTS) -> bool:
                 if orderly.extract.defaults.has_transition_metal(agent):
                     return True
             return False
+        
+        def contains_charcoal(procedure_details) -> bool:
+            if "charcoal" in procedure_details.lower():
+                return True
+            return False
 
         # Check if any agent contains a transition metal
-        if contains_transition_metal(agents):
+        if (contains_transition_metal(agents) or contains_charcoal(procedure_details)):
             # Remove "[C]" and "C" from agents
             agents = [a for a in agents if a not in ["[C]", "C"]]
 
@@ -1031,10 +1046,7 @@ def contains_transition_metal(agents: AGENTS) -> bool:
             _yields = [None] * len(products)
         yields = _yields
 
-        procedure_details = OrdExtractor.procedure_details_extractor(rxn)
-        date_of_experiment = OrdExtractor.date_of_experiment_extractor(rxn)
-        rxn_time = OrdExtractor.rxn_time_extractor(rxn)
-        temperature = OrdExtractor.temperature_extractor(rxn)
+        
         if ice_present and (
             temperature is None
         ):  # We trust the labelled temperature more, but if there is no labelled temperature, and they added ice, we should set the temperature to 0C
@@ -1058,7 +1070,8 @@ def contains_transition_metal(agents: AGENTS) -> bool:
             yields,
             temperature,
             rxn_time,
-            rxn_str,
+            
+            #NB this needs to be before the removal of [C] and Crxn_str,
             procedure_details,
             date_of_experiment,
             is_mapped,
@@ -1101,7 +1114,8 @@ def build_rxn_lists(
             "temperature": [],
             "rxn_time": [],
             "product": [],
-            "yield": [],
+            "
+            #NB this needs to be before the removal of [C] and Cyield": [],
             "procedure_details": [],
             "date_of_experiment": [],
             "is_mapped": [],
@@ -1129,7 +1143,8 @@ def build_rxn_lists(
                 yields,
                 temperature,
                 rxn_time,
-                rxn_str,
+                
+                #NB this needs to be before the removal of [C] and Crxn_str,
                 procedure_details,
                 date_of_experiment,
                 is_mapped,
@@ -1147,7 +1162,9 @@ def build_rxn_lists(
             rxn_lists["temperature"].append(temperature)
             rxn_lists["rxn_time"].append(rxn_time)
             rxn_lists["product"].append(products)
-            rxn_lists["yield"].append(yields)
+            rxn_lists["
+            #NB this needs to be before the removal of [C] and Cyield"].append(yields)
+            #NB this needs to be before the removal of [C] and C
             rxn_lists["procedure_details"].append(procedure_details)
             rxn_lists["date_of_experiment"].append(date_of_experiment)
             rxn_lists["is_mapped"].append(is_mapped)
@@ -1239,7 +1256,9 @@ def build_full_df(
             )
         )
         dfs.append(
-            OrdExtractor._to_dataframe(
+            OrdExtractor._to
+            #NB this needs to be before the removal of [C] and C_dataframe(
+                #NB this needs to be before the removal of [C] and C
                 data_lists["procedure_details"], base_string=["procedure_details"]
             )
             .fillna("<missing>")

From f73e2e071d685f02e1435b2201ee90b6cb56f196 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Sat, 20 Jan 2024 21:19:28 +0000
Subject: [PATCH 20/29] scramble reactant, prod, solv, but not agent

---
 orderly/clean/cleaner.py | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/orderly/clean/cleaner.py b/orderly/clean/cleaner.py
index a61c5383..a53dc09f 100644
--- a/orderly/clean/cleaner.py
+++ b/orderly/clean/cleaner.py
@@ -475,10 +475,12 @@ def _move_none_to_after_data(
     @staticmethod
     def _scramble(
         df: pd.DataFrame,
-        components: Tuple[str, ...] = ("agent", "solvent", "catalyst", "reagent"),
+        components: Tuple[str, ...] = ("reactant", "product", "solvent", "catalyst", "reagent"),
         seed: int = 42,
     ) -> pd.DataFrame:
-        """Scrambles the order of the reactants (ie between reactant_001, reactant_002, etc). Ordering of prodcuts, agents, solvents, reagents, and catalysts will also be scrambled. This is done to prevent the model from learning the order of the molecules, which is not important for the reaction prediction task. It only done at the very end because scrambling can be non-deterministic between versions/operating systems, so it would be difficult to debug if done earlier in the pipeline."""
+        """Scrambles the order of the reactants (ie between reactant_001, reactant_002, etc). Ordering of products, solvents, reagents, and catalysts will also be scrambled. This is done to prevent the model from learning the order of the molecules, which is not important for the reaction prediction task. It only done at the very end because scrambling can be non-deterministic between versions/operating systems, so it would be difficult to debug if done earlier in the pipeline.
+        
+        NB: agents not scrambled by default, since we need their ordering to let transition metals be first."""
         list_of_dfs = []
         all_component_cols = []
         np.random.seed(seed)
@@ -823,8 +825,8 @@ def _get_dataframe(self) -> pd.DataFrame:
         df.reset_index(inplace=True, drop=True)
 
         if self.scramble:
-            LOG.info(f"Scrambling the order of the components")
-            components = ("agent", "solvent", "reagent", "catalyst")
+            components = ("reactant", "product", "solvent", "catalyst", "reagent")
+            LOG.info(f"Scrambling the order of the components: {components=}")
             df = Cleaner._scramble(df, components)
             df = Cleaner._move_none_to_after_data(df, components)
             df = Cleaner._replace_None_with_NA(df, components)
@@ -1035,7 +1037,7 @@ def get_matching_indices(
     "--scramble",
     type=bool,
     default=True,
-    help="If True, the order of the reactants be scrambled (ie between reactant_001, reactant_002, etc). Ordering of prodcuts, agents, solvents, reagents, and catalysts will also be scrambled. Will also scramble the reaction indices. This is done to prevent the model from learning the order of the molecules, which is not important for the reaction prediction task. It only done at the very end because scrambling can be non-deterministic between versions/operating systems, so it would be difficult to debug if done earlier in the pipeline.",
+    help="If True, the order of the reactants be scrambled (ie between reactant_001, reactant_002, etc). Ordering of prodcuts, agents, solvents, reagents, and catalysts will also be scrambled. Will also scramble the reaction indices. This is done to prevent the model from learning the order of the molecules, which is not important for the reaction prediction task. It only done at the very end because scrambling can be non-deterministic between versions/operating systems, so it would be difficult to debug if done earlier in the pipeline.NB: agents not scrambled by default, since we need their ordering to let transition metals be first.",
 )
 @click.option(
     "--train_size",

From eda76fd3898b13522de38fe2681afd3de8f8bf3b Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Sat, 20 Jan 2024 21:19:45 +0000
Subject: [PATCH 21/29] added comment on sorting

---
 orderly/extract/extractor.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/orderly/extract/extractor.py b/orderly/extract/extractor.py
index 1acd3761..6d4a6115 100644
--- a/orderly/extract/extractor.py
+++ b/orderly/extract/extractor.py
@@ -297,6 +297,7 @@ def extract_info_from_rxn_str(
                 cleaned_agents.remove("[H][H]")
                 reactants.append("[H][H]")
 
+        # NB: We need to sort the lists, otherwise the order is non-deterministic, and the tests won't pass.
         return (
             sorted(list(set(reactants))),
             sorted(list(set(cleaned_agents))),

From 0075db0f64d9082bb03b6d93fb9dc9cb2e81ed59 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Sat, 20 Jan 2024 21:20:25 +0000
Subject: [PATCH 22/29] added no_min_freq as orderly-cond dataset

---
 Makefile | 43 ++++++++++++++++++++++++++++++-------------
 1 file changed, 30 insertions(+), 13 deletions(-)

diff --git a/Makefile b/Makefile
index 7676e0d2..946b42de 100644
--- a/Makefile
+++ b/Makefile
@@ -123,10 +123,11 @@ run_python_310:
 # 3. Plot histograms of the number of non-empty columns of each type (reactants, products, solvents, agents)
 # 4. Run a cleaning with decided upon number of columns to keep
 # 5. Plot histogram showing dataset size as a function of min_frequency_of_occurrence (can probably use the min_frequency code from the cleaner within the plotter)
-# 6. Generate the four datasets we need for the paper (split into train and test set)
+# 6. Generate the six condition prediction datasets we need for the paper (split into train and test set)
 # 7. Plot histograms with the occurrence of the most common reactants, products, solvents, agents
-# 8. Generate fingerprints for each dataset
-# 9. Train & evaluate a model on each dataset
+# 8. Generate the final benchmark datasets (split into train and test set)
+# 9. Generate fingerprints for each dataset
+# 10. Train & evaluate a model on each dataset
 
 # 1. Extract 
 
@@ -185,11 +186,7 @@ paper_plot_uspto_with_trust_filtered_min_frequency_of_occurrence_100_1000:
 paper_5 : paper_plot_uspto_no_trust_filtered_min_frequency_of_occurrence_10_100 paper_plot_uspto_no_trust_filtered_min_frequency_of_occurrence_100_1000 paper_plot_uspto_with_trust_filtered_min_frequency_of_occurrence_10_100 paper_plot_uspto_with_trust_filtered_min_frequency_of_occurrence_100_1000
 
 
-# 6. clean (final)
-# NB: I changed this one, min_freq=0, train_size=1
-paper_gen_orderly_cond_prelim_100: #requires: paper_extract_uspto_no_trust
-	python -m orderly.clean --output_path="data/orderly/datasets_$(dataset_version)/orderly_cond_prelim_100.parquet" --ord_extraction_path="data/orderly/uspto_no_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_no_trust/all_molecule_names.csv" --min_frequency_of_occurrence=100 --map_rare_molecules_to_other=False --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --scramble=True --train_size=1
-
+# 6. ORDerly-condition alternatives cleaning
 paper_gen_uspto_no_trust_no_map: #requires: paper_extract_uspto_no_trust
 	python -m orderly.clean --output_path="data/orderly/datasets_$(dataset_version)/orderly_no_trust_no_map.parquet" --ord_extraction_path="data/orderly/uspto_no_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_no_trust/all_molecule_names.csv" --min_frequency_of_occurrence=100 --map_rare_molecules_to_other=False --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --scramble=True --train_size=0.9
 
@@ -202,18 +199,38 @@ paper_gen_uspto_with_trust_with_map: #requires: paper_extract_uspto_with_trust
 paper_gen_uspto_with_trust_no_map: #requires: paper_extract_uspto_with_trust
 	python -m orderly.clean --output_path="data/orderly/datasets_$(dataset_version)/orderly_with_trust_no_map.parquet" --ord_extraction_path="data/orderly/uspto_with_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_with_trust/all_molecule_names.csv" --min_frequency_of_occurrence=100 --map_rare_molecules_to_other=False --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=0 --num_cat=1 --num_reag=2 --consistent_yield=False --scramble=True --train_size=0.9
 
-paper_6: paper_gen_uspto_no_trust_no_map paper_gen_uspto_no_trust_with_map paper_gen_uspto_with_trust_with_map paper_gen_uspto_with_trust_no_map
+paper_gen_uspto_no_trust_no_min_freq: #requires: paper_extract_uspto_no_trust
+	python -m orderly.clean --output_path="data/orderly/datasets_$(dataset_version)/orderly_no_trust_no_min_freq.parquet" --ord_extraction_path="data/orderly/uspto_no_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_no_trust/all_molecule_names.csv" --min_frequency_of_occurrence=0 --map_rare_molecules_to_other=False --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=3 --num_cat=0 --num_reag=0 --consistent_yield=False --scramble=True --train_size=0.9
+
+paper_gen_uspto_with_trust_no_min_freq: #requires: paper_extract_uspto_with_trust
+	python -m orderly.clean --output_path="data/orderly/datasets_$(dataset_version)/orderly_with_trust_no_min_freq.parquet" --ord_extraction_path="data/orderly/uspto_with_trust/extracted_ords" --molecules_to_remove_path="data/orderly/uspto_with_trust/all_molecule_names.csv" --min_frequency_of_occurrence=0 --map_rare_molecules_to_other=False --set_unresolved_names_to_none_if_mapped_rxn_str_exists_else_del_rxn=True --remove_rxn_with_unresolved_names=False --set_unresolved_names_to_none=False --num_product=1 --num_reactant=2 --num_solv=2 --num_agent=0 --num_cat=1 --num_reag=2 --consistent_yield=False --scramble=True --train_size=0.9
+
+
+paper_6: paper_gen_uspto_no_trust_no_map paper_gen_uspto_no_trust_with_map paper_gen_uspto_with_trust_with_map paper_gen_uspto_with_trust_no_map paper_gen_uspto_no_trust_no_min_freq paper_gen_uspto_with_trust_no_min_freq
 
 # 7. Plot plot_molecule_popularity_histograms 
 paper_plot_uspto_no_trust_no_map:
-	python -m orderly.plot --clean_data_path="data/orderly/datasets/orderly_no_trust_no_map_train.parquet" --plot_output_path="data/orderly/plot_no_trust/" --plot_num_rxn_components_bool=False --plot_frequency_of_occurrence_bool=False --plot_molecule_popularity_histograms=True 
+	python -m orderly.plot --clean_data_path="data/orderly/datasets_$(dataset_version)/orderly_no_trust_no_map_train.parquet" --plot_output_path="data/orderly/plot_no_trust/" --plot_num_rxn_components_bool=False --plot_frequency_of_occurrence_bool=False --plot_molecule_popularity_histograms=True 
 
 paper_plot_uspto_with_trust_no_map:
-	python -m orderly.plot --clean_data_path="data/orderly/datasets/orderly_with_trust_no_map_train.parquet" --plot_output_path="data/orderly/plot_with_trust/" --plot_num_rxn_components_bool=False --plot_frequency_of_occurrence_bool=False --plot_molecule_popularity_histograms=True
+	python -m orderly.plot --clean_data_path="data/orderly/datasets_$(dataset_version)/orderly_with_trust_no_map_train.parquet" --plot_output_path="data/orderly/plot_with_trust/" --plot_num_rxn_components_bool=False --plot_frequency_of_occurrence_bool=False --plot_molecule_popularity_histograms=True
 	
 paper_7 : paper_plot_uspto_no_trust_no_map  paper_plot_uspto_with_trust_no_map
 
-# 8. gen fp
+# 8. Gen benchmarks (final datasets)
+
+paper_orderly_condition:
+
+
+paper_orderly_forward:
+
+paper_orderly_retro:
+
+paper_orderly_yield:
+
+
+
+# 9. gen fp
 
 fp_no_trust_no_map_test:
 	python -m orderly.gen_fp --clean_data_folder_path="data/orderly/datasets_$(dataset_version)/orderly_no_trust_no_map_test.parquet" --fp_size=2048 --overwrite=False
@@ -240,7 +257,7 @@ paper_8: fp_no_trust_no_map_test fp_no_trust_no_map_train fp_no_trust_with_map_t
 #Generate datasets for paper
 paper_get_datasets: paper_1 paper_6 paper_8
 
-paper_gen_all: paper_1 paper_2 paper_3 paper_4 paper_5 paper_6 paper_8
+paper_gen_all: paper_1 paper_2 paper_3 paper_4 paper_5 paper_6 paper_8 paper_9
 
 # 9. train models
 #Remember to switch env here (must contain TF, e.g. tf_mac_m1)

From e544336d3bf449cbfb451b21e6dc20a9efd7a40b Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Sat, 20 Jan 2024 21:22:23 +0000
Subject: [PATCH 23/29] looking for bugs in prelim orderly-condition

---
 notebooks/dataset_freq.ipynb | 826 ++++++++++++++++++++++++++++++++++-
 1 file changed, 824 insertions(+), 2 deletions(-)

diff --git a/notebooks/dataset_freq.ipynb b/notebooks/dataset_freq.ipynb
index 454c4b3f..d1829454 100644
--- a/notebooks/dataset_freq.ipynb
+++ b/notebooks/dataset_freq.ipynb
@@ -237,12 +237,834 @@
     "df['product_000'].value_counts().shape"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Most common agents in orderly-cond"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import pandas as pd"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "path = \"/Users/dsw46/Projects_local/ORDerly-project/ORDerly/data/orderly/datasets_v6/orderly_no_trust_no_min_freq_train.parquet\"\n",
+    "df = pd.read_parquet(path)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 43,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>original_index</th>\n",
+       "      <th>agent_000</th>\n",
+       "      <th>agent_001</th>\n",
+       "      <th>agent_002</th>\n",
+       "      <th>date_of_experiment</th>\n",
+       "      <th>extracted_from_file</th>\n",
+       "      <th>grant_date</th>\n",
+       "      <th>is_mapped</th>\n",
+       "      <th>procedure_details</th>\n",
+       "      <th>product_000</th>\n",
+       "      <th>reactant_000</th>\n",
+       "      <th>reactant_001</th>\n",
+       "      <th>rxn_str</th>\n",
+       "      <th>rxn_time</th>\n",
+       "      <th>solvent_000</th>\n",
+       "      <th>solvent_001</th>\n",
+       "      <th>temperature</th>\n",
+       "      <th>yield_000</th>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>index</th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th>371755</th>\n",
+       "      <td>864236</td>\n",
+       "      <td>O=C[O-]</td>\n",
+       "      <td>[Pd]</td>\n",
+       "      <td>[NH4+]</td>\n",
+       "      <td>NaT</td>\n",
+       "      <td>ord_dataset-b8b98725045d45bdbd73512048f4b47e</td>\n",
+       "      <td>2009-01-01 00:02:00</td>\n",
+       "      <td>True</td>\n",
+       "      <td>A solution of 740 mg of 5-benzyloxy-3-[1-tert-...</td>\n",
+       "      <td>CC(C)(C)OC(=O)n1nc(-c2cc3cc(OCCN4CCOCC4)ccc3n2...</td>\n",
+       "      <td>CC(C)(C)OC(=O)n1nc(-c2cc3cc(OCCN4CCOCC4)ccc3n2...</td>\n",
+       "      <td>NULL</td>\n",
+       "      <td>[C:1]([O:5][C:6]([N:8]1[C:16]2[C:11](=[CH:12][...</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>CCO</td>\n",
+       "      <td>None</td>\n",
+       "      <td>25.0</td>\n",
+       "      <td>83.2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>111108</th>\n",
+       "      <td>245483</td>\n",
+       "      <td>CNC</td>\n",
+       "      <td>[Pd]</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaT</td>\n",
+       "      <td>ord_dataset-5eb2900a93c842ee98f26c305e657b61</td>\n",
+       "      <td>1992-01-01 00:04:00</td>\n",
+       "      <td>True</td>\n",
+       "      <td>10 g (0.03 mol) of 2-(2'-hydroxy-3',5'-di-t-bu...</td>\n",
+       "      <td>CC(C)(C)c1cc(-n2nc3ccccc3n2)c(O)c(C(C)(C)C)c1</td>\n",
+       "      <td>CC(C)(C)c1cc(-n2nc3ccccc3[n+]2[O-])c(O)c(C(C)(...</td>\n",
+       "      <td>[H][H]</td>\n",
+       "      <td>[OH:1][C:2]1[C:7]([C:8]([CH3:11])([CH3:10])[CH...</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>Cc1ccccc1</td>\n",
+       "      <td>O</td>\n",
+       "      <td>50.0</td>\n",
+       "      <td>73.2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>67346</th>\n",
+       "      <td>150389</td>\n",
+       "      <td>Cl</td>\n",
+       "      <td>[Pd]</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaT</td>\n",
+       "      <td>ord_dataset-b9a9e369e9da4413999591aa08f4c3e3</td>\n",
+       "      <td>1986-01-01 00:11:00</td>\n",
+       "      <td>True</td>\n",
+       "      <td>4.36 g. of the resultant compound of Example 5...</td>\n",
+       "      <td>COc1ccc2c(c1OC)CC1CNCC21</td>\n",
+       "      <td>COc1ccc2c(c1OC)CC1CN(Cc3ccccc3)CC21</td>\n",
+       "      <td>NULL</td>\n",
+       "      <td>[ClH:1].C([N:9]1[CH2:13][CH:12]2[CH2:14][C:15]...</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>CO</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>121435</th>\n",
+       "      <td>269124</td>\n",
+       "      <td>[BH4-]</td>\n",
+       "      <td>[Pd]</td>\n",
+       "      <td>[Na+]</td>\n",
+       "      <td>NaT</td>\n",
+       "      <td>ord_dataset-a20aed058d7b40bc81fdf50bc5b03f97</td>\n",
+       "      <td>1993-01-01 00:06:00</td>\n",
+       "      <td>True</td>\n",
+       "      <td>4-(2-hydroxyethoxy)-6,6'-dimethyl-2,2'-bipyrid...</td>\n",
+       "      <td>Cc1cccc(-c2cc(OCCO)cc(C)n2)n1</td>\n",
+       "      <td>Cc1cccc(-c2cc(OCCO)cc(C)[n+]2[O-])n1</td>\n",
+       "      <td>NULL</td>\n",
+       "      <td>[OH:1][CH2:2][CH2:3][O:4][C:5]1[CH:6]=[C:7]([C...</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>CO</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>21375</th>\n",
+       "      <td>42319</td>\n",
+       "      <td>NC(=O)C(=O)O</td>\n",
+       "      <td>[Pd]</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaT</td>\n",
+       "      <td>ord_dataset-4c8627b52d564809adb9b494879c07c0</td>\n",
+       "      <td>1978-01-01 00:07:00</td>\n",
+       "      <td>True</td>\n",
+       "      <td>A suspension of 10 parts of [[[2-[3-(dimethyla...</td>\n",
+       "      <td>CCOC(=O)CN(Cc1ccccc1OCCCN(C)C)C(=O)C=Cc1ccc(N)cc1</td>\n",
+       "      <td>CCOC(=O)CN(Cc1ccccc1OCCCN(C)C)C(=O)C=Cc1ccc([N...</td>\n",
+       "      <td>[H][H]</td>\n",
+       "      <td>[C:1]([OH:6])(=[O:5])[C:2]([NH2:4])=[O:3].[CH3...</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>CCO</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>...</th>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>752297</th>\n",
+       "      <td>1767946</td>\n",
+       "      <td>c1ccc(P(c2ccccc2)c2ccccc2)cc1</td>\n",
+       "      <td>[Pd]</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaT</td>\n",
+       "      <td>ord_dataset-0b32b90cc77b4a3db47ad263e0bbc1a8</td>\n",
+       "      <td>2016-01-01 00:09:00</td>\n",
+       "      <td>True</td>\n",
+       "      <td>A mixture of dimethyl(2S,3S,5S)-5-[(allyloxy)c...</td>\n",
+       "      <td>C=CCO[C@H]1C[C@H](C(=O)OC)[C@@H](C(=O)N2CCN(c3...</td>\n",
+       "      <td>C1CCOC1</td>\n",
+       "      <td>C=CCOC(=O)O[C@H]1C[C@H](C(=O)OC)[C@@H](C(=O)N2...</td>\n",
+       "      <td>C(O[C:5]([O:7][C@@H:8]1[CH2:13][N:12]([C:14]([...</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>None</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>179187</th>\n",
+       "      <td>419812</td>\n",
+       "      <td>Cl</td>\n",
+       "      <td>[Pd]</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaT</td>\n",
+       "      <td>ord_dataset-94e21e9990034c729ea727e7d2ab0eb0</td>\n",
+       "      <td>1998-01-01 00:12:00</td>\n",
+       "      <td>True</td>\n",
+       "      <td>To a solution of 4-amino-3-nitrophenol (25.0 g...</td>\n",
+       "      <td>Nc1ccc(O)cc1N</td>\n",
+       "      <td>Nc1ccc(O)cc1[N+](=O)[O-]</td>\n",
+       "      <td>NULL</td>\n",
+       "      <td>[NH2:1][C:2]1[CH:7]=[CH:6][C:5]([OH:8])=[CH:4]...</td>\n",
+       "      <td>12.0</td>\n",
+       "      <td>CO</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>147296</th>\n",
+       "      <td>329677</td>\n",
+       "      <td>NN</td>\n",
+       "      <td>[Pd]</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaT</td>\n",
+       "      <td>ord_dataset-2c460e2ef9934444aaf26fec1f75741f</td>\n",
+       "      <td>1996-01-01 00:05:00</td>\n",
+       "      <td>True</td>\n",
+       "      <td>To a solution of 5 mmol of 6,7-dihydro-5-(4-ni...</td>\n",
+       "      <td>Nc1ccc(C(=O)N2CCc3cccn3-c3ccccc32)cc1</td>\n",
+       "      <td>O=C(c1ccc([N+](=O)[O-])cc1)N1CCc2cccn2-c2ccccc21</td>\n",
+       "      <td>NULL</td>\n",
+       "      <td>[N+:1]([C:4]1[CH:25]=[CH:24][C:7]([C:8]([N:10]...</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>CCO</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>236303</th>\n",
+       "      <td>560550</td>\n",
+       "      <td>c1ccc(P(c2ccccc2)c2ccccc2)cc1</td>\n",
+       "      <td>[Pd]</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaT</td>\n",
+       "      <td>ord_dataset-7c28974b7fcf4c9c86d5f2a42ba328a2</td>\n",
+       "      <td>2002-01-01 00:09:00</td>\n",
+       "      <td>True</td>\n",
+       "      <td>A mixture of 6-(3-bromo-phenyl)-4,4-dimethyl-1...</td>\n",
+       "      <td>CC1(C)OC(=O)Nc2ccc(-c3cccc(C#C[Si](C)(C)C)c3)cc21</td>\n",
+       "      <td>C#C[Si](C)(C)C</td>\n",
+       "      <td>CC1(C)OC(=O)Nc2ccc(-c3cccc(Br)c3)cc21</td>\n",
+       "      <td>Br[C:2]1[CH:3]=[C:4]([C:8]2[CH:20]=[CH:19][C:1...</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>CCN(CC)CC</td>\n",
+       "      <td>None</td>\n",
+       "      <td>80.0</td>\n",
+       "      <td>91.8</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>578996</th>\n",
+       "      <td>1328400</td>\n",
+       "      <td>[H][H]</td>\n",
+       "      <td>[Pd]</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaT</td>\n",
+       "      <td>ord_dataset-cfad8b3f00044bcda60a96b019f09872</td>\n",
+       "      <td>2013-01-01 00:08:00</td>\n",
+       "      <td>True</td>\n",
+       "      <td>2-(Trifluoromethyl)imidazo[1,2-a]pyrazine (1.6...</td>\n",
+       "      <td>FC(F)(F)c1cn2c(n1)CNCC2</td>\n",
+       "      <td>FC(F)(F)c1cn2ccncc2n1</td>\n",
+       "      <td>NULL</td>\n",
+       "      <td>[F:1][C:2]([F:13])([F:12])[C:3]1[N:4]=[C:5]2[C...</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>CO</td>\n",
+       "      <td>None</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>93.7</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "<p>2683 rows × 18 columns</p>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "        original_index                      agent_000 agent_001 agent_002  \\\n",
+       "index                                                                       \n",
+       "371755          864236                        O=C[O-]      [Pd]    [NH4+]   \n",
+       "111108          245483                            CNC      [Pd]      None   \n",
+       "67346           150389                             Cl      [Pd]      None   \n",
+       "121435          269124                         [BH4-]      [Pd]     [Na+]   \n",
+       "21375            42319                   NC(=O)C(=O)O      [Pd]      None   \n",
+       "...                ...                            ...       ...       ...   \n",
+       "752297         1767946  c1ccc(P(c2ccccc2)c2ccccc2)cc1      [Pd]      None   \n",
+       "179187          419812                             Cl      [Pd]      None   \n",
+       "147296          329677                             NN      [Pd]      None   \n",
+       "236303          560550  c1ccc(P(c2ccccc2)c2ccccc2)cc1      [Pd]      None   \n",
+       "578996         1328400                         [H][H]      [Pd]      None   \n",
+       "\n",
+       "       date_of_experiment                           extracted_from_file  \\\n",
+       "index                                                                     \n",
+       "371755                NaT  ord_dataset-b8b98725045d45bdbd73512048f4b47e   \n",
+       "111108                NaT  ord_dataset-5eb2900a93c842ee98f26c305e657b61   \n",
+       "67346                 NaT  ord_dataset-b9a9e369e9da4413999591aa08f4c3e3   \n",
+       "121435                NaT  ord_dataset-a20aed058d7b40bc81fdf50bc5b03f97   \n",
+       "21375                 NaT  ord_dataset-4c8627b52d564809adb9b494879c07c0   \n",
+       "...                   ...                                           ...   \n",
+       "752297                NaT  ord_dataset-0b32b90cc77b4a3db47ad263e0bbc1a8   \n",
+       "179187                NaT  ord_dataset-94e21e9990034c729ea727e7d2ab0eb0   \n",
+       "147296                NaT  ord_dataset-2c460e2ef9934444aaf26fec1f75741f   \n",
+       "236303                NaT  ord_dataset-7c28974b7fcf4c9c86d5f2a42ba328a2   \n",
+       "578996                NaT  ord_dataset-cfad8b3f00044bcda60a96b019f09872   \n",
+       "\n",
+       "                grant_date  is_mapped  \\\n",
+       "index                                   \n",
+       "371755 2009-01-01 00:02:00       True   \n",
+       "111108 1992-01-01 00:04:00       True   \n",
+       "67346  1986-01-01 00:11:00       True   \n",
+       "121435 1993-01-01 00:06:00       True   \n",
+       "21375  1978-01-01 00:07:00       True   \n",
+       "...                    ...        ...   \n",
+       "752297 2016-01-01 00:09:00       True   \n",
+       "179187 1998-01-01 00:12:00       True   \n",
+       "147296 1996-01-01 00:05:00       True   \n",
+       "236303 2002-01-01 00:09:00       True   \n",
+       "578996 2013-01-01 00:08:00       True   \n",
+       "\n",
+       "                                        procedure_details  \\\n",
+       "index                                                       \n",
+       "371755  A solution of 740 mg of 5-benzyloxy-3-[1-tert-...   \n",
+       "111108  10 g (0.03 mol) of 2-(2'-hydroxy-3',5'-di-t-bu...   \n",
+       "67346   4.36 g. of the resultant compound of Example 5...   \n",
+       "121435  4-(2-hydroxyethoxy)-6,6'-dimethyl-2,2'-bipyrid...   \n",
+       "21375   A suspension of 10 parts of [[[2-[3-(dimethyla...   \n",
+       "...                                                   ...   \n",
+       "752297  A mixture of dimethyl(2S,3S,5S)-5-[(allyloxy)c...   \n",
+       "179187  To a solution of 4-amino-3-nitrophenol (25.0 g...   \n",
+       "147296  To a solution of 5 mmol of 6,7-dihydro-5-(4-ni...   \n",
+       "236303  A mixture of 6-(3-bromo-phenyl)-4,4-dimethyl-1...   \n",
+       "578996  2-(Trifluoromethyl)imidazo[1,2-a]pyrazine (1.6...   \n",
+       "\n",
+       "                                              product_000  \\\n",
+       "index                                                       \n",
+       "371755  CC(C)(C)OC(=O)n1nc(-c2cc3cc(OCCN4CCOCC4)ccc3n2...   \n",
+       "111108      CC(C)(C)c1cc(-n2nc3ccccc3n2)c(O)c(C(C)(C)C)c1   \n",
+       "67346                            COc1ccc2c(c1OC)CC1CNCC21   \n",
+       "121435                      Cc1cccc(-c2cc(OCCO)cc(C)n2)n1   \n",
+       "21375   CCOC(=O)CN(Cc1ccccc1OCCCN(C)C)C(=O)C=Cc1ccc(N)cc1   \n",
+       "...                                                   ...   \n",
+       "752297  C=CCO[C@H]1C[C@H](C(=O)OC)[C@@H](C(=O)N2CCN(c3...   \n",
+       "179187                                      Nc1ccc(O)cc1N   \n",
+       "147296              Nc1ccc(C(=O)N2CCc3cccn3-c3ccccc32)cc1   \n",
+       "236303  CC1(C)OC(=O)Nc2ccc(-c3cccc(C#C[Si](C)(C)C)c3)cc21   \n",
+       "578996                            FC(F)(F)c1cn2c(n1)CNCC2   \n",
+       "\n",
+       "                                             reactant_000  \\\n",
+       "index                                                       \n",
+       "371755  CC(C)(C)OC(=O)n1nc(-c2cc3cc(OCCN4CCOCC4)ccc3n2...   \n",
+       "111108  CC(C)(C)c1cc(-n2nc3ccccc3[n+]2[O-])c(O)c(C(C)(...   \n",
+       "67346                 COc1ccc2c(c1OC)CC1CN(Cc3ccccc3)CC21   \n",
+       "121435               Cc1cccc(-c2cc(OCCO)cc(C)[n+]2[O-])n1   \n",
+       "21375   CCOC(=O)CN(Cc1ccccc1OCCCN(C)C)C(=O)C=Cc1ccc([N...   \n",
+       "...                                                   ...   \n",
+       "752297                                            C1CCOC1   \n",
+       "179187                           Nc1ccc(O)cc1[N+](=O)[O-]   \n",
+       "147296   O=C(c1ccc([N+](=O)[O-])cc1)N1CCc2cccn2-c2ccccc21   \n",
+       "236303                                     C#C[Si](C)(C)C   \n",
+       "578996                              FC(F)(F)c1cn2ccncc2n1   \n",
+       "\n",
+       "                                             reactant_001  \\\n",
+       "index                                                       \n",
+       "371755                                               NULL   \n",
+       "111108                                             [H][H]   \n",
+       "67346                                                NULL   \n",
+       "121435                                               NULL   \n",
+       "21375                                              [H][H]   \n",
+       "...                                                   ...   \n",
+       "752297  C=CCOC(=O)O[C@H]1C[C@H](C(=O)OC)[C@@H](C(=O)N2...   \n",
+       "179187                                               NULL   \n",
+       "147296                                               NULL   \n",
+       "236303              CC1(C)OC(=O)Nc2ccc(-c3cccc(Br)c3)cc21   \n",
+       "578996                                               NULL   \n",
+       "\n",
+       "                                                  rxn_str  rxn_time  \\\n",
+       "index                                                                 \n",
+       "371755  [C:1]([O:5][C:6]([N:8]1[C:16]2[C:11](=[CH:12][...       NaN   \n",
+       "111108  [OH:1][C:2]1[C:7]([C:8]([CH3:11])([CH3:10])[CH...       NaN   \n",
+       "67346   [ClH:1].C([N:9]1[CH2:13][CH:12]2[CH2:14][C:15]...       NaN   \n",
+       "121435  [OH:1][CH2:2][CH2:3][O:4][C:5]1[CH:6]=[C:7]([C...       NaN   \n",
+       "21375   [C:1]([OH:6])(=[O:5])[C:2]([NH2:4])=[O:3].[CH3...       NaN   \n",
+       "...                                                   ...       ...   \n",
+       "752297  C(O[C:5]([O:7][C@@H:8]1[CH2:13][N:12]([C:14]([...       NaN   \n",
+       "179187  [NH2:1][C:2]1[CH:7]=[CH:6][C:5]([OH:8])=[CH:4]...      12.0   \n",
+       "147296  [N+:1]([C:4]1[CH:25]=[CH:24][C:7]([C:8]([N:10]...       NaN   \n",
+       "236303  Br[C:2]1[CH:3]=[C:4]([C:8]2[CH:20]=[CH:19][C:1...       NaN   \n",
+       "578996  [F:1][C:2]([F:13])([F:12])[C:3]1[N:4]=[C:5]2[C...       NaN   \n",
+       "\n",
+       "       solvent_000 solvent_001  temperature  yield_000  \n",
+       "index                                                   \n",
+       "371755         CCO        None         25.0       83.2  \n",
+       "111108   Cc1ccccc1           O         50.0       73.2  \n",
+       "67346           CO        None          NaN        NaN  \n",
+       "121435          CO        None          NaN        NaN  \n",
+       "21375          CCO        None          NaN        NaN  \n",
+       "...            ...         ...          ...        ...  \n",
+       "752297        None        None          NaN        NaN  \n",
+       "179187          CO        None          NaN        NaN  \n",
+       "147296         CCO        None          NaN        NaN  \n",
+       "236303   CCN(CC)CC        None         80.0       91.8  \n",
+       "578996          CO        None          NaN       93.7  \n",
+       "\n",
+       "[2683 rows x 18 columns]"
+      ]
+     },
+     "execution_count": 43,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# 1. Check [Pd]/C not present\n",
+    "df[df['agent_001'] == \"[Pd]\"]\n",
+    "# OMG there's a bug, the [Pd] should be first!!\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 24,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "To a stirred solution of 3-(4-propyl-2-methoxyphenoxy)-2-nitropyridine (1.1 g, 3.8 mmol) in methanol (15 ml) was added anhydrous Ferric chloride (55 mg, 5% by wt) and activated charcoal (55 mg, 5% by wt). The resulting mixture was heated to reflux and hydrazine hydrate (570 mg, 11.45 mmol) was added dropwise. The reaction was allowed to stir under reflux condition overnight, then filtered through celite. The filtrate was concentrated under reduced pressure, taken in ethyl acetate (150 ml). The organic layer was washed with water followed by brine, dried over anhydrous sodium sulfate and concentrated under reduced pressure. The residue was passed through silica column (eluant ethylacetate/pet ether 1:3 to get 900 mg (91.3%) of title compound as a pale yellow solid.\n",
+      "619281\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Bug 2: Charcol is being written as \"C\" -> this becomes CH4\n",
+    "print(df[df['agent_001'] == \"C\"]['procedure_details'].iloc[0])\n",
+    "print(df[df['agent_001'] == \"C\"]['procedure_details'].index[0])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 23,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'[CH2:1]([C:4]1[CH:19]=[CH:18][C:7]([O:8][C:9]2[C:10]([N+:15]([O-])=O)=[N:11][CH:12]=[CH:13][CH:14]=2)=[C:6]([O:20][CH3:21])[CH:5]=1)[CH2:2][CH3:3].C.O.NN>CO>[CH3:21][O:20][C:6]1[CH:5]=[C:4]([CH2:1][CH2:2][CH3:3])[CH:19]=[CH:18][C:7]=1[O:8][C:9]1[C:10]([NH2:15])=[N:11][CH:12]=[CH:13][CH:14]=1'"
+      ]
+     },
+     "execution_count": 23,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df.loc[619281]['rxn_str']"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
    "metadata": {},
    "outputs": [],
-   "source": []
+   "source": [
+    "'[CH2:1]([C:4]1[CH:19]=[CH:18][C:7]([O:8][C:9]2[C:10]([N+:15]([O-])=O)=[N:11][CH:12]=[CH:13][CH:14]=2)=[C:6]([O:20][CH3:21])[CH:5]=1)[CH2:2][CH3:3].C.O.NN>CO>[CH3:21][O:20][C:6]1[CH:5]=[C:4]([CH2:1][CH2:2][CH3:3])[CH:19]=[CH:18][C:7]=1[O:8][C:9]1[C:10]([NH2:15])=[N:11][CH:12]=[CH:13][CH:14]=1'"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 41,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>original_index</th>\n",
+       "      <th>agent_000</th>\n",
+       "      <th>agent_001</th>\n",
+       "      <th>agent_002</th>\n",
+       "      <th>date_of_experiment</th>\n",
+       "      <th>extracted_from_file</th>\n",
+       "      <th>grant_date</th>\n",
+       "      <th>is_mapped</th>\n",
+       "      <th>procedure_details</th>\n",
+       "      <th>product_000</th>\n",
+       "      <th>reactant_000</th>\n",
+       "      <th>reactant_001</th>\n",
+       "      <th>rxn_str</th>\n",
+       "      <th>rxn_time</th>\n",
+       "      <th>solvent_000</th>\n",
+       "      <th>solvent_001</th>\n",
+       "      <th>temperature</th>\n",
+       "      <th>yield_000</th>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>index</th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "Empty DataFrame\n",
+       "Columns: [original_index, agent_000, agent_001, agent_002, date_of_experiment, extracted_from_file, grant_date, is_mapped, procedure_details, product_000, reactant_000, reactant_001, rxn_str, rxn_time, solvent_000, solvent_001, temperature, yield_000]\n",
+       "Index: []"
+      ]
+     },
+     "execution_count": 41,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df[df['procedure_details'] == \"zeolite\"]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 40,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "57"
+      ]
+     },
+     "execution_count": 40,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "len(df[df['agent_002'] == \"C\"])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "'C(O[CH:4](OCC)[CH2:5][NH:6][C:7]1[N:11]([CH3:12])[N:10]=[N:9][N:8]=1)C.[ClH:16].[OH:17][C:18]1[CH:19]=[C:20]([CH:24]=[CH:25][C:26]=1[OH:27])[CH2:21][CH2:22][NH2:23].Cl.C>O.C(O)C>[ClH:16].[CH3:12][N:11]1[C:7]([NH:6][CH2:5][CH:4]2[C:24]3[C:20](=[CH:19][C:18]([OH:17])=[C:26]([OH:27])[CH:25]=3)[CH2:21][CH2:22][NH:23]2)=[N:8][N:9]=[N:10]1'"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 25,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "agent_000\n",
+       "Cl                                                                                                                                        63102\n",
+       "[Na+]                                                                                                                                     56010\n",
+       "[K+]                                                                                                                                      25356\n",
+       "[OH-]                                                                                                                                     22763\n",
+       "[Pd]                                                                                                                                      22616\n",
+       "O=C([O-])[O-]                                                                                                                             21063\n",
+       "[H-]                                                                                                                                      12680\n",
+       "[Li+]                                                                                                                                      7218\n",
+       "O=S(=O)(O)O                                                                                                                                6866\n",
+       "[Cl-]                                                                                                                                      6442\n",
+       "O=C([O-])O                                                                                                                                 6336\n",
+       "[Li]CCCC                                                                                                                                   5612\n",
+       "c1ccc(P(c2ccccc2)c2ccccc2)cc1                                                                                                              5063\n",
+       "Cc1ccc(S(=O)(=O)O)cc1                                                                                                                      4410\n",
+       "CN(C)c1ccncc1                                                                                                                              4319\n",
+       "[BH4-]                                                                                                                                     3900\n",
+       "[NH4+]                                                                                                                                     3668\n",
+       "[Cs+]                                                                                                                                      3584\n",
+       "c1ccc([P](c2ccccc2)(c2ccccc2)[Pd]([P](c2ccccc2)(c2ccccc2)c2ccccc2)([P](c2ccccc2)(c2ccccc2)c2ccccc2)[P](c2ccccc2)(c2ccccc2)c2ccccc2)cc1     3504\n",
+       "Br                                                                                                                                         3178\n",
+       "Name: count, dtype: int64"
+      ]
+     },
+     "execution_count": 25,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# most common agents in agent_000\n",
+    "df['agent_000'].value_counts().head(20)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 28,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'[C:1]([O:5][C:6]([C:8]1[C:9]([C:14]2[CH:19]=[CH:18][C:17]([CH2:20][N:21]3[C:25]([CH:26]=O)=[C:24]([CH:28]=[CH2:29])[N:23]=[C:22]3[O:30][CH2:31][CH3:32])=[C:16]([F:33])[CH:15]=2)=[CH:10][CH:11]=[CH:12][CH:13]=1)=[O:7])([CH3:4])([CH3:3])[CH3:2].Cl.[NH2:35][OH:36]>N1C=CC=CC=1.O>[C:1]([O:5][C:6]([C:8]1[C:9]([C:14]2[CH:19]=[CH:18][C:17]([CH2:20][N:21]3[C:25]([CH:26]=[N:35][OH:36])=[C:24]([CH:28]=[CH2:29])[N:23]=[C:22]3[O:30][CH2:31][CH3:32])=[C:16]([F:33])[CH:15]=2)=[CH:10][CH:11]=[CH:12][CH:13]=1)=[O:7])([CH3:2])([CH3:4])[CH3:3]'"
+      ]
+     },
+     "execution_count": 28,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df[df['agent_000']==\"Cl\"]['rxn_str'].iloc[0]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 51,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'Intermediate (15b) (19.5 g, 43.4 mmol) was dissolved in pyridine (100 mL, 1 μmol). Hydroxylamine hydrochloride (9.0 g, 130 mmol) was added, followed by water (50 mL, 3 mol), and the mixture was stirred at room temperature overnight. Water (100 mL) was then added and the mixture was stirred for 20 minutes. The precipitant was filtered off and dried to yield intermediate (15c) (13.5 g). MS m/z: [M+H+] calcd for C26H28FN2O4, 466.2; found 466.4. 1H-NMR (CDCl3): 9.78 (1H, s), 7.81 (1H, d), 7.48 (2H, m), 7.26 (1H, s), 7.0 (4H, m), 6.20 (1H, d), 5.53 (1H, d), 5.50 (2H, s), 4.55 (2H, q), 1.43 (3H, t), 1.25 (9H, s).'"
+      ]
+     },
+     "execution_count": 51,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df[df['agent_000']==\"Cl\"]['procedure_details'].iloc[0]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "'[C:1]([O:5][C:6]([C:8]1[C:9]([C:14]2[CH:19]=[CH:18][C:17]([CH2:20][N:21]3[C:25]([CH:26]=O)=[C:24]([CH:28]=[CH2:29])[N:23]=[C:22]3[O:30][CH2:31][CH3:32])=[C:16]([F:33])[CH:15]=2)=[CH:10][CH:11]=[CH:12][CH:13]=1)=[O:7])([CH3:4])([CH3:3])[CH3:2].Cl.[NH2:35][OH:36]>N1C=CC=CC=1.O>[C:1]([O:5][C:6]([C:8]1[C:9]([C:14]2[CH:19]=[CH:18][C:17]([CH2:20][N:21]3[C:25]([CH:26]=[N:35][OH:36])=[C:24]([CH:28]=[CH2:29])[N:23]=[C:22]3[O:30][CH2:31][CH3:32])=[C:16]([F:33])[CH:15]=2)=[CH:10][CH:11]=[CH:12][CH:13]=1)=[O:7])([CH3:2])([CH3:4])[CH3:3]'\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 29,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'[Cl:1][C:2]1[CH:3]=[CH:4][C:5]([O:12][CH2:13][C:14]([N:16]2[CH2:21][C@H:20]([CH3:22])[N:19]([CH2:23][C:24]3[CH:29]=[CH:28][C:27]([F:30])=[CH:26][CH:25]=3)[CH2:18][C@H:17]2[CH3:31])=[O:15])=[C:6]([CH:11]=1)[O:7][CH2:8][C:9]#[N:10].[Cl-].[NH4+].[N-:34]=[N+:35]=[N-:36].[Na+]>CN(C)C=O.C(OCC)(=O)C>[Cl:1][C:2]1[CH:3]=[CH:4][C:5]([O:12][CH2:13][C:14]([N:16]2[CH2:21][C@H:20]([CH3:22])[N:19]([CH2:23][C:24]3[CH:25]=[CH:26][C:27]([F:30])=[CH:28][CH:29]=3)[CH2:18][C@H:17]2[CH3:31])=[O:15])=[C:6]([O:7][CH2:8][C:9]2[N:34]=[N:35][NH:36][N:10]=2)[CH:11]=1'"
+      ]
+     },
+     "execution_count": 29,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df[df['agent_000']==\"[Cl-]\"]['rxn_str'].iloc[0]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "'[Cl:1][C:2]1[CH:3]=[CH:4][C:5]([O:12][CH2:13][C:14]([N:16]2[CH2:21][C@H:20]([CH3:22])[N:19]([CH2:23][C:24]3[CH:29]=[CH:28][C:27]([F:30])=[CH:26][CH:25]=3)[CH2:18][C@H:17]2[CH3:31])=[O:15])=[C:6]([CH:11]=1)[O:7][CH2:8][C:9]#[N:10].[Cl-].[NH4+].[N-:34]=[N+:35]=[N-:36].[Na+]>CN(C)C=O.C(OCC)(=O)C>[Cl:1][C:2]1[CH:3]=[CH:4][C:5]([O:12][CH2:13][C:14]([N:16]2[CH2:21][C@H:20]([CH3:22])[N:19]([CH2:23][C:24]3[CH:25]=[CH:26][C:27]([F:30])=[CH:28][CH:29]=3)[CH2:18][C@H:17]2[CH3:31])=[O:15])=[C:6]([O:7][CH2:8][C:9]2[N:34]=[N:35][NH:36][N:10]=2)[CH:11]=1'\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 34,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'[CH3:1][N:2]([CH3:13])[C:3]1[CH:8]=[CH:7][CH:6]=[C:5]([N+:9]([O-])=O)[C:4]=1[CH3:12]>CO.[H][H].[Ni]>[CH3:1][N:2]([CH3:13])[C:3]1[C:4]([CH3:12])=[C:5]([CH:6]=[CH:7][CH:8]=1)[NH2:9]'"
+      ]
+     },
+     "execution_count": 34,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# Check reactions with H2\n",
+    "df[df['agent_000']==\"[H][H]\"]['rxn_str'].iloc[0]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# NB: atom mapping does not involve H atoms\n",
+    "'[CH3:1][N:2]([CH3:13])[C:3]1[CH:8]=[CH:7][CH:6]=[C:5]([N+:9]([O-])=O)[C:4]=1[CH3:12]>CO.[H][H].[Ni]>[CH3:1][N:2]([CH3:13])[C:3]1[C:4]([CH3:12])=[C:5]([CH:6]=[CH:7][CH:8]=1)[NH2:9]'\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 35,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'[N:1]([CH:4]([CH:12]([OH:26])[CH:13]([OH:25])[CH:14]([N:22]=[N+]=[N-])[CH2:15][C:16]1[CH:21]=[CH:20][CH:19]=[CH:18][CH:17]=1)[CH2:5][C:6]1[CH:11]=[CH:10][CH:9]=[CH:8][CH:7]=1)=[N+]=[N-].[H][H]>CO.[Pd]>[NH2:1][CH:4]([CH:12]([OH:26])[CH:13]([OH:25])[CH:14]([NH2:22])[CH2:15][C:16]1[CH:21]=[CH:20][CH:19]=[CH:18][CH:17]=1)[CH2:5][C:6]1[CH:11]=[CH:10][CH:9]=[CH:8][CH:7]=1'"
+      ]
+     },
+     "execution_count": 35,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df[df['reactant_000']==\"[H][H]\"]['rxn_str'].iloc[0]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 50,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "index\n",
+      "98490     [OH-:1].[Na+].[CH2:3]([Sn:7](Cl)(Cl)[CH2:8][CH...\n",
+      "297681    [Cl].[Br:2][C:3]1[CH:11]=[CH:10][C:6]([C:7]([O...\n",
+      "719131    [CH:1]1([C:4](=[O:11])[CH2:5][C:6]([O:8][CH2:9...\n",
+      "Name: rxn_str, dtype: object\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Bug 2: Charcol is being written as \"C\" -> this becomes CH4\n",
+    "print(df[df['agent_000'] == \"[Cl]\"]['rxn_str'])"
+   ]
   }
  ],
  "metadata": {
@@ -261,7 +1083,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.10.13"
+   "version": "3.10.12"
   },
   "orig_nbformat": 4
  },

From ac5fbfacd0ed8e66f2fe7c56feb2348fe0166dfd Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Sat, 20 Jan 2024 21:31:39 +0000
Subject: [PATCH 24/29] move benchmark generation

---
 Makefile | 31 +++++++++++++------------------
 1 file changed, 13 insertions(+), 18 deletions(-)

diff --git a/Makefile b/Makefile
index 946b42de..6908248d 100644
--- a/Makefile
+++ b/Makefile
@@ -125,9 +125,13 @@ run_python_310:
 # 5. Plot histogram showing dataset size as a function of min_frequency_of_occurrence (can probably use the min_frequency code from the cleaner within the plotter)
 # 6. Generate the six condition prediction datasets we need for the paper (split into train and test set)
 # 7. Plot histograms with the occurrence of the most common reactants, products, solvents, agents
-# 8. Generate the final benchmark datasets (split into train and test set)
-# 9. Generate fingerprints for each dataset
-# 10. Train & evaluate a model on each dataset
+# 8. Generate fingerprints for each dataset
+# 9. Train & evaluate a model on each dataset
+
+### Benchmark generation
+# I. Extract data
+# II. Clean data
+
 
 # 1. Extract 
 
@@ -217,20 +221,8 @@ paper_plot_uspto_with_trust_no_map:
 	
 paper_7 : paper_plot_uspto_no_trust_no_map  paper_plot_uspto_with_trust_no_map
 
-# 8. Gen benchmarks (final datasets)
-
-paper_orderly_condition:
-
-
-paper_orderly_forward:
-
-paper_orderly_retro:
-
-paper_orderly_yield:
-
-
 
-# 9. gen fp
+# 8. gen fp
 
 fp_no_trust_no_map_test:
 	python -m orderly.gen_fp --clean_data_folder_path="data/orderly/datasets_$(dataset_version)/orderly_no_trust_no_map_test.parquet" --fp_size=2048 --overwrite=False
@@ -255,9 +247,9 @@ fp_with_trust_no_map_train:
 paper_8: fp_no_trust_no_map_test fp_no_trust_no_map_train fp_no_trust_with_map_test fp_no_trust_with_map_train fp_with_trust_with_map_test fp_with_trust_with_map_train fp_with_trust_no_map_test fp_with_trust_no_map_train
 
 #Generate datasets for paper
-paper_get_datasets: paper_1 paper_6 paper_8
+paper_get_datasets: paper_1 paper_6
 
-paper_gen_all: paper_1 paper_2 paper_3 paper_4 paper_5 paper_6 paper_8 paper_9
+paper_gen_all: paper_1 paper_2 paper_3 paper_4 paper_5 paper_6 paper_8
 
 # 9. train models
 #Remember to switch env here (must contain TF, e.g. tf_mac_m1)
@@ -287,6 +279,9 @@ with_trust_no_map_train_20:
 with_trust_with_map_train_20:
 	python -m condition_prediction --train_data_path="data/orderly/datasets_$(dataset_version)/orderly_with_trust_with_map_train.parquet" --test_data_path="data/orderly/datasets_$(dataset_version)/orderly_with_trust_with_map_test.parquet" --output_folder_path="models/with_trust_with_map_20"  --train_fraction=0.2 --train_val_split=0.8 --overwrite=False --epochs=20 --evaluate_on_test_data=True --early_stopping_patience=5 --wandb_entity=$(WANDB_ENTITY)
 
+################################################
+# Generate ORDerly benchmarks
+################################################
 
 # Sweeps
 RANDOM_SEEDS = 12345 54321 98765

From 9611e08f9f6cb9a2b1ebf6d96d8faf9aaa7ac16c Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Sat, 20 Jan 2024 21:43:20 +0000
Subject: [PATCH 25/29] make black

---
 orderly/clean/cleaner.py | 13 ++++++++++---
 1 file changed, 10 insertions(+), 3 deletions(-)

diff --git a/orderly/clean/cleaner.py b/orderly/clean/cleaner.py
index a53dc09f..f84d83bf 100644
--- a/orderly/clean/cleaner.py
+++ b/orderly/clean/cleaner.py
@@ -475,12 +475,19 @@ def _move_none_to_after_data(
     @staticmethod
     def _scramble(
         df: pd.DataFrame,
-        components: Tuple[str, ...] = ("reactant", "product", "solvent", "catalyst", "reagent"),
+        components: Tuple[str, ...] = (
+            "reactant",
+            "product",
+            "solvent",
+            "catalyst",
+            "reagent",
+        ),
         seed: int = 42,
     ) -> pd.DataFrame:
         """Scrambles the order of the reactants (ie between reactant_001, reactant_002, etc). Ordering of products, solvents, reagents, and catalysts will also be scrambled. This is done to prevent the model from learning the order of the molecules, which is not important for the reaction prediction task. It only done at the very end because scrambling can be non-deterministic between versions/operating systems, so it would be difficult to debug if done earlier in the pipeline.
-        
-        NB: agents not scrambled by default, since we need their ordering to let transition metals be first."""
+
+        NB: agents not scrambled by default, since we need their ordering to let transition metals be first.
+        """
         list_of_dfs = []
         all_component_cols = []
         np.random.seed(seed)

From e7b8516495541756e2413ce9af34578a83276b59 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Sat, 20 Jan 2024 21:48:13 +0000
Subject: [PATCH 26/29] restore extractor from random comment insertions

---
 orderly/extract/extractor.py | 27 ++++++++-------------------
 1 file changed, 8 insertions(+), 19 deletions(-)

diff --git a/orderly/extract/extractor.py b/orderly/extract/extractor.py
index 6d4a6115..a364886d 100644
--- a/orderly/extract/extractor.py
+++ b/orderly/extract/extractor.py
@@ -474,15 +474,11 @@ def rxn_time_extractor(rxn: ord_reaction_pb2.Reaction) -> Optional[RXN_TIME]:
         else:
             return None  # no time found
 
-    @sta
-    #NB this needs to be before the removal of [C] and Cticmethod
+    @staticmethod
     def procedure_details_extractor(
         rxn: ord_reaction_pb2.Reaction,
-    ) ->
-    #NB this needs to be before the removal of [C] and C str:  # TODO check does it re
-    #NB this needs to be before the removal of [C] and Cturn empty string or none
-        procedure_d
-        #NB this needs to be before the removal of [C] and Cetails = rxn.notes.procedure_details
+    ) -> str:  # TODO check does it return empty string or none
+        procedure_details = rxn.notes.procedure_details
         return str(procedure_details)
 
     @staticmethod
@@ -1071,8 +1067,7 @@ def contains_charcoal(procedure_details) -> bool:
             yields,
             temperature,
             rxn_time,
-            
-            #NB this needs to be before the removal of [C] and Crxn_str,
+            rxn_str,
             procedure_details,
             date_of_experiment,
             is_mapped,
@@ -1115,8 +1110,7 @@ def build_rxn_lists(
             "temperature": [],
             "rxn_time": [],
             "product": [],
-            "
-            #NB this needs to be before the removal of [C] and Cyield": [],
+            "yield": [],
             "procedure_details": [],
             "date_of_experiment": [],
             "is_mapped": [],
@@ -1144,8 +1138,7 @@ def build_rxn_lists(
                 yields,
                 temperature,
                 rxn_time,
-                
-                #NB this needs to be before the removal of [C] and Crxn_str,
+                rxn_str,
                 procedure_details,
                 date_of_experiment,
                 is_mapped,
@@ -1163,9 +1156,7 @@ def build_rxn_lists(
             rxn_lists["temperature"].append(temperature)
             rxn_lists["rxn_time"].append(rxn_time)
             rxn_lists["product"].append(products)
-            rxn_lists["
-            #NB this needs to be before the removal of [C] and Cyield"].append(yields)
-            #NB this needs to be before the removal of [C] and C
+            rxn_lists["yield"].append(yields)
             rxn_lists["procedure_details"].append(procedure_details)
             rxn_lists["date_of_experiment"].append(date_of_experiment)
             rxn_lists["is_mapped"].append(is_mapped)
@@ -1257,9 +1248,7 @@ def build_full_df(
             )
         )
         dfs.append(
-            OrdExtractor._to
-            #NB this needs to be before the removal of [C] and C_dataframe(
-                #NB this needs to be before the removal of [C] and C
+            OrdExtractor._to_dataframe(
                 data_lists["procedure_details"], base_string=["procedure_details"]
             )
             .fillna("<missing>")

From 776ed921a4eb7b4180ada88f394012e6e2093030 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Sat, 20 Jan 2024 21:48:33 +0000
Subject: [PATCH 27/29] make black

---
 orderly/extract/extractor.py | 18 ++++++++----------
 1 file changed, 8 insertions(+), 10 deletions(-)

diff --git a/orderly/extract/extractor.py b/orderly/extract/extractor.py
index a364886d..a205e83c 100644
--- a/orderly/extract/extractor.py
+++ b/orderly/extract/extractor.py
@@ -248,11 +248,10 @@ def extract_info_from_rxn_str(
                 # check reactant is mapped and also that it's not in the products
                 mol = rdkit_Chem.MolFromSmiles(r_map)
                 if mol != None:
-                    if (
-                        any(atom.HasProp("molAtomMapNumber") for atom in mol.GetAtoms())
-                        and (  # any(generator)
-                            r_clean not in products_from_rxn_without_mapping
-                        )
+                    if any(
+                        atom.HasProp("molAtomMapNumber") for atom in mol.GetAtoms()
+                    ) and (  # any(generator)
+                        r_clean not in products_from_rxn_without_mapping
                     ):
                         reactants.append(r_clean)
                     else:
@@ -1015,8 +1014,8 @@ def move_unresolvable_names_to_end_of_list(
         catalysts, _ = move_unresolvable_names_to_end_of_list(
             catalysts, rxn_non_smiles_names_set
         )
-        
-        #NB this needs to be before the removal of [C] and C
+
+        # NB this needs to be before the removal of [C] and C
         procedure_details = OrdExtractor.procedure_details_extractor(rxn)
         date_of_experiment = OrdExtractor.date_of_experiment_extractor(rxn)
         rxn_time = OrdExtractor.rxn_time_extractor(rxn)
@@ -1028,14 +1027,14 @@ def contains_transition_metal(agents: AGENTS) -> bool:
                 if orderly.extract.defaults.has_transition_metal(agent):
                     return True
             return False
-        
+
         def contains_charcoal(procedure_details) -> bool:
             if "charcoal" in procedure_details.lower():
                 return True
             return False
 
         # Check if any agent contains a transition metal
-        if (contains_transition_metal(agents) or contains_charcoal(procedure_details)):
+        if contains_transition_metal(agents) or contains_charcoal(procedure_details):
             # Remove "[C]" and "C" from agents
             agents = [a for a in agents if a not in ["[C]", "C"]]
 
@@ -1043,7 +1042,6 @@ def contains_charcoal(procedure_details) -> bool:
             _yields = [None] * len(products)
         yields = _yields
 
-        
         if ice_present and (
             temperature is None
         ):  # We trust the labelled temperature more, but if there is no labelled temperature, and they added ice, we should set the temperature to 0C

From ac67db133c2fb45c15cd5e5340125f5a8f3d64e3 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Sat, 20 Jan 2024 22:09:52 +0000
Subject: [PATCH 28/29] regen test data

---
 .../uspto-grants-1976_01.parquet              | Bin 835352 -> 835298 bytes
 .../uspto-grants-1978_12.parquet              | Bin 379848 -> 379809 bytes
 .../uspto-grants-1979_09.parquet              | Bin 319631 -> 319620 bytes
 .../uspto-grants-1980_08.parquet              | Bin 403061 -> 403046 bytes
 .../uspto-grants-1982_08.parquet              | Bin 580054 -> 580010 bytes
 .../uspto-grants-1986_09.parquet              | Bin 453362 -> 453362 bytes
 .../uspto-grants-1995_11.parquet              | Bin 678280 -> 678277 bytes
 .../uspto-grants-2001_07.parquet              | Bin 2067818 -> 2067777 bytes
 8 files changed, 0 insertions(+), 0 deletions(-)

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1976_01.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1976_01.parquet
index 90e2f09e9cc325aaaecde7a5ffc3cda563dace30..28aaa77c68c644fb0bb71e46fd0bb471f5085975 100644
GIT binary patch
delta 9024
zcmZvB3wRS%+W$FirkqJ9r^z%kZPF%f(k4xM&$JDtkkDjkX)7d!Lb;RzAyDoG%N>+j
zRs>d%#X$xtC`C~V0!6MC7AXn>LgnfzP@pO*peP6~uA2Q$m;d9pAN-%^J(-!jXU?2?
zZ@>3_XKhIM@(&@d+%wR@0dDQ8aPHQ$bnaXNxrpEp0H=;ijsg5mM;lfDf(Mx9<0K_=
z`R(|tsIqb(M1$HaP#X_~A|R*$y@UMGbxfoqs81E;V+9Aa&(oh#n&t`aE=Fc=pzzl5
zyJ;kfq9-fXF4{!9O)Sb(+x((ENJqRjofgJ}MitOc)i}EWeoQoQ@E(Toiw>8SH=+LT
z0?{6(@CBA#BtIu&@B-vA?b#wM1Do5)%B2sS$E5WUCd~r2pNh_RtHA&}g@CUhRcen&
ztiP@hM1CFa@r6o+nLyO2;2ULoXc)6>(WFX%rAVt8Q#p)Sm*%8$R`m$XF^<NACs1j=
zBg)3}uqFk-;HKYME<-2>mP=nMqCxJ78sIoO6a|9<pMinEj$jI`u|UYo!OPz5u)LrF
zR5OtRhb)!q;&ExTP~EmqAx^=1X?C~7dK_u*!H&}69>_Co<PMeK>XK0B)6MznEYT&7
zXgd48TSx^pDqzaymnp}wu#f&NL~bu;K}`v;Xq<bf)B6ezL<%C%I0X)v*07CC;M@uz
ztOWy|0L0+WFtWLPZBMC*v>kR`BrMr!fei4oXAwOpHqfG%Sr7}onm_P08Ydwj;$6{v
zK*~UI_DxFRVb^|6W&k>;poS`l0yv-s6*|f^nSFr~p6ODOrzd|$dKz&sTv~z>ORqCR
zZ(#4Ku~A4-lEJ<ix$~~VZ+Aik`5UT={B1Fm+g%GZ9KStMTka72<vE^F3V%6_KsLXF
zQ!;|xp99Oy(b7*yFJ!V@IiwlCOLRFAgroUfjl$(0=K(JL%2;9*<CK9H8Dc&nbhG<t
zlM=6#deObL-)Qw4j9MXR3p(udX$XN0kCRr?R!en)?y#{h26p(d>NuOKF})7NhA$OW
zPdP^eVJbNKqp1Tsq)T)HoG1N@j)ym;7-dSVkVp@tBXk*)l0m`Ol%Rt6f_B_;{1>TG
zIYhsM#Uqo+1D>N9(eCmaD-snBq5Wlm-;h35X2W*rp|WRGk0f@eang)1SdNDw{tw9;
z*|X$3qlE^dswbW1%%<ATxW7irUaWM)nNA#?TJthA#ijODT8^><)KGadtZk~NrH#_h
zPG%bz;G_QCypJ^*%3=VwNbi%5Mmjw&#^Zy$!sR=6Io!0OhYyIUcashLjR&|!2`ok3
z`8hFa;$}%cxB$2ef5nGmh&PE2$`1UQNr32lt<;MtPLvE(Z+$VbFA>(@9LBjfOp`1$
zKw<*a6yU1iq)kju_*dyFV^P#6V@~R(vf~<ajTf}JjVEm&Sg=5HtE%+b8JjmMY8_^X
zV}p)hRc%SNdlh8U=+qyjU8>Bump(99f*!BFTwNp)8%ga?TZhn8Jm{z2_A>4fX|~f5
z#@*2{w(Nw|2j^JgllV83MK7i<XY9p_pd%CuG_mk3@UojN*0f|+q)ok3dr0f?;>m5F
zm(2zXqrsTP;Ft~wlff8bbMKnW+-0FV@YF6|5B&Hdb{+zh;EAPZ)Nf+iva#T)0tED<
zGB%`B%Wu?o#iSL#F%G*|(2r+^zYbrX9}e4z!!2u9l$L6+a31J4VBxm3Dk?~;N&LXs
z2v3myh#D3-FvW0!TCMPm*Kd>p>S4+U>9|hC59XRiNgt_gktGh@$mZqxMkOqhepU~S
zSmgjVR+a{6s$iY8O*0~<hm-Qd_M^gI3xIiOV!*Y{Se6g)OHv|hhQCT8`!_XCEZUZ8
z5CRQ7N<;6s#=Ika#r^`7(xvER)fth5KkCc2oTfbfH7QanLZjr@z8LopaoO7>CVd_Y
za`1IAKA0N})&_&Q6vOtF9%!Gn%q;ZvEH1S=Z3<dDG6D#GNLxl4^|e`Hmh%AzyTMC5
z?Sdq8z2Hgd0@nt0(i&cZb-PkyFho~HA}*hKpi~%p7}iKoR|=n#p3)iME@`4J1)8K)
zIvX4<wd(Zn$6Xh7V<CJX6~uLF53UU7o*Vez0CV-@E`X7K&yL(x9)BRbo};NbivHti
zVv4h#ok)qpHzE?@g!ce9>@yk^_XD%P15`~Y2|S`NlfW1{VhqvwXxL6d(q4t=AD4wc
zq)Eg&(XFU}=|4VAP^oF2jsUvA->Ke<KTrtMj6F3>PzXA3P;>Nm{7|j1iTFVjrZuYi
zG%zDG4Oo8lEGcTM`+=(YmIU#luQ)#yO&uh*l)6_*J<XJ4R44FXBVePB<5LJsq2Ga#
z$l>zA_sL#Es{bXg3gG$Cpd!S>Nj$VPU($zl_UQ^bim=dD;M@v;V<LTe0r21bY+T&L
z(La1wKeZNOAcpe5TY10)k>BN?6>I5F&~6e7xwLj0WEcQIpj1tqE3Mkys%;+tLb?A-
zFgNez4TpSM(3lQ=JwcroZ0^VYs^n6r=*@K}qJ#yY&H#!Ux?VxKc@C)HWnMp*p+<GN
zgkhA&au|s|v91CL&jP9sj%Mg}R1~bB=YsJ^ZvnrxNJAyIJtwu9v;)$$3)Mh)8iYum
zV6)6uhz|ajSVd4+ycMc|E%SO5u;bTszrJpvO*W1eRs&!hJ_smmKHL_oCYb3+G-ngZ
z`>9$GbWaDy2+3;BtiMPrd_MlP2H1*X*#kNtde`Ga!-0_91xJ6C0hd%v1Iu+p_K)Ys
z0rBn_yplyE;IopFEC&Bh6q{o&n5$=}#=*4C_wW=PPf)^6G6?_T$?Wr5fNxyJN8_Lg
z#+SY#90qi>cU+v1+y!WCMRXYcL%L|rW}~^fSJ9y$Xv_wUDS&FlZ%Q#qUe#Rfz$^ec
zBqoW!beC$9GQ;9k(wd~y)K{VuLU7MH5~Yepuxpi;ZmGpO{HZ$i1}!WG+^e`F7XPkR
zREhZS(ygQc$zzk7e`*^Q_?`WWnhBCushwt7oM)Vi3<&>|Cb6nG$!L~FCG#*!nw{*a
z&$SA5T^oe^nJRyV;B*0sZ%YS#=>2Sb3K2X`HBgMg?&DWs{R17{4n*U5otcY-iQ1BC
zp+a<+9BQ$uVq$d>t(aoB?+ATjvwmVLN+GF{5KVE_xsB2A4tdR#qTgzYfvLLK97g2<
zgK3)DB|5~4L_1Lce<%j|1X15)gS{;4YDNJ;56aesmywh&N1EPC8-9gA$#e^cs#s(z
zwzDPD3QK}3fo2aWg{Qid*LEt|7j)Ft)^@={XsMo%d#PUYi?FMU%l&{SNAq7WKs;9+
zY06NDzRr5-drNV*mubFTg$1dH)mYCau=^42h^A<!q}t}!Z=@HQKp>1SVyRWAlvqSq
zBt3{Pps83zzlqL`VBu-V#v>g?^{IP5<*(f!<rl1w%JW7lg(QG;6G^bZ??>W;oR*{u
z89~S5Ur^ANfgOpUmX#fN1*gvts6n`2<RlshV?Zx^B(P#cQxo_&o*$*gnUMrA@hJ6C
zw>l`;gN~+WM;57tVc-sdO%=PGr+%yb)loAZSoYRMNL-3dImF@~18J+9%9Z-2<nwDV
za2jhF3gi7LsHECSF(8lMC%uv~u+OSY_q{Oh_%N>q2sNP92&e__8Gz#Iw6S?+l7Ic@
z5%}%=%oX{#FS@aqt+e77O+bWeNg|A|C_CrkRw>PvoO#}4Tx<Y7=aE?IlIK{>P=Fm-
zz{ht8B);E?Q3Qj$5bYJ9X`#aJVxN_swIxPV##n)5<f30-Kgy>dj5kk^T5RdDGka=n
zBsukE;q@jU;B8vV>*QhjSET#48HP?{WgVbDWw*sN6{|^BFWPIr$_G(_dMb;(B0X<+
zDTl_t_YJ6d4&YtVC-!{i8{OF{p>!7^U%C{PYK=;aL*kTjoEfK40Ff<|`lLP`-)bj1
z;=38}^N|ce96N4N;+L~Kqd}E^sB|_p-Vu=y_`*N|$7in^I$!2RtMEG@I6#Mt&)Q5q
zgkoTM;nMDiM3U)%3YSP?&ppYQ77K?<-O~mr-gB{S((<&Kitnv>nWRWB)8oE6C7dZJ
z!thWB{??&&iK4g8-nk0vqzUPhNCI>|-DBt|M;SsjzMveO#18QR+pvlJ?8Mm{8KD{A
zTaqJ#<D13Wy<rxeSF6DPW?-|X9N4YlerZHT2{WhykBY_9NfJ`KLTbr~ft#e`85QWO
z5+JtiELb8XWe$P&r3IP!ac2f|RYGCc+i+#I%e&PK16*(CI3<X{ol<+IP;@>Ae~L1G
z0YQC1DeLD*=rOJ-p{~}ARBOz5nh!VA_*Opg(p&jZEi`pu60sRt)jvbw5JR&{3H@tU
zsnj<sF?rbtl922#Y>L7qJcLB0p||0kx1s&*&?4}N>J~{aW$BUzz{;l-prmsQ5cTC{
zH%o3*cJgdrJFwmC^Fz)bL!cR-kxpi1!@bgPS-s&psb}`9@Uvb2$bJYpj=r}J3}7`#
z=ub)j4H{?w0w*;U@zf#$+p{H@lLZe+jvQ-h=t0E%2N9tvNN7_P1WbTflcOuu?<xcR
z1ayja)IG1N`3>x9$T?4Ie_9*PeQx~koMQhel2hn*HMsJi|6Y#bo(qIp!~ef@r!}0&
zvqGcpN)xI*e^GbU8~%2Gn67WT8?@&FxC2%7o485EQrX^tZ2Nra?m(ATmAiSs)pQKP
z*rwP}88xYlV$-A%z4Nrs#D$)T`|_DMpkf;gK;#!1NU5;7_o9fWl8C?TmNE)!;4{+d
z!hz7Z>)XO{FsfCs6m-eELm)J}<1ba-B~@NXMi?34WGKj>$cP}L+rBC<${nu#Y$?UP
z)(QXnjkoT5V?AP&^?Z|3r);1Cx$LOO!B5xp0qhVJzN;cpifz<zSBST>N4ZP_+r=QT
zfirkM)Hg?wT*0#)7ex{OB2|K|vz)V$d6qm*O{_wp9f#cc0Fh|mn*-q5Sr<)BYSCx&
z*^6Eb1RAy*XKBF<X9i$ra0IvOhv$-#g1s8;&IP6~n-sW=0WQ2i4Q8<Mq~a-_1>6My
z9)#yJdx8^7y}bbNSGoXRp>fXz8ZG@(!(ehkjWwsSe5qD*fbGuU9c+OB@f<#Iij;qD
zAFwo1K4%%|Y=LFPrvdC&Spqy;9B<%PX;MGd&u|1<ylTi9JEvp{0H>?%4eow`)>mGJ
za0;F1c@c^Ja?|&0-{{UTsAfnf!?xlXeh;mv4Pr*=&H|nVvuYadP#NQ*u!A;@^fb14
z_+yA)tvtOR&(MP4&D^FT5Z<r?AJ%d=mmZyrKSru09z84qqG@0~2-4RK@N}yx0R-*+
zi}WhJezlUGe|l;2Kd3y~1A>}iRGuPeYpX$yjM~6Q#AzuT!F#2W9+rAXMQ4#ebOwrT
zO2NO-OfV-XXokaCv>~;TTm>B&&X*|@jq6$dHe(3^zc;897Q_EQfoZD#EE-GQ(N`<=
z8&KX4O%v5s;uHdjyQ58C>ukCX=wZhrt~>r#Z}dh&qX_|jQ4jqx<mr`3sRL1XlFs@u
z1006Qm{;nIuovD&XPQCr2;j^G&TK$=3o3!n4Dd!-)r<`Fl{C*P^W3B@(y-7Z`d)za
z(`<)wWD;;s1n|21Z-9L^+Oo5(5wL6MqO+t&;>i3=xuSuTO&=j&A`n`_iUu<kB#sVD
z32Q5)NQ%#Lv3RtO-LL)Ous%JOJrx_6h_5jkjKyejS4-UD>ja7<5>OS6AXZIK($o!O
zUhVvJfM1vQHlqyG5lw$aU1tOG?Pin|yIa|XT!szHt`qsLiN$lGp%FOnYT3F4S?kSu
z_N<EH0{QIfC@^iH7Bny6V^#WFdSi;J%S+1QsN(i2z-Fr){qWZ+PY(ZDoN<X-bqqN4
z_3CjWGY58K|Dj%d6BNVJwib35OYfzs;oE`UnhH`(&`CvkFwX?p`<UdQMtTWm8HL~9
z^n9ijlqLeZAb+w2sSR%$GFO4cZ4`x1;4WL7#!loiC16RWlK5@%!!5`N@5?b;QF?kP
zfp{#*l9tVdM@)^zehH-Q(33%eeVVfe*gW@Q4xJ6{PAh<487IoqwjwUtmY@&?cCr?C
z#8B^=Qb;a7OqRDId)z4&pH56|PcXIOkO{uZh=R+{E+M9k)$--7NLPR45e5hez|e_b
zNuWCe#|UEZPpe47q?u^K94F*IN<XNMqYFsFkH?a`o0dWa5ZZprJP4gHTWaP35(E3j
zud8DhCT8aG?HXX}ZF>I3k&xN+h|FeFO%bfo1M7aZP8m?JQ7U{n8M%B)zvm)$sV$4m
zY4Wr0q8R3y9&c3{))&tNn36ZXi_Eb>LuWiwVXt-I(`FL2#0t+n%owZ}<V)|O!lXvM
zzyp2;^6Vi_nrJv$3w*S{ARgkyDD|$2pB{}b$gXWDtzunL(?zHt6~4PSWN+$^PS|rE
zSE9h`aTIiB)R>Rz8gqfi1N{0hx+tC9XQe7t1K~UfEtAZf7+eucvS$hi;bBBCfweud
z0;jM}K0p?Rm*j7^p>*p;B3`I429)OZM4B^Q)kmq#K@+WPoCk>8;CG@(#-;C&du&HV
zMfXxy>{P9RZ#LDGcHnOqx&S}038d9Qz<+MT9savan9RkHW}v1S`1whx!bWJH+=TJ4
z1iXh9N&uWf!tM+5hueu8;P@MN?0s;#{E#dRr^;>uVT$XS{5^TY4%8Dm<S%xhtZt~P
zG=p6cFRNRS6&B0gTTl(Gl3!~<qcw?E6m<BH-l8bahRQk6-CK^@iOf+;Q$XgJ1W21q
zT`h$&xo{`a!-aCyPE-iLme=n@gY#Dld%CwB)J&pm1J3J`3O0u8c7qxGF{Sx~6!UUg
zm<d@Aho*=k|B~s%8iI&bcIszRS?rfn-a`h(mP{<mCGR0;SSU|^?me>COY*_@&{L*4
zZtRL)R)$+68>+h!*0JgMbET)dVv2~fqUg`@cG<QI86ve#ki$l5i6q9z<-3qJ=5nfs
zJuwJ>8VLl;CAR};GI3rc{z_i33njw=^2S{#Mmyck4^%3Af(nBQoJUGXQLmFf*@f!h
z2-&?GO@ia)=G|zU;rH$XoOq_1V)%We+~v3S0<}C%TTbE#jIHvFJ*YptBDd{9`O&4l
z`_BNi8qJB{@IKTqoETw@%<M&Z>bYKgBU<6F#cOhmLz3|;@_@Z)P+VD_Cl0TT#J4g@
zx0GLBs25Y0^$q+)9Cna=crUWV&gqG&gSG74Sj}^$mVfC$AclUwVTpd0{KsCD3VGSK
z4~^CQy$}!AOd{Yq=#`bJLsfXEykQ@T&$?KG8E?@-n93KrNDZ(s9=>7dPPBJ55`UpX
zJ~4;=SxY}d_tDq)YpB7D{M|kxnXR&FKbp&D_x@o5JI`D&FpS-0rZ2D~)J<dYMETYI
z=p<CgHSePexJ_<-AN7sv?}ugco>67g4owig=n7N%8rJz6Fpb!hC&#y<g1BNg&WdG6
z3sa-;`~-HkjUs(n_Mr=u;1lxGt;hy9$c?Qi1@4iLw4w>>lY^-~{D-XSD|#vR#j?HR
zf&-{KbjfuGh^>AuuRB1r(<n0s(KNJtxG51-;WhGK4x))r*!ROhlm+3hvf&U}HBT-+
zMBI_wc!;Qni1QG#dcg3m2IZa+ZIPtf5ZV3x@D>8r&LDsEHg50dj3M3hOaxF>amI-O
z`~%VzCvjdMMVsd6FE~TZBiA&`DIcJ0m?01S015C}`Na>A3BD+A`vC1_L_f8OG&uNm
zxwyAH?J(lfs>|6Z4RDpAGSh3o-*(Hmu4a({;9hPb=!56-{CFDvf~|<4UzEJ#FcE%-
zeBm&f3eU>kHdK*1Z(waAV8i1BQ}6{9HQ02kr~n@vYwoXf?+Dh&VWb3a`-iF$lpCw$
z586;Y{7vqC1kHt_y#EN1!)^JyBgk$x4=FtlMDxZ<L~`iOJ31>-VM#U4N(8ysVZamA
zY@wX<A(9j`hU1sylOLkB@Mk&r5xKu!mOetiD{Q0KPi5&S${=cLJBoHfB+ocTUV5bb
z$}ys}3-aHOq4DbCTF41YE1G&X2sl#BKa_JnMs3EHc~p0NCDQbx{)$qg1r<=s<}scO
zHZ+>Vu_bc!aWsZJZT4}L4>!v@j-x!7E8jRy?7DQH`UFzKZeypK9Z~Eu*>n<>(e_3B
zWHrgF2g@^0qMLAvJn|FN3}4uXPN6jrX3DRhMw7YR1r!0d(pmUh2G2`cEVA0ke5;($
zjt)Vue6=0<p;FHLlpJ=+3qM8EiQE48DJp|+?JGEgq9FWME<1~Qklj3g7P*z*uXZm2
zyf2%7M?Q5Hjf8jPv~$F=4$6bh5oMLj^UfhB9JsIL9P)%|TepOB-#h+0FW+#HR3Zuc
zk}sk(ifbGRZ+x(QuFA85)Ov<YYZ%8h&Xz}igB<eoZxB~MoPdIf!Iw<9Qd!WC4&9>&
z)!w53#(`|cmr?D)tu$~!WoSDfR7-x>1Lu0cCUK!f{GU{?h(DsGLlSgKFysS|=#mfc
z<8z7SvtuIpLX0OT00F9XQJ1n=0@QD*%t?Ho(7mufsq|iGV%Q_PL{9z*oc{!E_leFP
zs6+Q72yMR~(M~};L5IVmF{v%>^v|RaP{~hUMkVlw+<KXaCr`e18R_dUxi`H*16{*S
zg-gsUV@MGv(ik{GCC*1Qzr<PeDuLn8Dtm_phdrXb<Rr)Yh~CQaa<}k0CI7cxxP=ye
zgt&h`Q{@X(vlnCCdzJXEV`4d#U|paJ-KGm})7YmZp*$j&sjYQnP?O#A+$$(*_{j|C
zM}-Yu<CH?pGWrh`B$806*@a%d4{5jZ|5kbj!O-QxKk4M<Li|%t(th=MpMrO8L(6Ti
z2bfp8fe8=s0$=F`%Wd{|gK@Z5e_QUnf}*+2iJ=34&?fTR42qjcX+w~P?m#KZ_#_qS
z=~0N~<$?~BV4jw{aC3yZ)?xQK@Q@TBc5KFPC!ak<zDOo4yB%um?QN77bfCQY(mtCG
zQFy4bXbgMW-IZsl@}1|q%|1bgK8BB|?qkSeEpUZKCBHQ)In*CMqB?TY0UnXN0{}Z)
zrs2Qo&G$$p?JZxRFSbO2TD?1!AaF)%=t`;gO6koo6i5@C%{Vj$61p=6X5^2-n>A|E
zHE~shPJv0M06r6yn$-8s_c8qUF`MU^^Hf4BU|&gRd#d=2UWH!_iN%DxqCDX$O0R0k
zWamUHY#B|LL6AL_ZeYMPzCW`)Iz2QY>=9j>5QcxmRy$Yl&#caz0->+7gbL-s;ylQ2
z8gglfDzp+5uLSM))qnQ4*O2luAGgRqUPU}Jp%hh#g^+`3`YKs-4Vk!>?!X_K2i)_3
z@IqI=<D3eZXUKi8A=^{$4aPs`0v|LkKCFuH-HF6=+0d4XKdF+Ow}-vf9yTDbIMsAU
zFPCI8Kv)iNoHmxU1RUgxto4)rz(4hx^}5_j7DT&AG)?m1pX6KDkX4}{ZapSzucLIs
z?UC%4>HN}2JwFWiUn>ctd|=Ayk-%?0+mjtBS6?StO?P?SbyNUn$rr98Z_JuuT8Bv1
z7eg-%`;*w0<cv;|Ug0GrornyY4CTI(&dAJTaYt5#^Q))p0FQc+&%C;L@~IwhVWh!j
zT`#YCo6#A*SsBi68Av9qpZ@raV@KqV-)0Q@=LEt&b_71bzb}W#62``L;ryb6t~vLg
zm~&i`^WI_1+;J)F@!QXAlc#jC&TM;vb?QU;jV{)y51(Ld`dq&B4rAvxeNI^C68_IC
zJ^WnOH8bhl!!F+cecLxT$-|qOq%d-<Z)PlE_qWS$H8Y9a{q5n8Prb53zTCCSl^sv8
zS`NtZTNn%1a=`ugT@P=|m0Or(?%{3E<E%T9yt0dRNAf<-`sp3{_!h=m{?j{8uv+f+
ziRTy{2XwJa-MBz-W}tTFsO)S97sj7U;1?~U_>-NG2%XG(eAR!QligdHc*ECQ!ud5_
z%KO*3;>YLwO`fxrv1<RexbNefm*ky<!@ace3C_da^7mUAQ}>6vpI~kLc}zT~{LgMt
z-Q3zyGbc_R_h_eQhVkd($R0NSEcbbrv4-D!DJT!v#zf1{z00V1a`pK0pa1&*O?aP7
z;K<bz6J)u38>6)dj~?*36wWV=Cr@}@9{c#z<-7hfRZiPRmRi2+34Y_rE`BoCc=8F>
zo7d!J+qx{~n!J4*V-5TMuKd|H#?*EEwd?xcO4+oXT$5v;?d1B*DtXFwVo_&SjeFcM
N`Stsnwlj0o{{tf28)*Ol

delta 9111
zcmZ{K3tSV|w*KCb!9B@jlMKl~0tqCLknjj(kccSJWCQ^<QBhh&v@wG3*ivgzd!%?u
zTh#iR9!IfyMC)zwRTPz4t(8`-)M|a5R_hDedU~v_ZEb7c@xOCDzy8m)_x@OunaP^H
zGqcuSYkl7uKKY$r8|D*7dPY07K)ZTXv^G2^OM5PvoWy7m0Np3VhGCp|CYRdnIz%<}
zdV#>F0bU7APvR6gD)QTLPb{AV=%IWZ2<3oK65xv?`tll3o&spAMJXFOg?B*vJpC!T
zX&&#6C>?`<%v;CprjZ~Bo~#7BU=!>%p*T}z^9%MM9kX#Z%})g_N}!*icJ%?=q&VQ@
z+)O6p7o0_$5shdAf;~#+3pAf2AKe;UkBUtDUg4X;){|uBvOBJKL@CCku4l`Xbb(t1
zM%gI@e1+*^cT7sdRT(dEYjFQdN&z~6pjN^y<$Gut-`p^x65y2zYci(t7@;oHMdhvP
zA9&JOjmJ)-GJS8!Th7DUbO3{!e`dAW{77K=_!Afh`gR9_^VnDv4DwtC20S~CDNIcO
zydxJk@3?AyRt=~&Vg)U-RH{m9GHJfLYp+b0h2zD?`lK`{6!!k?1Pyi}&zy-nlzdUd
zLcO830#%MsBur>Md%(?SfLdbmT<%$U4GV|pw;(dSgax&wz@m2TqfT`ezOLW}pmy<E
zU|PdUOyJxLfL{wny8wvC=P|Ms`MQRwh_#(|odTBbwm=5>+0%#~(-3IazsQ0F=vDuY
zuh2LZ0RgwiaT9g$EVZzD1FZsc%eFGY6Pdj0B;eP};KIVc%lV%ocllm6pz{iAslr%*
zlW0((qf9f|zcTz&5!ZPJaJN{U9997zMOpNWZ+f-XR9W`58DN#)9t=7s>p)8baP!2V
zi&a&^luW)jzpxfkOWju)eh{z^P}?XZE6rwG6}!Kc`Ry(!BVR*hvA-*xa=Sx7t>w0h
zr;q`v#9K(8l*wvGLYnax2}Ldh;Uk<=Ei3ZZctDYUnP{T5aGBVT9;EqHqi->4_@FK5
zw9}^`1U5WbTtZu8$0zHK7>8hBhwmy+u<7DkWQarj2c2x3r=jiYiE5-x*3y)otT@*(
zP^JLBZVJr8{}fqydO}VLJ(`ZuWzR?hg`ZKp65_L(nq~NDag_Wa{SFparIFuxnq~xh
zk>5BnMdsvBuLgLN__jP3o)o{64~TW8vSW=i9*=?}@i@f&A{HqIa5sz=8VJe(bfznp
z>N?~83^gyUbco9p2JPc%Y$HQVd#|mNH5$s}fp)Qen0Q3t@;sk}4{)*~-?_`trWgAA
zfRJ%J&A{FG0oTfbrMMqAH(o{JE!77Xf+E8b{AN6fD#1zFf!~n|i1Ys@_G3!mN^ucm
zgsixM833OXFEJL`Q)!qLqm*_WG}n4Tm)rQR4Fn78#s11FeRlTNO|p>F3~@rx8LSGG
zR=Zb0Hj~cyQrx9<^a0+RS#=7PcwHF}w~0S0`zw;|bc}It984%bE)KwXP$$m83Cw1T
z)t1Hzv}sU=OB8WagLj~eo>$%sNOqar1di*#nuh`0VKQqk^ZkG))btYY<9}e+c0lo-
z1d2w(r)M@#0mBLr&`-$TkVT!m(GbZoFZ{|l?Ot9#m9g9HlTSTrdU^GNXn2}LQS)h8
zS(zI1=Yf6$=Kqz>n?UPDj}G|tYhry)adqq%T1j#Q&PBLJ{5p1=!k=#Fre2hJrs_9{
z6{>OaJ6X6+$&J+p4vPC#wm!v9-9&GOOuVUjIA*yM*w`1vq3SAlNZg{H09~tDEnF&c
ztQp=F`?D)m(W>H|8G1g@*uN~iwP^BI@e+F%_7Ts-r72GdB%4v6bmg_?bL&N@5n!TN
ztXZ0PL1=!3q?n5dAP-*^l7fAM!B8;Rmtxo<;w{ZzWRDH@6^OI6i7_o+vhdG}FK7qC
zcg6R#U62!(aUz`4krIy~x-1a8_{@IMm2d<G#UB&O;9o?K&H&rRNxF1s6ra=CV71t;
z)59-2PU@yWct^}l?A7dA5v@Hs`aes=NALc-L?}8+k`6{UP$^7lbyaCwhit@Dnk1?x
zXgw<-Dvgr!F9PilExjNaXl=`BP%<3My9uZ<;Z$&6K1l_W<uN2J@Q=XLB!P6w1b<B~
zK1!4P(v6A>nGsVnd8LZx=op|29Hs^}9g*>Kj04q7kPkZX6!o#+@y`t^nN7fVW3gPV
z%mj}+GJ)l1&&g-|N7sH%(s1#oT0a#>9U`kIqf<^j#*}7PCv!Ituu&)QDFkNGuR;ZK
z7Wv=-l25}*uu%!9O?CHT!b`x$C7^a1!1LljMQt*;tp}9;whZ<TmQ@XN-GjGR0csWR
z0S&-8U3Th8;Q#hq4L7~&tIuXnhZqRqeDGiSzyy)M$p1u~WdungGgzq2<Ufz(N`?Ud
z1cs<c;AB?2_iI{)fN=W%7jHVqU%lagPXk)|L*D>UrvY1svOmbR=~Ud-+1+w}A*eHe
z;>O6+G|0EsfeKFK^m7?%QY0V@lRQU|;1lX90RI<24Z*Pty^e~76?8qAdh7=9TRmzj
zrE6hHF@P>A8TxqgbmXy>)C(Cl%_0@x9|K_(F_C3~OmK2{5@bQXX**N`o8w9>u;Uip
z&n*_(WYf~b!64Z<ehiS=eE7`-6)A|$6mu5wh@T39p!-o^j1rOA(Qt;A`Fvcr8rX^x
z*iIb~yle5{@qo{bRFx$X(59Bqz|w=r^Wb@jK=^htUdkepv)ScxG8uf6cs$QuI9Jb3
zPlTDhd+;=zB-qKHv&i2xlYLqP@Qus3D-oJtQrX}5BY=+c1`-K!A+~1yaZ&hZvD2K(
z#%k-<qr*Yak_%eW0rdv{yZE!&tDLJDodcj238})}Jh3#@5tV|)`qYe!m9a8DxbK`S
z=q#SVu9aK*WO#M>JypgHnr{Ny_4rZ({!t~X67T`>{nSxukEFGI(={pZ8~axk6C{rk
znrUguH`XHq!avd^ad}gXM$wbTL528mnx`Sx%GX715Wdb-`LlU$E@E2~sJJdW7((x3
zwdut8IMqNgGP{pkj`g>7^hqEX-`AO01x(SDR`V5t)8tSIRTa~#i(_Q7?Dieuqc-bN
zTX7nJJ~?rew%XaEg15<&rWgNITLMhgCCO1#J}{W(sEUL_r%+)hKH&DpBOfp5TWoNk
zWnJwg!0SQzn&|R4z~@CG;y>jRC9qr}1;cOAqK8>zE3va4ah4@{a1zZPl=H(P?(FJL
z^97xuP$*Ir!cXfzd|J=1iYU}~30fh0?N@kO9QPpugmdE+rfiwu>wQW*VJT@?L31aS
zIEQwd08}yFI&Yoj<#{@aIQ$ls88~qIU!ZppmM62XBm8k3MU%8+^XoUzi%lR9#d%q3
z6)Gd^=boG%!xhq0f^7J7*Cw#&6l9Z-j-rOt?MYwK2U33C3aNaI<UAL-8=RFw(hlCQ
zz*7!rH3TDM2c1oKQP7r+?I|F{u5{uhTD_I0#^5r=3>xs0fn5Qt7}3-;E`j4FsBor2
zCJ1<f>X=&<<n2Lc>t7~%Rs6X7LB%$6)UTmC&f2NKvcE1`ylu6~{TBCRNL$_EJTWD`
zfD<ur8AA+(@$Pg~TJ543kk9QEL+PX860s{iAyb`@Zzdq?zk?KSDsVhofQO<RO{~v~
zpEm&kt|b^4Uy=9L<4#d#OLLqt85bLX&-GRU^|9x8?ZW_P<NzP{Ax|pAt$0~5$mPU&
zd1zWB^B1x6#7bLAT$C|^CtzIg^X!`i6og6U8nM}yl`wmN#zt_dF9)wP0UmGDSpH6a
zOut$@WqaIk%~)9n=nvR!@vS8)0_X*M=t=>I4Kz?W>?N_<UL+rzwCA6o_Gy6I#h2{`
z%%{4uv%;1lB0h`wFS|9?kcfm?BXMS;S_TAmsA$M|EUD8@9K!W6;Aa#Jsaba1F2^f!
zd*eWrew4T^BS{~f9Jpklfb+xkhThfOU@QL{AmxJ&8yAQ@8JzxBu}X*_96VMd^Es8w
zji4+T;$|q!4~eSGQL^nt>|wDcbFS>B6+bIp%`DgV8A@R5TwWH1$2#$6PEC;@c<b!F
zt1u^evStu8v?I$?`r$~F%~#_K^0BGxNFT6`o6gNknYW4I+W`L5PojIV4HTKq6B~VR
z=QssHN}Gky{wRyitC8V*3~W=61a|9hJ~2PLG;Kr$o|J$eB?u<8B%9w1x|0dui9r7A
z1fqnhaF<%YUTn>-KvznE(6zhp8S%I5hu~M@7)Jr@7GH4a;K$-C4u0sFJp2L5z6*kS
zzKr#2Nun|?FQu;4O;oAPc$yEl(fIQM640L)Ko#E_31zZ|Xyu4(nNtYQF8k-~vM>cL
zlxWOJ2`-*M(96ETt+BY2gOE5myctlN!^V88F+aQl+!yx>5-{?>!a(wZwCK%=RG}Tz
ztRaQ}fQ%mkD~HKIY42np=tq>_EWJ_L%dtc4z;<)US9!lb1kJcx+?bOKUlTvb83fme
zvAOHv#~s^qe}eH8?HwOgNy70^BqB$#YINKHcwP(?bJSu23?32B^EvRc7@cR$@UEAI
zmGHiNtb~9GkcH<Ye3JN_R0C4JieIaHMp;`1J1X+d)0%&+jn;l<{7(S0_k99jbR88%
z`Ecx#JXvKbAU_>m5&eJ3k`>V)2K&{j2pp;Q{8bfU9RAbuqI5$~yTR$c0KSH*hELy2
zvsA9v$DUpw?jBvFQTE+B>Y5FMD0V|axSX0%P7#V>8<el9PYl;5eo~(Zl<Z#&K;RY_
zaD({Rpv5smQ_13K7k?g93+u&429JjNj*h`KFm^w0DU7(|5>I5_rN5?opG)^bvPF?C
znrt$%QDloDTc6H!FX}s9bLDwTyS5kp=X3AsobwW5<PGJm@<tTs$4*v^om)E;u;Z2Z
zdnL(W>;|>=I!O-f39W;GfD#bcq&0Ys>vzNwAmZ7LOJhkm5*lIat<c)Yb(R9HiY%Ev
zni_?>03Z@_z9#`(`@}`l%PPTV^Vy4+2c|UcHr8vv<F0JL&eIYQFurdC;T&vloVzbD
z-DvHDM=+oW6R$naCKKvY@&s@f0(c0XcMJgCOT7aD@K+WQlCE~w1GR?!rtx8NKus|_
zSZ=sReUL3=@E*31hj;-O_<+zqw+~oaD4(kw^mf4Vl2ZT<uPg<gZO$!tBhB#k$DIv<
z4zCJojlFXmJis|Bd!u_ep!JoPA)H00c$Ol;KhpFiTOQX-kt#$w9<~=6xYub}D2QX_
z?i}D*_(W~vw^VjbEbdF2CVE=BhH%}8ljNuVj^}AW@Mgc(@etmy10Rmp-duVtgx^ET
z=RA5?3<T3?5Nx2Y8{nxomB}DzA5pAV>h+SGUT|t@+s9Ns?EyjcI4WNjv~^8K&g@X&
zD2ZvxM$(H?PLIp@R!QfOFLVZqZOy<x(@Zcg$g9W0Cul>)267g3X1o4InP|M2<!&>U
zFz|b8<$M$T@5*7Ct-pY%+mrQ=$n~30{#UKj2@9V^DrRS#=`)>8cNKcr8A#g?f1x)P
zE1=PYfWNrEei`x%bRbnA7SGmM-(!H&Fca(KIwKs2+i8axluQ7wzQC0WC~sjU@DaMd
zMY{YdGBk{%Ijub3O&TbTi%f#=Sx7&|_Q)ru0{3(Pd)zAl`*fUTcX<n7*U`miNjJsm
zxFmm}k&v%<kuL@C?NP;JnF^A0N2f=14W<Obf)WyNP{$tBe04;hlfZtE5SWfXVbmCx
z#F0@QiA_DEa)~%fDRTy~a;BW7ZW!}J3$g&-EIrePvQbYQ{UO!EPLU3@q0|II2uPtb
zY-sKw_Fb2N7sNp$aAHaNz71I$EPD2WlF|kWSSc3F8La_r&vA)L{a1QphB6W*BNC~S
z(^Y`YQ#yy^&y}7$t~JrPM5R0qochJ8nu(6leb|pxO*cUaEbD4#_p|hVsv5o$7^JQs
zjD=1u&WHIX$o_~4zA@6zVJ#!``&*xJSV36|u=CQoZAfK!#o$;4nz|?ozmFqpaSl66
z>kxq@jY{R(rQ_R>5&kIM*oLxHFoi@c$)3&i@NH9z(Jz^{qeS~Gb$K%27ey58dx)y6
zIUy75*34v?z_TF@?un;%n9>QRpCHw2NA`vfSbQcW<7~3&4Ghh2JtOc%es&32)HqH9
z?S_Y#yr5H{juIME6PeQF?=8p2?pF&!A`HEFZ8F^(I8Ley_m~n0jihadEs1==zxl=0
zV+#qFz*EVDt<OUl;Jbcx9D=UDS!(A4k`BuQN$cv^#uP^ZcUlchgG|rdcspz{-50aP
zR9g&d^}u>erIQC_Y^)MrOG8CIzvmygNM&g<J2Zgb8s)YDb~8hoaTy@&11=w6Khfju
zO2bPfa{)$DXa_PU)Eat|m<oHyiO-lxP7^9TcW|u1`jPbN4m3D*gPtHm?s4SVM?yH&
za4ZCTw7)P3;wD4_kb5)^e<Ue(qRfhyQd|E4Wu#)e2SN5r{jq8L&f`iHc(I0puIyU#
z30+HH;PH^wLKIz`#lB&sDwU(*d<ZQw%$pfpnLtoF1^DPFV#&bT{yBm7agx+bCWcp~
zeLGQ>b&CnEhcaV8Za$krYfabmu_|+|iI%s_2PCNQtFZ*W>3gI;yHIiQ4;e4)R<40B
zx7L<j#s6gJLj04BKMTABZr3;7=59G)nl_#w40RjubF(w}P0&8G72^rXcpuG|0yv8#
z<z{KsE)pc_q#L*GLvV$3oJ<TKl^9aqWS=^?AEcSDp#iY3^vY`}2QgI<u>HAGy80Tj
z!ZJzLj%wjVX=XcmM4f6yL8t%N7Zl~$Q28Wud!;Mw$Q-*g9XKW@L)v7DG&4p>%H2p0
z8ztLrG#Gv^P2Y{iLP>gSH%f57oo-%6^K)VEc$|>TEjM+qA*FGmo%)W@m4i;>dc+q-
z*596?K-)oxkFt*bg{(QYHATPEfm@{bJ;)UWU6Owfx!<eO(miOHX@MIT#Wk1XHx!N4
z5d=In3tyCb`pISqI473Ah<8gr?m-4uybI*93JtMFjV;j1h79yjHjN?0&E7WDau{FB
z@UY!u@COROTRwI>f!cxd75JZ0c?U{^W28A9C|>ipof|Ee`FJG;6*!;pS3#d7wRE65
zI7w3OMKj<GseUioX}H&Klnc*QQ4H5f$TqiYAP8|ZZFvvJU~HGX`_Kq@U0S&h6~ql6
zG~#g(Qmeav#hs{eJXw}{>GD35uUhEEH{xXe5MGyOd?*dCl~VSjF^MDcJ&Aa&0)OEk
zT~zMxgY`m2^U%OIB+MU@R_sTXgardobuh$!o1lK$)c%VO1mfufjnC=prT6!v45*WS
z+>fTHj||4+)iX#j9`m;q84oM*UTIb*O3JxdietUSi(m#P6cLKCC<$&c^ds)ProfkU
z$S35n-)ZQl=pp*W!y9XHtkl^_EVE0x(uwN1yg^@0W9OR-M@O+8X8Hm<LDf11&yr@m
zj^2ZPr2GS@0=7%d2hdPBQaXD8>0p6$>i{ZDEOTR$?jPZ2$KnOa>^vJqnz-z*MW7VF
zFAaVJ+2AH=`Wq-6zAmkL15Hz%8cPk~-eHxW($8aG0y{{e52Ai>u;e*NR_bMG+Ck!q
z4btU<XbxI2-jo8W@H%PqAv7K4cOE)~av=Oo`tnUOYk{OaOrnttJWPB+Y<C!0J<;PM
zeb3h?bSVgP5!i#n@itP)T|w^HEqr#kE1opv9SESRlI-qL+z+I=&VHoQ$I_-J^%q>>
z9TV5=knSEvxiD8se+%*OX=(gh$OM;5PrZfq4=?dkn@KB$Th|wRNrKhCnM;ird$%{5
zl&620uuqO38Og?~K~Xs>H*Ey|t}l!))h^}%+^?MuhTwXRn@YpG*osJ{WTiz%h~ayr
zEl1F7ctKKkp^A*BMu$=W8=VxGg-K}}Yx<(N5Wh9WJVNf?6Ree@2=VXwNLdL+8YfE2
zyHEoROZaV64@;!y-X><bEp@()>}JbDW#@rl-cpGO=HB{RZzURBT8(p3Kws=M;AtvW
zkfPo}qHNxHyh>X44q6NEO8ikWez7$9DEdv7If*?dO+JRQiPu&hL%UNIL71;qc1<!C
z$Fto+@mTz0YHzghLp97BI}E?1nLZx}q$$UVdq0wv9Y<4DWg)2LpRZ^g(8yzjiu+ZH
zeiwBaJLXgU@D+vWYyB0uS_3MehRtU@*=+a`lJJ&DpS+7ElN%{MfePTO(xMY6ANG^>
zo*>J8c<1#KNDli<oo#l;vdg4zx=}fuwV0c!BH;XCNq7(4gma|qlc)_gcYbsdt%1-f
zg-)TF+I|ZuQYp)xz@IaCerl7zYG!h8O1Dp;!?0M|c^dhlQu;lz+b@my06j{g`S1s*
z9JY1RXHYDJUr4&Ms6V-zinGWq|N2GuV!-)wx$V;WvuGmxR{G^ES%Ys$spp6vE2PqM
z$OXrCHk?DAD9w>=(b~Jt{{;7&{y|7ha_4s!P$tENC&C#Mp=TxGjwud%lvW#_4`$34
z1LZ=)k}81H7<|=)$H@!*boekuBy^Yp7*EM%eA(4S_)QuVL3wx=AQB><F9Fv}fVFDF
zP23++*u=3Ps=_NcZUq<K30*rODYo~e&Ih;|eMyk8H409^I79#lROw<r&SQbHoNZH@
ztz7s{)E`oIC+v^9FLmVLD%f}xbbU{p_LD07eGHLZ-^ZLzhIpn9$HZYmVeRyHgixrY
zftOJ!JSH_=CiW?iwp~X0hO6$)TWFwbyeTUZ^2-@Q2n8AgXLy+FzAPB#Dqc?t_J`vJ
zg~vwSm;L0x<-ITe<??F3!0Y6k7y-BZ*AR~^V5)q9YSMpq@0a86ozq8B$<~F+@GZLV
z7EMBuq<Fe76AC%AshRGEC4k?1AM3!2ga)6=cD*~e(QBMDSlvwjj)KG{N)_Aa<;Ytn
zfx9R7j)CE8ga45Ct_{ZD4Imw3pLZC%eG6J{c|E`^xq%6f_X1zp1<Nh=p21k{)!$C`
z4gyb%3F`ro&3ceA3F%)=39|=5_#oI!KHES^8=+35GF`ZW(&fpiO45L&5UZ1@t0>uA
z*LTs@7*)t=_c`(SbRb;ainpbmeV@D=O>Vvw-Z03!K`Ob5^4Z}-HXo+&!}8+E>=}0i
zx~ZfWt|DE-<h1bQw0Y#S5#E<Na&Q$830(!i&X%jWn}fK48KjNo3k<~;1qkUq8Kj2i
zW`wVod9Rn<j6wnOg0T&U9gs-K0b5=Ic&l1fLAZ4V>Cc`f0Ow3>hIQ!eFXOo{<G0Q?
z=PP*;uxqo}0ZPnm@yh%{xTJ)Lw?yKvp)7ZYgMBhiX3K8941(+jSq28o;YKjwd*(l6
z$vrdv8e8q!!e4TF^9u8ax=N|=hUk(F(cD()rE4fDwWAP1ZhaiLQ?IEKNLq7JG3-e_
zo>_*-Db&JD{Y2^VHDuCu^aFkpRPOnJe>T!qIj02Xd6N1%vJKle7T@g)e9+i*L>c3|
zt-$qccxS~QQb~?aM{PVEH7d}QVLGF4DRmIEx(wh%O#*2wILSL;>**1|Kl{4%Q|TEp
zA?hTVHr0o}llERmR+(YEwOhJ&9c3AAPh>yI;+|LNb>o14V`&uS1G7#|1b*{{0c=1@
z?;+TyPzv^-LRc?t>_J|B-8hX?AoC01Rpb5;cGWll1Hbb-Af7dF)Jb0V?}l%PFb4H=
zY%d}kO*VNa+pBOKPdv0LntSz89pKPF^0HZ%ME-LCXrsvM3~Reo^$MdiTwD>&-5yOY
zXn*v<D>l9(t$KwqjCzLr3b*kc_@I9C4=_o~=(IqWKnJGQgyv4l&CSzBap#h`)$5|U
z<;i3ix#q!P?Jr9AUSZ7I_7|fbRK31Sa=l7aukZSkYVT2Lc0{%J=$}+eE=e!H%GkLj
zmx$_I^8a|so=eiFuQFNMo`~*$8erKLDXoo3jUxN9HpUWl<u&QiHYP<n+s4F29c`Dk
zwlP`U(e~&E$E`i&mQ352IO&%*My(}h4_?sorIfObN#lCH^gO8g@I~+av9q@^DlIvC
z@PdnN(&}xDReQ1RPpY+dq*L3Nq~zzDDeh=5B*tBPXUK!&ubh|Ab|%U2#kOc}Q^apq
z&X+uR&2N%tJ7d-S)-?1%%`ehIqS5}c=}(%^_DTD<GbZh4`~IX__PzAwb|xi??C}o9
z8g*@j<l4ccX|AmZJ~;n3gwQb-?KhwOFB++L2cwY=?O;@zZ=@*?j{HgLiH!WI<4>Ar
zPu?Ht*hwt<?8!grSN2F1JBgdgzJ4dUy^D9GO*<J=WdBa&{Mt(CR^*)Q@h)<HXtm_p
W#iVEtt*&`+HF3*2XYOL2RQ*3B=pG{g

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1978_12.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1978_12.parquet
index e4fd5e803068db1a8e7ec2a183f0253f996382ee..7945d516e36377cbaa337b53264e0d192e58e63a 100644
GIT binary patch
delta 4825
zcmZXX4Ompwy2sx=?1g*J%wEjwnZ03#abSRf8NPBhG9d=Z>_9$b4@X0f5O$^kR8my(
zdOAHfky2AEO1C+MW#UOolT=j7QWH&&rjmJ-bF=a@PvTc6>(<Sbd)DZE?sE?CtUc>l
z{MP5Y-t}MaIvk4b2t`{f3dVZ`5H|0igxe3M3wJnDh!hxrRokowG}&l6U6wZeOTDjn
zWJWm+b0;z|ac-nE9wZLX<ND%^nk2^&o80JW3<jkj6-j%71^KEmBBptnp?X4HO`bO@
zV!a)d{w7+U(TD-`)YCAd&|{YUvg{c~8TvcXYPzD<nN>9<Qfh<jT1gD|efeC^dc8}L
zr8>bNxk**WtABq208`srw$QS_u4(Zz8sMBccMnMGp&&r}n$Q55sW%oR7dX-we{-Hv
zpa{QF59KZZX(Zqu*v#d~xf*p-W7&MCN}ZrxG7=Xlfs+MBIw;{>mi+eo%sz0T>h<5i
zpPvs@6J1@#RZA~IV5V7C^5nb;QZ&?az|O9Z2WKSStCbP~jE=WJ5r1OA)$)#BanJIU
zmVt8)t)|ennca>uN0Y&=$npf^8+9JqJ<Hptk){KBtYf>t2A<bfkDFdoa2)NYmB34u
z+~qp<3uB@*F+7Mc73GT?v}`2I;6Kp5KPrOXqm`ZnY6=p;<FIA8L3X^&xMcU`vHCo}
zTlwL(x{bERJr6Ff7L;E{G45c{JBzA2ZU}m21%vB>4tlKR>Q&lFb*Tnqk6gu}G|Z_Z
zQ7Uy>eH>+_p3pfy(@Q~D@B$QahFL*RmK&eYK|J2h=@ZEg^lXI0^5r07UfqrC^f@g)
ztDZ&U4H`Ytf#N$r=chy6QG0|sCbGKn5DP750vlWYgj==v&8VdZKw16(D4RA$v`llK
zZ#h-^9!TGXH`;lOY+KP{qm&#jv46Pm4eihC&_Q0w$b5sAP06-VZ#%xD)^+On<Lb|m
z8O@3I!$$VIr6KD>Q3d_9EGNILH7j}56N+nkj?j4%@J6k1H0^hnH)cO1wZp-&dO4#e
zvFNWfip`to`-!$VELy6b*BX(DKg79L@^5Mj-=ZFc8det=ZeOn}IOGcg{)^sm#n$Kv
z1{;GOH^=b18LAwc3={(yvUC~>6nmOhvH-O#_g6}a5APS1o4bqYNv9SO@8tj6fKM7h
zZtPS)9~Kv%QuuNHiKwC%A`xFuGC3%oGFL7JzwdoY+6D#MqS2~>v5E=lL6=<Jm{uc^
zL9_Tij?R+uE7dWKvw4}ej|Fczcrj4pc~M8PsJ76#;(1+f4wPO8$pnGamSo5h$LpL2
zz%}{l3s+DFJ#M<COZZq|sc#v7UIqS22Y&1GI$~^oEdHW_N<;gk`EZk*Vh+7Q7iTQr
z10LMVfj^<bU|kV4>XPN{_GbXIW<#?-uiF2V*sM1YKg;u%-hlj=sKD`WM6`jasI5^H
zP|_eDYdRJ#t#PiUiLb5l@*sTy?5*s%XaN+L{@sFa8lj}aP0J;vW5N?QK7tr>>Wjui
zvU908A`SNCR+w9q0=~~^Hm7YE#isI=W;~XoC}{TJVLC=}$@otuegbC*2D4<TCmB4G
z&p<kVQ0p1ZN&HkDEy;qHHK1H}nj`;ms{0?-$)+0@a}xojw|H=8ltK2$io}1-)3TIr
z05(UOOtGdZk$?^{O7Y0jewNL;5NjZdm77Y|@t-gtTlS##3;$@A=H5H?m2HvWxLQy-
zB^^*anLHPdFfk{G>lXk&gT+7KzziIX1D{)~h8rJ&k&Y7>F_Z;dC>8yJ59{q^W^hv$
zgY1^wL9vXZ9QE@+vVvoWp{UxIQDZf80|uCH0O?r>T>JHZeBI!Tu(z3-^LOc_lJFGp
zdyMnp%sB0U0Ez&GyZJt?Ma%Xihm00$NwF~=MEAGmJcZ~@-tmxFjZs47?^$pC4H&S(
z8Y>Jef$k-MXR~+<Lo06noRI0DoXK7s>8R~r!fq(6)e{rRc(?_WmQ4f6R8ca8evg}{
z2#Z=eVskH|=VE}}ETkZ|<p>+o-UawAy_wjnys<|7#?B~8eCt$GVURcz+0@;>osq@_
zJdq<J>aC*ZcK{m?aRnDzdLW4ZCbS$PqWK}?441NXtXu3fS{n>K*TvZe%E-878&yP<
zCfkf1{4hLO0GUryXX+BlKV^Dmx2#$8=Z*wW*fLPCq;+m~<Vd+Nk=z@J&`zx6;h+K`
z$pgx2P|N^Pj%@shqvcJgp;%`zU%{qc=ZTwE)vI6Y(y4_it50lx$4UeUt;w8XIii%T
z13w?jV67E@Xf9m_fm5`SXkB3ycJOmzSf4kJqsaNSZnbwbWVRYD2Bancn>i97xugwW
zP2&E59~Te;&-80(9vMe!0O2*r;m1bf|Fm~Jg-fHSpRU|XyZo5~o5_LqR)pkhrUp2;
zpEhuiT_XJm&Mj6`V(JlYd=hx$^<-%iYr+dOHWa4u{W>WQ@H)==N}Ln}#s?(<0=?oH
zmiMJdwy=(l&j%#X%Jlq=!%*}y7qC<V8=opzR=A5Hf|lEBk}|{oG}Fmc%&ie`IQiy>
z4U0a*D`NV!xHZYx;qZw~HY4wQ8X-&mgSw>=RDSqi6ioF_prp3|JxttRoZ{|T01D#m
zHX}}sFW{J<C&*t)pc_5-QNg`jPwplkg%B-O!o(hqOx;@vQY;icfumV(67Z^(ngK3c
zW&_3kgq_j5<wQl0h+zHZ9wnGsZsPIGo8|m8Zs|q-lumjfytU2;q~E#PM}cqx$GLYc
zpk<HvvxOm{1YXE2z1?w<+;NUOd<o`um8Za5aXSl>*lYD~GlkRe5;Y1BH{IMw3Fj04
zHO%>C`@?uA<!iix>RyJj#srX5K;!2nX`~ThL*B{*sWd!vZX!)B(ECXg&@fs<B^IrW
zkY)k!1=m1+^$fV@x2Yox(MbytJMS?#asas{34qIwMywlR9IpN=N9NOcem;-&U00tp
zSTbH>XW3+;X+#TVc_%RSW5L%018LBj2F|gL*}iANXN3ATsJDXBt!^`j)HU_6!B=&M
zVMxdjNRU#1{Rxeuo<=M+MREdPq0`HOG>MI;+(g>$U8b000<AICjm2p-!C>ws=y?S_
zNtH=Cu)ZqZqe!xnB`co5Aw7w&vaI9OT4TEOhHkwZPsQ~kpq=Q&B(O|EXQ{jVV%G6r
z>BN&sUp0-ygQWVG0p;LRj6JCCH#(_j)GJ2sD1j}C1A0P6?qv$4Q(__F-;Fx^L*&3D
zDr1r(ah^tp-Yfc}0UdRK^a~jMf*rAr0J2&+^>JQFTWJ7UoW~P+m;4^%NOx?edl<6j
z$8czF%U`8hD2yxojKp8=z1B?i4F4E)R(;Uq&~>pOJpm|}IGH-#WXqX&I~)KK-$p*_
z*ZM^0y|;|2NgyqQy6*gbT1p7-rm@7igqCaV9{pL7GWz+a)K5$qy2~+)OOfzLdVEp+
z#WaozsE?Sv&8Y*H_6LC&Tv$$G(=%0o%E=1K?+Nl7T#P)U!X?je8|giNG%l(i6V}?3
zR=_TU>z=B+irgf5y&7o+;Ma!}H9WZw&>bkAIlO@ZXA{WwZVScol{`X%t!~D_T1!m9
zLKIWC(6N*uL1%!8;xCaSBE7dl5Yj}}CcEw{eCI(u88(Q}nl-{eCp2^d#CEPXkSIct
z*$8`#I!3V4-*X9yIz>p*wVIq>XCFQm*S}hQNr<7V`Aya8Mj<xIo6AT^VNZu)*kPjc
z6ui)EtTHmhisYZh1jOW5sQp5$snkraX?&FJf2y(8)bj}_^RBDWV#fG=M(gX*4TU;X
z=9`MZg6AZ}_%mv%OVYBexyj<CXi8|GMop{=>vQ`A=uLrwy-nG1y_IltB%qNc-@iUO
z)rAkL2j3tOt4G|!;{Q-riHS+ZIB5ov?h8}+Mwec8yX2xwO~YuAio?@FFE!0|X8xvr
zBU-9H6Z9lFmHG5o{44{0JSiEE(~ApCcooxEqeUxxbx08TZ%w=`X0{sKhta~KbKolv
zYpD(X?~lBB_HCtv56%Co3~RSnM4L1jt%*>d7xp7GXHyfkcLemV3_mC?pK3bANDly7
z>zH}}$$}EwzgJgSQZ2W2X20xlkw?YwDDtSdFpWg14R>(cF7>qKfyk>Y`0eXet=0C#
zz%n3-TXwcI0v$e^2LpHD<{e<~3_J(_pUyr9m#-05y}4IYvbS(dfB`*jFw`A2xcnr-
zAJ`Mow<iMq4ax>SkN)WMXlg)OqH&six?xBRH2kWlZPur$PIXx9aPv(eph9rqMYQQf
z)G~~S<^=q)niIQ?-ko%2w|XNsM?)v8DK;ndzt5N3UMROw<kOHZf_(G`<u=syaz)<`
zN_;!V0wVK}7r<g8b^}2}jF`MR$0q140$5_$z|^|NxsQ(+;Sn_A9Zr006D77Lkv5xi
z1~=*s`8Okr?dn!ayl5shx*r={<F(Mj%_zbA+MK~)@BGk5n~}r3cm7~<WohX9aGNVj
zsljMn$k~Y!g>@^3kiIpcnog7;^sO0!dfyGL45Qw6hoEC!p`&4RtZN9`zdNLPpP>D_
zhoI|+LZiay`k^6c$BEGFFxqiq2)g}c=$$aS{pAp}?^5V2LFRpzhM*-O^A?mUF1bO-
z9q!+D?AVRaqgznAa4gLJeKWh-La%N?2^x~`+Jc<a`=N7Nh^^mWsTo|8^uLw#vE<bu
z<XCs8aw~EQ$GT$&H#&be^zv4eB%Hr%8;lOW5~8*tbLi`>$Y4JF%J9K_?W%3aDXwiF
z0xte}cxY?5rGFbTilj88<sSR}3a#6aL0G%TKDffZ4WTWhb79|xxWVY#Pr^tD4TPJ1
z`^gY~=;(c((?$59qeJ-l)AzVD`~Hb|rJE8@vc%udpB_@R>s%<*g&acHxgqGMmqN$8
zP`rlZ-*=&8&8MqF72C;*)3%dQKV6+Zc-FlK!aPanJun2de|5j<o8hMIUk%~6uH5H+
zVSek%5WfG%`#fp~;ro9a!Z+Q$&x^x+)7>Hb*5c629Vk&l@)bK!GPO0dYX?do+1EQz
y3bi0a?<A!Kp_HAZ)E4sZL=KY8*@;Y=lWRkZccOUV<l2nE9xYz|;NzW$6aNVc*5JVa

delta 4963
zcmZXXe{>Vow#WBr=aiZJI3$zIv`I_aG^9=F56A$q7FtQh7TVegg0xRc(HL5xKoR8P
z)%zeVT;!Kl=^)XFK()Yv%g<6k;DS^HK?SAv+N+2NOF@ulA9`I^d}4ej@Q;gRW!9|8
zCue5Q{_)-WbVO25MN+NfeG@za2!HKH!p5R};kKS$qzEj);yu>mN#UMMknyWJ|F{ta
z)d=R-uuwBMMGAnV2Yf;|uAqLf{bYvR>}d{#q!3DBJR#qxF?G%3>O;drA(!lKZXO=;
zNKy#^dt$kvZel}yiC34rDIJvSa~OG6GXc<{Kf|oDLbDu@WzPU)xV{a)&WxYt95%Tw
zB`meU$e<*~pLgX#&zm}zB1;XtL2}do+S{)_0YK2+*6oZOXqfZ-GfAL#7X5Zi+62C-
zjDH@EF;#iqbdX$N&tn56imyue6^$-#0ciw~ey*?$J7359(O9n3siIF9myE@wRl$#a
zMkXYY`#2`#lY0Km4ZU3Z+ARo_mI9i`)K%&0q!+-~YnGJ~xn!c03XOW;o@&ejX9_v2
zksJU<`vy?NA2@KecB+Gz+~&RZD*HTxTao37#&r!I#y!W|mn6*qJf&l|zy+7+>K>X=
z?>mFvWt8BHmf|)och2aLYT|bmp2Vf%TN*9}X0^To%Fg?f`9m6MCZMTU1dlzVzzwp!
zlclb{@}RCH;8uS4S>3}}Gh4u=p9ab=Ue+B7dFP;pGlr07PAIeqn2^U>seZ>e(Js}1
zZ7HfA=MjxM0uMpG>SK8LkTY8Q2RbR_3SEXWy<twsGu%ziXd#R2*6SS9?>$?=QN0>u
zGEn_HwlNnp<O}s8o?!T0i?yKmk29tD(D0!R)G;Y_bv_<sfh^*nwJ*4-^-tt^4Ou&I
z#XF#^&I4uJrsURX?hCD-RBi$3>v)izr)bE^!$v75e#Je0`Ew>v(lG!^Du$KnG+gdT
z4eGRijo#eB@+Z}sDFxw7TaS_Zc7<vkInZ~7k>#9K8naSTH*s8j&k3evBH5%dj$#7t
z>ZXySr6`<`b#g(yqx@qAr<c_DZ!#9UMS~joWFt25@9N!a_}4UL8_*+A&uN1eTZh(n
z(jNkHTxY+Q(d-F@nnND9p5@<WQ8m#S5DNve^eOlhTdqd309sZ23zD+pVME2m``8{A
zT1qpM?>CSSjUYE~RZkDdoZu)sTzYz7>s;@W6wLc79u0|S%#)snfPW{F_JB`Qejg`@
zl@BB70f_3s%Q4bOFb@I%|GX2nd|Fg>&3W|_wPiTpt7nEwr4Oq`tTVh!V~=hgl!7F0
z1$YUdTApYThYQQ>?SIjpErQBxAekVT)0z###Y(O79dJE5|MDf=!8|m>(j^=gIP?vz
zk@4goS_oMGqNQ=*A0X!%Q64@fErNbp@Or$C8CS6SAb7}0Jp{7H8LTS@-v5=QGx{8a
zNz7yjl+*>D7N5}>Xd;#bD!+$P!@%I_FL}I~9Y3vJQ9#LqQj*lsvf@$aMusM1yO#&)
z9B@C5oR3>zT;&%Qa@7bG9Zp8Bs2meF*!Uo3sT!1zPT}UGOez_+=dCcmJQw^Y8Lp^p
z0OE%5)n@WQSdS33>vs*SxMVWd#8>GJq0n$y>d6Mrqvs)?KcexB(o6hfT1JuuFK58T
z+;n^Cl^pj?>%*q+FBUrh=Ldf!n+6(WkE}@iKY2!$N)5mjNsl7VG&KeANmdy*qVfvI
z4ZoZke_456#hd(L7G%qjFpggSA8Gy_F__er0`~8Glcwea&Z5R|9fS4Mo%({tFz`=u
z<QfUiB!mRdT9rY@2VsQ$V*)I)fC~@7{~_<_Y<@Gik;NdpWp_yQ>yf>20Z3M`?=_Uy
z`3vf;T7Ap_iwqzwgy46-{M)}DoZ)2KQd4QSR;q}bKsK|^2MVp4m;j0ZWrz858jFVO
z$!RlMtQBL8Ss=Q<F_$P*kxI_26(<m!P0iE~<GhXEL(B^6tq@xd`<DZm$&qa=qqzA#
zq0o+;IbMB=ecJV9+~%@rI_f2vm@S~RZjI%HQF#u+|3xMVOItdu#h3BoG~l)ixtMD`
z#id2NfJAj>>ayzQdhwi%Rg|oiQ%z+dnol&YfA@E%7_-U4dMdBp$;_e-;Ibgoce%9(
zLgcK_dXmcN0oED68m8skVyDsCVCcCa&NLt+>y|T6IS`rTn@Ji^#tR_x+3HkncJ*<#
zXHM(-rEhg)gThsTLOdGhc6*Uj3^lZ)IO63++O;qqW61-`>!6qc2HHoGB75r_P>(q0
zI6lA)`I@JhI(f2sMw^eq>i=mS;q_K3Pk4RdB+F@~;!Ozf=`2B3vddh#3WEP+6o++%
zRp{oYr*VF-O^;}8TesLcnhKX1Ee5P+16Mc#Ag7{@TpX-_oYV*ygJ;GKynsGOY69U^
zDB?>~$<wxur-?6h#;22xGOj?Oz!g$8_%T`Xhfz!qy?O}tdQYR+GH6-`BYn~|aBi{w
zo;~E4e!^hzrt0YX)zrt|(%4ky;4f*VOdzl6t>H{54UG3m0t8Qry&PYdD`mu`c0ws&
zfl+3Zb_~F&&wWOude}PCx1?+jOEs^!$0QZT^J?}WcKrNsz4*D4_ZpVIOIp*eCy@n%
zjU9Hs7|by8{--duln&L-_!X7L3&i)NcOsHD0Di#wfesYkW>7G1GZ~4H<<qkvPl)f$
zW|}=@tl(a)qdm-zFlMAlQ1iB)TJD_$QaY49K?ZW(!N4!JqFLY~#f=%D*hE(6mK};9
zQQi90eZOF8y-F(CH_G_O^()TvA8Dm!@$fo_N+qu7{UCfowC;m{W@L|e#lq4_gZwwU
z;>V8jv=`cM^FE?KSe*;=#lLY-%iU;P$CgbeE$VIkAfaz75^g#EI2YY`WMkyvC^ql)
zlMN_bd<{1&e&a@TKH%yFnMsY6@M8%y)<UBdifs&P(FN#`(Bw@*j`Ek1B|5I~zN;X=
z@+r6%tw>rq5v0fCZpD=s7b2+^gi2a2VazR<0;i8+w<H1ROL@$>A<gc({`Ki+nUVls
z#QDD+KzkWCS~97P2IGUvxj7j!RXwT}bG!=MI2Qap5X*y=dEgvtf6BiQ{6=VOi`%5^
z(Fjj>O+(mgzu7oDUh4f;;H>3W)x!oq`dN(`(wtRVogB=QxGdzR(s!p16P69GOrv96
zUVSK3{5ABfQb!rn(O9+Cm@h@No806H(l`vFR5ohCQj0I5-}qL}{#iav==d<$Omj$T
zd<k$7`IxoMRQDO3=uhewMz0TZNG{I=W@16{6@>GVScdtx244IjC02_HYV8i9(Mm_+
z`HBBlKF@$Zw1ZRt#(-eMoZW`4R#vU#m4cTHAd7Q(D)>^sV;td5FLV#UR;Vf-smJqM
z|3Pb`EVJwbI;y#kTHWe#ehT_r%`(}Im=n?kzJ>!D8E-KlS$)WqQM~dipe;B4Oh39w
zXU{vWkG|OCT996ahW(}287VuSUB(LMaz>tJ^XM*$$Qa<8)s3bC?WHu<rAXwEj(n+p
zW_k!sQJrRQ_#_desUWPT_uD)f+!t^;`b8?gqt&nI)8tv>UGglqkvZ~|aZ%%#cwl<1
zK<ozB9l>{%yXn_EG)Za)^5P(>lD)?O--dCs#U>V<^FX$3vmlOFc#MUNhK3XMmNef&
zoYt_=zJjG=YLIQ<|3#}t>SXA+a#ua@+p*_h{c`|TT3#8l>uxvLaa2d`2w}WFBE)t;
z(++_29qSEr>QH1Z*=971JP=%32KH+-aC}(}?P?Yyww61<dQG40SH}p0wM$G+ud|OF
z%e=luog<_%kMP?bQCA7+&eOHV0%>sEu`qyGO-zYG7MhI(MwZU!^haYhW=kv74}^4+
zY^IH#yvJRCy1CxeLx;r$SJYcV!GwcGYkO)_sTNoHpTuAx4cTddg8I6Oyy4cu9PvUb
z5~9;l&E&YeM<>GBT<{&8SD1Nr67-J%Jfh-%UmG>VPEM#N*3qHZBOc*Mv-*ta7_7;Z
zW>Mw8Je7aKrIXz*xx6r`X%tA~?j*sB^QJose^QT%mdVEj9i4w=-kDAovJfCuIpiUo
z_^gSnW&7$ixZU4?1(E;K#E%vWml)iq@L$Wn0)KT}g)`^?cjrU*SG#v3;gtE0qVnoq
zPpT<tXv_heqw(>0ug9~uHKU`$K$;C_>E7p)d7rR!Lx4BhXWZQlVMp|aI@dD9@*k}^
zAbVVN!(tgow=I{a(IL3$7HQk3c3Xy~T;(8O>rk&-GQy3kV5I{ZOX6jUr(G;hrb-p<
zfLQLm-^yIL@-5Blrb9{5eW+|qkO3We!qD)(!4;sJiP*v9zJtm5Ur-f$KlRZ2sVF8D
zBB$9uU;q|_TYjk&OmuRk@w61sg2Tn=S%IUyC#Ej42GJ(<l68>zst{Bm6ng=0djYo&
zpmIEs>{X5FdzkLQ=eyOD=|xGmo$VP;^z)LOjAc0)h#pDwNTvsKASVNNy%gTrjl>U%
zEFiL@y#SUB>QxYusDpEQi!ubAMF2}07ksjzdH!R=hIxb}@wQ%U*@nb72h%mZMfWcH
zX-dK(>F%QQyOH=K9dg8<ru<=1%AH4Djnr?)+2*Tr|3L157s3&BJGPr&2;WQguZ*0E
zui3v6-HUEUV%xDpxVa{QbZ&@vci?QHb3+1Jw>2_5j@E5WK+(OC&NzzhO+YUniF^@9
zFCR%jhu)9yJ1IK!egay2E>aOki_ayXQ|BWsadhf@0&4pva)2Up+cycQFY^6PJVfmK
zo|4-%Vg3-sXn&+&7tR-={RyjDUWzpD!r4jm+`bDt(e23FyQr^kuSvSsqxAmqd-T1M
zK%xgCg}bp+h#pA4chSyRWcqGASlAiMxEHNm9qEXpwW|l+i~6H+WRCRj#s;B3n!tB=
zMofEfj(K-y0_u-cQzG;qvE6&qwk?t6d#IIdTQcuOUB@Et#!=U?1hnR($n7{<^HBoY
z*%$G4QM9u!0kvF;Jk^EmLd&HD^ln>Za~IA^qUTdxI47xnUF2pLbwm5Q{Cmw>_g);C
zBPDyW!MyIh1b!_N+)JN-?TZ9*?E2l6yXZ>c*!2Yd@=tg9wK#wIrv!fMw>#W9@^0C-
z??>V(j+X7M-xB_MY+0nd8#|Kdd1g1xL9s}CH_oP~ecd=0osE1MzdE~iAHBE`aqYu)
e@xsP}gbdx-5qWGM&Js3uBrq|0V9h?P7yloxjq}L>

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1979_09.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1979_09.parquet
index 32552c282686e4c47083ea84ca7e36d0986d6658..c8ac1436e26cefb383d6bc89ad2d82489977bd54 100644
GIT binary patch
delta 1260
zcmZY7Z){Ul6aesZyIa!kwRipV-rMW!sO@;$o_4wxQkKd{%45rxxON&K4%4tYJF~^i
zMUW3-oE;;WI5uL?i?xYEO+PW?Ah87eS1C#3M)v{<5rRv$!61YvW^*ROh_CL8zJ7SQ
zx%r)Q?l~uKJZm4x+P~ge&i*!vxN}Y~x5RM3ZQu+5nN~XZ0~XE>mcPGYUvCvi*ZB|a
zyPtcJ1OHAKc-tbmf*uG~fWHF{nqg1M-n#|%R3I_9eV*Z-HlA$}EOJA%CaiZ5(WE!-
z=&R84?&HC0fCNjVS#FLH!O(uyy-WYj_J@aK)}t{iQa@cLW-UA6vMXzd8;<_BpSu}@
zuWX(hN;L3EPvXmld`z~HN}e^K{id)Kj@a(n$*+8<E%ETkdxo$tLM%4Ebt{&_u__3Y
zj||8IJ+|c<WA`&6B<lHid29#B{7;)~)ph1!(%xh$_LMk;?t3~J{eE{JS_pRQvgPQr
z&>!$Rd6s}o8eNH3hitcNtby!A=ssWRJ#`%+@KwJi?!Dci--7*b%=f}((!bTvvsohJ
z!GIZ)<7}vG;}@pgLq;R{-i}@Y+4voVKRIS*LW8)+PuAF~uRNsNsXu3Sqm9*!N%pC$
z`K40w)@H=e;ozqL<f$ytBWU;U()F4<;j&eP;Ga7FL~~#fg}R`*#bA2*`MxOX@9Of$
z!8X?rIyCutM;%D57aj91N8H~6HW!2CdNS@OHwD`R;us`p!F>%!Wd6d@rna(7QyY#B
zJ5L@%{DRW~{IIkw0N`+{{{nD20I0o1r+&sB`>hr&<mgfnaxZCNj6N}~!dP4jN!nS2
z<bW3P^hgo%2eq(3&rf5CU3hm}V2Sx(TWR{&H1^ulL)w%S4b0#gl%jiQa4kAR6EoP&
zo|!<}xKaAe4ED04r%L1oT~$-r8)KE)oIGv3j72V=5w!C7675$h^7v8-DHqjsnj9QB
z1k;=GSkHmjz=3^@jXqAtFERYzYluG`RX4AOX^TFZpmW9d9!*qfX=REE1uU{FQzew6
zFBH`FbKjQ89Q~kxJ?vbrL=>7WU^i0eVgc9b(v$STEN1DGv)IC=Cq?b!Bg(pW6@5f0
z@zp#3y69gY@#e>fUt!b{SMQXB-dkQTHLaFn?=6>jivB*Uj+t635rx((c$3<4hl1<S
Z75bKfJ?e3?cuUeBi??KEQNgDx{sEc9Jv;ya

delta 1279
zcmZY7Urbw790%~<Dd%EK+e>@@9NMM~ds)FsYereh(u|o>OG<=RW44&EMzMjpWy~c@
zFfj_0O;H0*YvOV2VK8;Z=&&-ihM=)=_F=S8UkGuHN_0&V%=TbI*aPTo>Vw`s+}xb|
z`Tl<AckaCxQvAgfpLncJC{80WV|9uJtq5YRs0R45hrZKSW4dv+?)BPnw*{ELb-~C$
zkTY1j52+km<U{^wzuO0AUNktYUJ1PQ;O>V}Jsi8l4<CeMI%MxW@{3k%)7-EVgWnZt
z4l!m{(lMIu;d+MHX1wcwOonimzbnj&>fW5~QRd|%JIE*+)x$r*tOd|32L{4HR|Ejg
zj~J&$j7Yvz$q{4CSQEd#i^L0&|IWLW+&fUSDUQ3N5c}A*GT6fz8%Tp1jp{;<P?+1|
z+24ro@z~z0=hdP1Fw1kKw-Y-;gN+bk-&Z3EqTF^f*_9r#L#&NFUpE*4KlyP#_h74j
zlI=UBD-RkQL-(Yow9Gl%X>`XMRi*UkimwExSvL!QwxcO_f*UV38$GEHeD{dOIiE)e
z+8a;UBX^!)4r2EUn<t@z9qCj@I~;7>>oH*VUBOqo@0#xDq(;N0c=R0jHQz$$(%X8i
z?;MWGZAHr$5*xKL-{@^<U!zv%Z<pWK?FNUlWy$V)&3hSuefTEJAn0|M(6An!%nEk!
z?x+au^8Ai`$Du2r)(!NXjG$AgHyrofL%?mid}G%?>Td-{PujdGnTOl~;3O^fKg7n}
z?5`&7FV=jP{lsKj$6-JDZpsy`NxFhKGReee5ZR^|Z()LV=(Sr|;&%dvJpg8_{MUe}
zl0Q|;2EAQ|ji;4xl^RxLxH_bS1sW*B!Wkvx=}TqEk1AoCe!hYo!uHr<PeJ>?-IwT}
zE7-{|jVntMv@L_1QG$juxCN!@Xa?Ja^c+&==IEyx>=fqatK=@-l1qiVZ#5`u^3<Ni
zb}^qcDJ7n!&&WjL=_>l;2Qz&(iv_xn#Sb)1>>;ucF$0kwS2dZpRFcIkHjwzdMHzT=
zu@Y$M5LKk0C5g+3<m%*RGAOe@FaO)@_$oFC@=6&f(TP<o38kA=QlMY2VjC*ZjaA%=
zmfu^G&%Auyu52)yqi9V&S7;3z#MxYxPZujbRrcv(l~0r^{(adeN>#qNTk-p{3yZr|
kUZ9ORyk8!(J137>q^ELNl5aERRf2w1UL}$nIs94uKXSoEmH+?%

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1980_08.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1980_08.parquet
index da8960319c18bdc0bb8ed5c72f404cd36eb805a8..d1f4a143b96cbcf1723f48a39969f001c936f81c 100644
GIT binary patch
delta 5804
zcmZWt3wRS%zW<*x6K3)_Nlr4;ByH19+JvTULNlc)1X@Uk6bd#$DitXOLvD*eEz3h5
zUNAReQCvhc$|$}zxJ5xhk%x++^4bMO9_zxgEO@QRMGzk-vbt)yC+@x9w+r7lnaSk$
zKj-}4k9>c~_CJSgug*7G|NJTv)_E(02e1HwM&JN8PE1Qh!q#VubNeMft^;8d%X9T}
zeNAu6ogf>4^`gbH2oeX1^rkIZswhgQ7wLkt7<O>@c7d0}St|U!0b-%=5nIFEG^3m!
z(ZYiq`E_|6+EG7%X&3;hS#+E1hGS4M;!ROSsh=ws!nCJz&_H%D?~<z_ty;*~okC8>
zZ-C#*#+Yb)W)BDlcpiL@>}7oN%xaJq0y|{<i-^5PKeoJ0+ujVy!ABt0K8tvCY)E>d
zWgIk!P`?J)X?0W6S1m5|PTFqG8}uhVtRUyKplmsFu8@niDw$IdFo+P{^4;#y+~_u$
z@6VJfb{mp1Kk^*qlRM~0-(hvm;(56oIw;$f(g>Ch%PgO|r=uFAqhhfMSMwdE(BS~{
z+u-Q^ZDwrm68GdlE2k*Y;62&}<?xjgI$i+T0hl-PSD~5R%Dw_o>@(vkGkyREe~08+
zIBWpWb@P`er!QNpb${XYQEykE{WRH?4hSRceo>wbU(xdoXb$_#g#>FvYw$SUQILL6
zkWgj%Wx>}{Zv9FNtX7^4jR-ecs0O~nMa_y_OBqoVvezb802q4O936ak30Mtg{I$W%
z=xbx%7@E*^4$CdDmNGE_bvnMbSmi`%BH+burU#04)RAr!%M0x|GKsv0(No>?o^1qq
z8SpK#<<~V9=|r$Bofa#ED(kRPB_Sue5~a<Gx3z0@TLKSH|5=n!AbnM=#_I|gMd|J(
zFB1re>GY7?Hm1Xh?@q7H9m1(rjvgsaPX8_UAyQo^kIFiVc_0jpZ!#b}JAJp!iS9~2
zX{%`XU6DK_dvI4v4pQ=4>ql4l+LYp;lS2-d-PvaVdQoi|4DIcyV=N!v({UVxVjSef
znfalgbLC;+ez{?0T4z6HJ-QVM>uvuU3ZLKhYbapHfqtX@e;;=Tb;Do>l22w;$WzsH
zB@&6Oc!7E|$OE(HTiy+aTcB?cd^HI8LF~!C97BDMq2narXj*cL)!VdA(faK7sJ(gP
zD~65-pw9t+asd18$AkeRZqbspoL=#kDx1R`yRiN_M$SQ#{+jM`R{B?H#s8#E8}}5I
zD&9nbD!2-1O6j*mRm_`^W8Tnjweke`_DyjAk^GWA<Qy|_umJ+=yztJw(D8Kkq<>y>
z9nhX`xJTP@1=s+&P2W+H1%eJp4=)t78V}PN3ym)5Se2eqScKB)WMLsQ$GTw7vUP>`
zuPa3KlS4oK)OCf-bjzuAIl^P}{#9|CUe7OL?%<O{9w&3n!UpvI!KcUx1HQ#748niq
z&}CHOFV(q@7&t10v)vqTommO;=B$x3qd_4?YmbY+W@?R%_O<cHal(wI><)k;?lnQD
zq5RB`l!Vj8U&qUh9Q9eLn7v|Bd~L<iR6=+k2ShHBknz1d>q<7flq0Dtj-5Ff_UaWS
zHqLq2;C|LU4-&}2B51^Ki1;Uyqze{+kVlrA{ewYpfGw=}9RP|wAKg6WNPlUiR4R@)
zadOu@{Dc+)8&0D=_!TpB;xzBgW1`B8JY0(1CmctuLb+p$*4d#aIRdJdwhHR9GRB^q
z%TY0_$Q_pUbr9R~1jKNym8tUfi1K~hmIol}KaZGbAZ$jD;wzZlOVbXT0t()vgQ#mN
zPm{@zPw)w&#6+WjAIR0myh@B*(?J)#(Fl4c4p7=X8l6747y?gjhv<jXXEfqEqdTwt
zgXKOTx+GWcgG?P0^ER{sS*~sI1N)n>xqdZ-zo|XuJ^=yJZJ}a>Ydw2(8OFT;w}?zv
z71rq~lf=!%Zw`Bo*JpEj8aKH>CV6@$QEDF!a%e&>9?atbmf#o*zF+HHV(6XUdovSI
zf+Zl&&sr02)a54?rL}2#Tm9S=CjZbw7rGx}@oiL$MBrSNNogHd-sbt1*tP6cCXpb=
zdG;Bpn!L|lxEn_It`4RQ>}%eI_ys`>5Qqi4iF^sTSZ%Ykgi$)tWMse>IPmfssO9IN
z`+Lr^Y-lZyYFO8%kEq0TE{Lo43I-?3Qg!alk%3Lydgpk~;F+HzX+C;0JbnK`eQUE%
z770g6L^K;6*DIy9&DG7QdO{7kO30TMXzrOt)?r*@6!+PvY}n6pQE#)mt-b&vM#@Cb
z>-6TV*P^-31!m_Ik=hsQ$_@yqTi999jcmd*nOQ0AJ=AAM^Yk`deYJuSW}W895n2;{
zF@u5a?>yk}Ekvt*vojHS5g>!`44KKfpQTQq&&yY6iIroD-0MbEj$iE`5ZC7JKZ9CP
zDc&zKiZ__b*?ku-vxtK%nakm0h%+fMH^8)fU<Wu&YNd}E;MvHdFHk=33GU!-`z-WG
zPR|r~OW6ve=Q7uOEj0p1%}gSI2dGOaZE~%hbwf;i)r_X0MsiUnUjR;40(}u+BU0fi
zZH{yx(CvNm*h2iY*mPibll&GUJFvl77liN^)R!?>Dent)_zOn{p-W_yK0KZAD&i%s
z!3E$In+saxoc61fW`G4n-FLuj$ESJXNlvI?ldR-TE`nxMBf50Bn}-CsY;>sHet&R6
zzl>gXLN?19QK~6)Z;n2Z2n267gq#L^++bLVr#W`FcMsa_V`4tMr?4*sheOb-f#{|w
z=uGAx(g)%?lY1=wP#Z4<8Gl@4i`FVMf6T+5=pP(PqsaYJT?j(Y_M3ti3So<+&D6j+
z#51Ka(m<fox63iWh1Q9(mWo0?H8_6HK{NJhQ2{jbuhDQkB-9<w6<@PDf5Lc}A^t5B
zsFb$s{tyDeKkHdZ@SB&>TCfcV*9>HCvt7jaeU7|gXjj7{GZ&cJD6eITqn`g(=L}>d
zCP2t+mC}{ya|P-%2Q9c<k58+A7|0A@hB?K<>g9oq4V9_dO2(-lOni@{;fM)06ILBm
z$#|2SE#$d8l2b$-HHxos5Varh2sr$s<K+T8*hDS+VIg!zIt-fd32K8Yxx}J16)Q9l
z$nt#8AP#BKH2kK+VfKWKfL&TV&}05RH!o0|>c*$cO#NKfE-T)HC4Xox$3$hvV;s53
z2Az%)5}F25(5s%VWCEyN#Q~;uWOFIyqLzH=T4jZM8-bX4rYlglxF0q<p|-C8{(1Qe
zAme=KSoe>z9cA-IO>ua8cs}m3;FSW-Q#S<|C-+qmM_UmbQa=hXL!^bC9v^#9TwR(8
zvLm0IFun;w1N)KvTxLHS%XG<jkLr_|d}LL}$V^EUn@<j#$Tp6=UyL8lmy9^KA~lt!
z@SQe@PU^>^d5?cXk3|>BVJUo9WNGGr)23!*Mkv)edp1I&JR@7y&%7uPhuxRtPb;V!
z;!6gDiYpvUSLtm5$9VN)nFh|>qV1nUskp~OlfY3Yn$i0dqr`%%5$+Ru-jzn+pH1@Z
zMyZ!7jXdKh)0ok->^*TYT!=Wa1UpYl@s+9jai=+3)$nLD`61VD0_g*YIvs9w9*y#x
z;==8S&?b}y$V<@&UF@(6vs=8y?#EmiS6N|6k`K`w>aO7pq}bK|z2L>u^CwLLbQY^F
zyp_Y|5;EH4xXC4qkdX7TL<Fv7J<yKhGVfYlLjJ2HvP|4-!BH(`y6IbfJ#^uGBYv+0
zs>CTyE>WeBKe_lV<gm@Wa`#z$r6jxrU$qC8Ky&<Fc0%1lTAci=VZ%w9N61#5=IE$&
zT-pvw-4r6{^rI~rKBLFyF`Xs6__8^^Pu%9kw-UU`N8Zrm?~Js(zoeZo8DfrOHtaDH
zFXzw)##b`ES5n_Ff!Z+%z|RSeeHG)20DXWK;&DdkL|gDyYxvN3KlI{Aq<wQT@)krL
z-FiopcDu$o4tJNKvb!4aYOQ&mo=z7QaUt!69?}Q$H_igIPgvAjg{Kwc@n%w7$S47F
zX9+}mV(hMP8{<tF=AdPCmPu_Bo`cTtR(p`Gk)H0IFLhDVb`kNkz?wh_NkwQ9-h!q5
zjzjq1lSx_$?j8k^1GL#tV)4JF{)R?EEF8lXM)DCBXILCPBCWKU9i&s*pviVI`FMwg
z_oJtU;h8(agP^wiN}U$sIr#5+=$-1@>J1Ps*2m~(J-MKGwg?;gGfFYKh?dZiq7a95
z(D~&z*ohm;?P!-g5|X>mGL4P(xsZC;-wJ4r+ey{Gv#@C*S{Av_vB|Z6Vx)iXqqr{D
zEr|E&B;N0X`XrzU(i&51Gwoo>En55%6$|C^gK-si=O%IGAmk5xCr#A1Hu~DoX;2zY
zLwy&ZQfLgG?_Xa6sgcfGJ+u^}Cr~P6eg&G&p7V`)7sWHuO7y9NSIt4jF=7Q_1~yI_
z45&PE!c}YvHFDt(#J6lPG&~)@Xz0Bl?hwVdig8@P2YA;84qCCZ+Obb3j<#}9Ie`DD
z&I&U5$v12>-Xg!(W&ZAO0&>uX7YA_xS2q(<OCl{Bx@ZF;HLw;`L5`Z1wwOoCZjhei
zS|@b{dwJxleu$22sr$$amYF<s)Imd|aX|&dmy(`5gxLMU^N~lzNDV~mA~hdgV}GeT
z+n;<gh`d{n`CeX(AR+#-tS4C%s&l#*iDQCn3;pG+r)NHX(Mqg5ZX|J9g##<TiQi}T
zADl&5J7FcgHN20XjJ&w4Qu-)(+}OUC_B!>WeW|J0N?4WeRCEXhO$@2#{e=*ALMeVQ
zm%TmmpN0pi0j`9uSoe3+Og%C8&Q;zqnf0M$oijX}Er^FAkKu{VF1ttm2K6jPKSIKT
zbmW*W+=hRk{RK{P(jwen#H%^hhgO*Rl-|9Z{yb9MTf>Y=);poIduVUceiJ#IZ`jFx
zhJR<E)Z(o+`w6_KNL~!*ZVeAbhzr!u{m`{R4>7{oy$?faV+DGg<QO~)Z47O$zYer~
zgxRU$+k?n-W5W_yPyqX1!nOIq5YSl-BN0E$?T?~9Gz(AF$Ja7SZDb0Y%akU^c|0XI
z=fv|+IDTZeDz@`<_R|kD#cEqy+VCPfJ?<RGN6?p$aP|ew*q;aFzQF|n?@IED-99)d
z06$!|f_ygHT=rIaU-P_RmIZfetiP6!FZc^(_!ezwjVUy$HnXeyth3SV9^2y=i^c6v
z<8mYaCQeRv(*zN9%P(a&XsLuq!hD3bwA*rNvnW0p#QOyC-4ambN8RN6@`k&i<#xaq
zY-AMI9rS!^#(Q0T`4G(q--!5TcJ4tp>6c5#_g)7yi`G(X$rUA+a}}p=relrxd`*uI
zYlos&5L$Y%=Z)PHGiMH}4-R1r=ymmpA<TdwQ*Dtg&io+EBSQ>Kr&pY4>Apy=74vSU
zIe;E_$<&(S0uy_NCY-&4L2-PXOD^Pwy?BiJ!w|+5^m#xD-tO4l=a2S_EcQe=PN8-7
z{W{b`&Haa={#!Jn-pG{J;!X@+`j<nGy*K6fyE-mJtD;W5GsNWjYVw^i9qEUW1_Fpp
z#kSvt&2{n;K!>^)7>+qM8v-UghnHS-Y}bn)saryf!#&G|Rv8-D_TYa2oinKt-N(Jc
z-Ot$M8las9U&F}rD1+aGm;&xe8_Z)L!GBSOp-dUts*V`S2syv_)mw)$9&}#4XDDr9
zK2g=7jAXUu;o&+ORCiRGNw#?3B;XeH&`_pgLRT@r3NOk_(b&WlqT@gnJ|~jjI3&Zy
za<C`uS~MJ+hP5xE=ry0?xf;6o;EQfkqzEtfHN+716FjZHZYkQ~hmH3GUYq9}&pUrm
zn}*S|7pWb?m||3^J~)gi%e$#u%p)77;$JLyX%*388jq=JgrEu^y8p;9=Jg}P*qWo`
z`(8!jA+H_8Piuz-0qptoUo#ML0Kl_(dcL5y3t+eL@#*bd^QTr;%0iC#02cqT6^Tb}
zbkXMN*DvZzs<#<5lBF+sJ#tcAWzZB`Pp(4OBm30-S!CanHz04TKWCA*pT7ZFsa9c%
ztlV@1@`*Y#i+r-<24tI>!J0C0+j|sw0RNk;KfI^*VoincLl*mYHT>MGx{MlE4t+Kl
zHSV0g^Xi>OO_8<l{0*e1)QnM6VLkQv4ajM=*QoJWPk&u-{r0b^c_xk9`r6LI>ygds
zEm>sq(HoF8>O&L}*POZm`T1<2y2Y%qs;5mFvq;~r-}LIWqF)I-fq$CvzUOnqcji%3
zQs8>pZQn`1(z^f8?Dy_Q;_`e-OM&ZYcRsAXYSswqZ_OH$u=8Qp_4vi7vbbBg_|y&1
zl^*q?StD7m^xOcgRSStmLTlA9(F~xE8H6fi?S_HZ@BW#(jv~TmAKidlxul*ZniA`k
zOE*9-t5yr$?&Y6uK=!Fk7EKA-r_Qy|V?R~bS~M>D*p+=-ub#`ktyeiKecP#4S~Zfm
ibA9CvF4ebQoo>|><@Bw8eT7x?nVDOf4=^q~KJdSTYsR$z

delta 5904
zcmZWt3w#q*y8li)6K2xnBss}U(zH!Gp$Sdfq|HE6NNFJ*QYobgwo4Ui!H{i1tmRVP
z2xg;J9wHz`M_o{ZTNG4W<e|X&C~FlIP>^L^6}1XnAHc;2y6Rf*>ALscz4rGznaRxm
z%$e_e&;K_Uj@hmrv;A?Q!E)k{NO;~;EUd-?2pWL{c>H?x%dw2GXRDzzJ<(7F>}uV_
zmgb3py&zNsc;Ko#z4Zs=n?TlsAj=|1EGW``Tckiyly(ow2+U#F(V-TBmqUYd@K<_>
z2ERgVIXC&HB7S@m4{~@AV5_9!>U^eV1SIFsWd`p|Lc<Ynj3`QVr<@D39!#5qO4!l7
zQ!WD`o6gy(2VJyMDUcQiu7d`WwzCPYF68&HQ6>_b+XKQ8o(J!}FEd_wZW+k;0z0PR
z3B=x}t1D{JwoU=%=z55TKO)03*r4=4(-f!?q56JcXI0(M{e{_y4s~No)~Gjiu!elB
z1!ecU=d-y;vy%EN`1K-0c3<fYa}!%+ei&1rWZICF`j+RY+}u%XcZ<b=G~FvKnb+nS
zvCOv>K${&*e*$~|>Qr6-GS?0MW=>Hefm^kUi{P>Y+8zhl4p?L0zlBD25Bm&6aG?p8
zns65mJc{H>IH3o~=;FVe)_uWJmT{WbMLZq;*7wQ5ZnrQ#vq+S0fXh^6dNiNi`dOSc
zp!@L@-fr(s2omykrvz`Jz;am&tX7@}wFpl#)7@Udd5wx(Nd-|9^0G}X2GIAk*xUHf
z3b5!+_@ds#=qjV0D4N>wF_xR)Au7ZO)Na3Mw#e~<xZi{K8;ud!Q=rsmgU7_n4UjB7
zdIx&;e7GvEmrG$l4|DY(omyVYbk~d78WD%5lNY+@i8eH%d$m|Blv>6XC~-O75ie*|
zJk1>wTjF?Z_fb(ozV0u?GW<vmqbObdWCsDixUt)3ZDHaTyrg@Bbqu%J!ch^#Y29yG
z?;&!oe9fTKm`b1`_Ou@1_U?H$2kPuzZ!1PyyARp2(zc1+@7S_+v$KQ$u*ze>^;FI5
zO_}dnUf+X+?Y4h+`Sa6$art!nk@UpFRI7czxV#=+P5_&5#RvVXX;4`js)UYsJU%G6
zzuE)?U9h$bF2{k54}{=U2wW`#KkKT0*0sF>xX!^e*UdeK+A0RU0nB0G6n}*5);i3V
zhp(d6DLbFix2=W&JDjou`_cyn1`If}QA<=#r%)a5s^Zus)ms=j4Rx7n&UCjpN}0Q~
z;?nMW9S+@UtvnUJIsmTM$Pe9z9Fr1ddhl=az>Bv-+m^vKe;J*2XzkwB!?J$Tq#LlI
zhiz!!^?!}~Uq^k5-d)R!SG4wYH`5#qhI87UKuyyTyLS7}nx-(cE&9Ke=z7TR8ihAn
zS-!V-Kt%sLBvEFNzR8Z7L<m1S&|z8Jafh0_O3g*|mqvf-s+!BpHecA5COoj<-xa^-
zIa3}phfl~QKo*&Vr_k}E50k&?@pTrV9KX+@Z&1FkD8qS5&lM;h#~f2^#@tenZy&UK
zsty!FEIlqynz1=L(c8kmgWF7KM(-$4#8-{bt}i<K4>~0yBX$t4F>utaC3kq8QSr76
zk0jfK*YPNki^pZ$#j^=#&8{@*p*5$^z85OiDN1yT<D}k|axI29GP4M3@gWfp7^RFr
z4hUIfjmb9}1UuM5H9l%rB{LgcJL%LgX`NIcPBn6J#{&F&E%<kQhz{dtP0)_-=RH|W
zM7ha<i?HjA{cVd-WZ$cG+@vF{fXbvj>f%Dio>;`uU5+F-n_GJzy8HJK#p5l^NKa3u
zyo%d>H$;5@iI|8#WI~(pcbNS%O+sh}D7Y^JBF<U7J>4|sBm6gm#6%*1@3!iq9wkcB
zbkIQ$G^SpR0TguAq4$rj0RMx}L*)3}n`&{$;L2(}ahDeePH@#;NQIcFr=}Um8f}vg
z*#8Kbsy9ODO65t{8Ss<UX1a?|rMoX9HO#dbPZODrkr?Z!kfeZ%T^n-$B}^@Y@5l8{
zkcmUb#0xUVfgHTviYs_rXb#ky@mj4Tq3>VP{~N}y1oA;%IA~1#bVhbUQJU*#w^Vl~
zG5Lv#jaET4wwG=r?mr(eD$P@hTHIGjrk4GdiO0!1JiApYBgeVVR>H*o?!cXT_F2yo
zyi*WI5r_s>6Zs-=(aJ_?1*5d18;~Af<iNvgppu`DR`pz9+2B?lm9x&L-=;gRazaei
z7c)5FO;)*fg-h!9_AhjQgy(*Wq=o1}Xm;N*U2~&X76p!s5YasJ7oAd2*;v+y%BGf+
z?+LkLhQ^**qzB`mL43_d_lEsE7x6T@TB>s(Y@kB)JVy_XuoiVXmYEzek=oakQTT>{
zx`cx{UC1VEP2HSK+(uotj;DvYR=~02$qAsKt4Ndlfr(Zo0e{GmQ?!73qk26%%<+cZ
zy990Y&P#>mrGWIpR&opHdYJlzAuC&<^;nK7a=!~P`1XIggJ0Zg?K_K_Q2{<8GKwdV
zOzT~Qy=Jl8occ0!5^+W)BkBT}l@06&hgr?{G9%ni^XLnd&AS8pxoPKu_onsCa5WXK
zGq}Ix`hQN2$Jd&exE~j)^Xc5g)=bs~QE`t6%|f;0Qil8)aI)eb2m>3Ia@T9qq$B<=
z&w-Om@FQaVk=}aw5F!V#-cc2R&==IVak_;HzdQq9wx<xfNbc2zW-}f|yvWr!0X$+;
zPLrJ0`YoMPXhsp&9GGW6pC!J-3FT~pl{|)}(1^-KR|bt~h?8$rw~vt$k{-XK*RR8^
z_TJX6QN3O!>cxGz13@?ugnkV~o}Piurhbs#5POWSI(%Fk%LN(t=h-5a3e6b{@F%*p
z6*P2QKUM`H_;5O2E`-dIHdPK&5YH4uNiBhP?+f-qC)y^;S}HC%uebk!gGTJpq8#uw
z^3T#3d{d}8X%+Wb90M4S)5pGI{H4<F-s9j8oYt|D;4|I9k7D=YG3qxyW*WMrMoJhF
zeXO>YFcLG0kWO`N36rn03)C}?nQ@U0f2bZTVMgSPbBHG`CNnhxTh@Onr#6*ga)Mu)
z!9<myR~;x}9J&hPy$_8;OsJ7?YC$REN$fV0ZCQlPqYfIuKXVYtJmMB`=zIG!Ie4^@
zTJWu0Xb-pPHK8+9*WX!*S!*1w&;TH7vfZONq(!svpX_#vJ7@sx)MD9f`UAJXUzzN}
zXH86Xr}JeCK8z(_u#;mVvVAj0qBdx^k08NWAO$??TcwO2%~8|*O!I`M0xCr%x#C=J
zf!k|=n0ThcU$rb9TO3e1kORLAUk;=k_nhqdUd}{=`64(xv@f&}cbM@80Xx)`pK)+s
z=5dsy;7av1KQl&J?(XrjN5$@fRDhlE;M9hC2$rOiJ}%XV>X;50_o<T1WFxCuB{TU6
zE}NV%lD!-`J{)h#mJHZloSaEB_X{?NOiyRgg8Qz}*P_qKNh!1_yfSsfk>zS|q<y)R
zU3Y1sQIQLjNb&?k33l}Cgj&juDSbQpggg#<FUseOWd~4hrB`u@o#`l;=C@Dv_VdPO
z%A{M)|IYqb5>)$T#-?ACx9?LZ5c}LTHN35AMlz1XqxQf|7Nyu^EHmUpoHznIK9*t|
zl56oi(_l%%btdwo)n^3h1VkKpF7!FN#(kCx%|XP`!P9Wq71`)y$9~q?R5IM9Iv;ly
zy7KuTO`xtIuO}X7>)!<rz9oD5bU>frtxmk3!`6IKXS6psh4B(fIKGpJjceKtl$m_P
zJMYgYcjbrg5??jrYqg9Lt-s1|hYp-;z^~@RNb$E0E<REr?>PD0<fP5CuJ;1|AwQJB
z-)H()Kx3?nom#a~i{IDpc#rxk+0WAi9g+Sb?E|GMMr1af(oOu44u6hmli|VNm|}_7
z#Dg9@o!~uQa!7~2HPG6=OFQ)jh}uuvu*X1(IXm$;jAZ)1OMcDxD<{PPZxigV7dPYq
zdIK%NjRt5(yYYTY=vadf`c2_*>+bL&h}c)_>{GP+ct<1d@}j~;HMm=ATA-tihgn=s
ziPSv?fA-FsA%R{Imi7nmoZ)z?iFk4u#ZMOGL!_sHeIeAsc;fo`=nmTFB%c!=f%edz
z%m7<1ZRuYkbx;$|C*p?!YXl`YA`eZ+d$H7Ke-j_wLd(GNYao1t5)CC9doTF`ngG#I
z6qgvt87#J&?N5lb%%%>IPARctuoPru;{#^ihqeggQa6W2K_&S~sM10#4gZjZUM!nd
zt%uNXU6d}@V+F;%SJ*L(QHE1jSV4P^Ts$EI+An>LC0ygnL@&q_p!Wh(TU%{~<OW|e
zp!;18s{faB>#s+5gzvQParQNbhxOlwLsplNCazK+En_T+>$G?m-3hYtqj7+{<!N!#
zDCF~hEnTl`uJyK{_d%(7AF4Y56+ms^^I_HbkeuLXanss|9ze<9Gf@A@`O--*p;$^<
zhd#C&e3Qa}jt|P*SjEBgg&^&iI{B6!uMMvwOiAtZ(SVA=XPj<ha55KqLo9mH1{I;%
z_(^^LMe%?r_72Ct74RFp^C=FRaekTowG6S&!bN01KCaFPFxiPiwwn%-$1_sj`Raij
zv*APl=WtbXA-N*lw4;Nv5UGYWqhj*5ab=Tfg6sn65w3Z9N8lWfoMrd4=^=S1*=e53
zLt7QpOg1bohS*B-L>5A9pYT|ClUP>{k*d1#<3F=MRedsyTo^?T=cKO6%MirH<Aptm
zykN-TS}INnuub&NXYQVb_(=<~@OUzb(c0@@b0v1C$#?YTqc;3E3;EA--pk*BJh-$}
zdONU8+xllpc&aCQlQRd)Um)92lYvmc$dEGLmkS{W6ySTT?2X}D^=qk7z6pM3>Fc6t
zyV-inde5ZP_TcSRj?g?dCl(BE#?6k7T(^7$^(;f*L)?ur$mxtw3;sLxEXw($rC2NC
zjU4MmYfOAn=empDzDC_&&P+;#9njuY(VxhiOipI&53=X*V|qGUoNmiJgZuL2WpMtw
zPz6Gqzq<3dbBB)Rzl==J1}La4M)#34y?eQhp#=M?zv&ESXNreMk-3JN6|gu5`kut&
zvjah(EgWV<yn*YBpaFC<o~esH#3+^F8LX8lNKA2i@~=&cWuZ{)&0bY(<!MVu#u}rQ
z%}p)%)=c`k^Bf;Wmmuz7%1zjp1!V2$9KUBBc{Ve1bU*-poNNL4lR<{rU+Q_*z0)My
zT**4$ctXD5KP$w)(FPwh2J0(RFLZt4m~3*@_4q`$xNi$CGVlj*Vy25Gifdi+lY<L1
zl|ndfdXw_n=d6@5ijM^FYl3(<9~AjM7x}TMW;rz7NYlKHT*Gw*+~-aBRp&r9M6$s<
zKDLYPJnAAEEA8^n*`ZOi7GTjT<~tqhIb9=dZNwerJsH?ofu2EV<>x)m_g<emdsJOH
zhS8&g>fK|Q5o2cA!n+;L04yM5^h~=)Y&Lg&PSWyu7t`oR_c>*1O>wc2JxlY>(b1sT
z`>o`nHRQqd>K9`eXJe@wltA3xJK&3?hnKm-9H-Et`&Jd|q2@kT;roq7)ESt9O5BdY
zL%&S)WcDZRKb5b!)B7O5dD273WM>ykPs%0;-D*;#Lh#tJdSj5WP7G!{q8TI|BMtZw
zn~d&TgsoNb3P8uY7VA&icj^5`JfD}Iu<z4}XL9w+<cWYwwY-Dvmb0OMvRmeb#*i`C
zp}rJkgi(*$<Z_^lgfD001-L%H_MCJhZPG@DHizyI4hjc?`yln6^gwK1ZI8Ic1`F7G
z@mZCtU<z&f1w?tjy+QBm1T?~=v^5x8ZB^+d>ZA(BP5IRP3dV*msB0@2$zsjI<1%Pe
zc9fdPAQhM_;A!eh6-@Ee`NR43_|~i>4Q^a4+E0wc=S8x>F6nm`L1rRTi^gGbZ0k~r
zrupnW<@DjBPq>WXJiNwR6GhlZ@SN(Zm1w^YcCG@vHOtY!JAPEhkEQRvO`SHD8IDG(
zE5|a0S<xafi|m+*e=_4cMiMQi0llgWAsC6@Ty=OX^W5RFZ24R4-~WijkJSO4MjZXh
z>})}oDL|%;kIioFSU9t^RHo^9H5R|wgT%LNnIOJ0du#w8Gn;<M10f9n+?#)g)H=OJ
zvd}-j#xAP~y=J)O^7<jjyXyWy<lP5{Ag9zzgUG4Jh9KM2d`yvTPY*%9RhtKqZ}$&D
zURO6^O`-Vu%M@9S|DWLNd(~4|Q!K3S9rDzsk5#il<4mLHa)ZW|w)s<ahC!2;v-#5@
zQ?GoA{-rb(WoXRmE`!Esp*Mf6im%jD294YD)qmyu8ablsj2f5a$iduSBd@68LFARU
zh9FO=w^2lR>fIs8rgQ4^MvXhN>D&-#-S4FTv9GSPH5OkO`=^QhznFOAV`<{y1=PfJ
z+OGvUp%$AoZsEk&Ly*%O)Rajhs4Gkwqi}kI^Vd(_xn&S{33qN80<GV!zGtF-vwr&!
zXs>D{8VT)HWuh5DPtyri!`>Ywzg~Sn?WTw@@b(a7{U!A{(d1j!Um60vs<LLf+^ath
zK|WBc%$j`kff_f{*ZxDj+pKZY)2_j*m(??aS1+sI4_=*CC5uK9Pj4?B;#Lnlu3m4^
X<fT3E_;brGnsX*Dkqs~<)ZqVbhmh)5

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1982_08.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1982_08.parquet
index 21677838771165edba7d8a30ef5a54fa1c2149f1..f80ab2818c17c777cdf8ba172a4b242e4a557030 100644
GIT binary patch
delta 6319
zcmZu#3s@6Z+Wt<MftipTk|8sM0RjOM62ip{7&UEBGJ@bGp<r#5;tG+txE2+QTeq!1
z6&3B)>R8kDR<Wg3TwO1%w$EA>FVsta>)L8t|JIiNtM)>vi{1TuGyW&qXD{;eoJq*@
z=A4<F?|Z-Z{m!NR8QuFc#ERldCl7r0TPps2Uk=}IBp;%A2H@geu?Mx#A2Qyu!o9C~
zT>$cww`s`+Is^bdwxZMs5(i-|%+`Y4nI_x(_I%kRKVo?X{FOX_Huufq4JX%`x9YNG
z*`FPJ5l{wG?Jsj?vEQ)K(r_3b5A=PV8-c0C#)%CzGWp|qNYMViytH_D-;0_+K+2!%
zazqvFK&XN8HWlYGvd53K?h7&Dc*u)~l8sDRZVk=A1ozBn={5*4%5s%0v*bvy6wEvO
z?uSx)RbDipB&J%P^MFu1RU=ndHQ-hjFH!!YijQsq^T^uk%2%r7=zp+iS=Tv56E!jg
zFJYK~v$1}zCeSDaa2uluIAs|NEXcwO%8V#SY_m8@qw#rU50|Q}kD8QF`4(k4)H%w2
z2x94K!LI?oo1$=o@^h4<#q%k9h4T1uTr3rEdM86&HrUgpDIjTqZl!#+4YSQvw4N%<
zb#;JL0IrXtq$CJx!Lq+EOUeK)#2WnpXP_})%T8nsWRDx2F_uf<E4E36cNW=A7GV!C
zzS<fz+~AXKKD%f3*OcT1B!Qn*UZs*-yn3jc3}0mfeMvLHQNy*}2UWr@HORP2t@F!r
zhR-_{q!L7KTi!}PLQR_G8V;6<zD$}mPhC6?*h#Kd6>IRi!RT%B+PqEn@j|rZ05q}r
zCiq<#Bs<W@nm%XRmVw;fr}I?%a%-m5V(CE`AV0HIt6xt)`NO>6!+<`=pT)I}Kd_v2
zxk0J}R}#?c#=pVZchIJy`=cZucxS6B3!tR1*n=JdA2(A&0#9dC6*H$Ozo)ISR1}>~
zO>mTaV%n=5rZZ-T)Pi46FK|;RSOgnHVCzZ1g{5ky$>F*h+%Xn6B~Tup3q^MZOs@$V
zzulkAZDAOatipTvYCaVRh?^3~D|cC6`hGNiR!zx1til?b$7iS6_3?lfgKVYL+&N`p
zbW%$rb9yz{@Ffi$i@#$jyUgWiIHT6pWHf6$D<81A6Tlq@(qDm_sWtC6*kpTSpmEFd
zG1#FNYt(cziB)-F-mkUe7a@}U+*ECG5I>(tIcnHg$SSVV<&&(+M;o!oQ#Khzb3f^I
zgj2qJ{F#;;Y3Lr~01z#;lx)L~a2i>5dJw}nJ$`YR{_41ePq303)h$w+_1=6VVIaML
zDxM|&kFM`HnCX|i+d!DBSYzD7&WH@58m>vy$#S{fZ@y9cQ+tz366gms6B`^g<%tO4
zZ7kknfc#wsZWXUQ7xM`9K^)$$+=!W(y)~}w+3Y`<x;B$to3bVad$8v#I!7EZ^bDX0
zn#MrDIe!ox0zdbHvY6>bbxH}F9%o3S{B~|*0yrOCmpzhK7P49NQ6r}pm6zF3=vCzk
zJJ!*kX}+Ne*eYz?%tQ#_el@<OrdhW+O{#{c?*nvEC(o&to7qNjk}^p>89k+JQx`EC
zjZ7e56Q?O(sNMPzNu)lZ%k>{1r+I#HynCHe5W5PU-+v@lLFgr=N_!vsN2f+Ew8?n0
z64IJl>T^&N<<Bh=()Q-C49pU*#o4F1KhV(`>Dsu48&VWSn@P;*0O!_`{r3Az3!IRN
zSLxV?jREHZ!pdCXtU+g!ZKZa+Fc<9jiUFKr3YH#*_Bh5@R+ychO8+L^AICl;*b>ke
z;-yqZ^2pf@g?M2hv`E{c9a3AQi2ie2A>4nHYfq58sE6XJ6D_?DRwm}z&)=mxo5n^X
z*Z70^^i1zOBN&R?@#86^_7cjKNsb!fu}nUXUdxots==G{@w`#xdvyeq?THqAHL3hF
zZ7Ro_8F`N1?TnIMgBJP;aN_Z9@Z%8%eYC41zV389+b=x;?pKXca-_bv?I66wlX|hj
z(8kf<Y5WUnEasH@G5A||+l<<Ygi$_;W;<pxQkBfvQb<j?n$gyce>O^U!90=sn?VwO
z(Fpn|Q!%r~Qb+H^g9h$QCf<`P=ibl$Z}ulCIMoc*;!M~0Um+Av2%i{js!C-!>0RIk
zrQ#{+GkLnLt<Sh1Kxi2#z@LrO1q$qq5+0L>cPq_tjuvjD5I0g|tFrmHVPo*E0@w7Y
z#3uKSD7@8e=`p{-1Z+O-3~OOR`me1gd%a&H?e9vS)XAif1%yM%l-z86nMY4{*xJ|E
zTBfleX;4vn0;dyjnv}jQkz0_~{wyBN)bFd;n+itYG-FXIbQbEWeGLvNam7^+D5*M(
znw31AAhKzV36PyNieAGlG}ftTsSy%~NeO^oQl{$?tdg2*VHrYlF<(t5Fl9AP?gtjn
zP(uNy^+I81D*iKz`<0D4YgW3`+A9{BV6hIaJ|w)OF4<}B0*`$zVbaAW;mKmWmu*;s
z$0=Xva*|KlS+nGZ4O&2JNC=1|Iv3~VvNPFeT%iop=e7)#6f<qxaNGzyQp4qCT~xcj
zq0t$QPnN5sGtl=PCC|Z4>E&mpF5_-7;=3#pk-V?mCEH47lX~}wn<}v5$IY%4lvESB
z3;W;R)r@6{cdy#ooKox9)&=Mmj%S6FoY>FS_o8v^)6$;+t;QGFwt=29Fwbdp1MV1&
z_vD-HSk9&F9zK@uW#7zwz%>Wx%|-FbExkikSgK@mriag^n%|&&Hr$b0>&b45%hLe&
zjwkGb-Nb|xp>6otd~i0b?;I?T7mlTw$5Nh(QBxDu&Opi3B4y;s1<6hi`CYP4D*h8A
zJEZ{w1edEIRUEBs<&xuOWpP;?*QXa|Fe!R*i*lCBK%Xeza>)~SyK5b;huP9p5awfC
z8_y7F9DWU?@)bmWwp<7MQ^iOC8>^FUfaU$3l<~<x1VTuP#ZRy}!yuL!@bm;I=Q1;K
zKzS&B82VCK5kEG*^{|TXzyIH@RqOsw<Ntu%D`&%^gNbUXGQzVd4*=-t=6~6!tMTBM
zMn<@EK?}}#=tYpVX!!MWYXNJNw|L_tEsP<OzyYcA$DTN*G2kpmAWuS|>aoL|4FQm{
zfcO!2518E-@{v>l{85_qCPzs*zZ4sy%~~ye&8tw-M^HacBj?s+Z235fp6nf|IECTp
zF=dvJg2pQCLb}ZTHdZPE+NL3fOPl*s*EfhJuA&1v5pGfkx6*VaOb~Cx7R70#CjnB~
z$B62Rw>2HhbrGNBZ(E5O<x8Qwh2&0UISls0N)N2efDK2$eu9{p9|j+Ul(G%KgTc8G
z1jNoD%xS0_L*gh!g6Wg&0h>I>9;lz8HIahv?0lL|1E=#{JtO@W==rWO>@=!M#$$Cv
zmH1JT6S`udip>TYS#dq%Y5-|i#8uU<{2R1?shKp(A#Irq!A9t~8B7C71g9Vo<GY8C
zM88eATeHZ%)17FQWem4ZA6@|$R=~Mtz_n`1?-7=3!M}E=#<`3Bhxb{U7xY=d=Ye+w
zGeY`qCb?x$gq2%WbVluu!5ykfCW@$Wa4dwy$Z(6$ex1|B(x;#-`|;RtHw<(G+ih1M
zciZu3&~i)ExJ1i2^^FZYUZi7WzwC+uX%ISnAn%x@@emUeF&PdJi*<dMDI!5fV(~)^
z(ab{;lj+JKx-#mQ7`wHw@5>m==AJ^H9XJQ{x#=U+;p>{b>l)%WGVl5g^PddWjgFc-
zhU4d73`hDxE&UOx#E-@!cApsq<Itq-)*vP~U1)+d4e0<%tG&53;`2JX%e=h%BOv9T
zWHIf-bK)eSB~l~Wy{4Nu;w8rdKMfcho5T>8kLmc}RH&Vs{SJdK3&=@4QpALq^7O&c
zkwCQnHPUe8o;K%7X+Upmn??7mkD}?xT_fotVxQT)TyGNgX$^KV&o0m23lN95YUq62
zk<2}zlU@e54$(EuJ-f_4ij=U`31FW+9i$17r$eWxQZK&3k#bCOp{P}bP5h3z=nI6g
zQI9n4y~5(pL0SO(XCwh4&Q(i4bmdbO?*XIvk8MQ|aDQYp{h(Du(JiDVqQe-+5xdy;
z0I5~N3tZ2|qE2PDIGlYUl^Fic4aJw0?c!5NQmje&crb-LLhaUF9r$%NeW4Oa@*}IV
zKgnbJO2GYs>mRZDm!<I}k+)O=y_1`l!hQ~kXeAK}TiZr_AA@PdnCwF<m08Kv`sZz8
zN(R`tvnd*zU$(p7+y7p28LFFNPe&5>LJD~#VJ%O1AdE@%k{+Qr@?*gwPUP@e12>iv
zr<w79Ags<e*aMQiae?qKXQJc0jX<BqQMm4Ug0uv%%p3fk3es+1Zx-I44F0@|bV_M8
z87(`D9^>myAv!O+o45Rkd${l`hz$2Di2HT~K9)wDUstAb$duFav72S}dZI-#S3sUh
z#^nW2BSe|8i4|@er_6+Z(kz=6xJ8-VIiX`QG~(MPs<F{EAptxVEmh^@vaG_B30#Q_
zuM=sOlI+}?1X`2AoAxBQd=S0~uA3kjG#C=H&su@UrOE6is6wx|Ac`B41QMT!qw{qh
zV%GxN9|T)|V!qBtjJe#^22R&O0o+K&d1<)cfc+y#R~L}jaoFMJj9K1spo!QFPM<H`
z0vL`a<SyP<3NQ+dLfiN&=o`y1-5pD!U3qaF-%6-=>GBA=8_gtpzq=ch&r(fEm(1Ek
zEQ+o~;0_9<2~bPAt~>`=&bZ%I64R1fN}RzmkQhYxf@h>AlRbAz{)&Qn0V+^539r+x
zS8F7@zaTgX8s-v8bbU+toL&7Ixk`45*M~!3dMeAxU^dg>la8wFO~P+Ik;cvqPH&p7
zYimz2&30a-S)N0zv^kOrdH!fduJ&WA-Ce7^o;G&c{<xs`U$u$e=lu?Z^CQdWnYWkj
zLfjlPYX%Q~u>@<zkcPqLuflt63@Jh=O614%(VaizpGNcS{3uHDp{%YYor{<%SrW#X
zN~1EADziD}(R~;dnG2QQnH?<)Y(2LSqMY7zZ<<)j)c%dH%-cw%7ncR6MY=cr6nd67
z8_@!KSDz)`%x7$}hd%)J6ivzW(w{S>XMmmn<)XT*GzZMz;Zd<rQ(McBSXv!ssWd`y
z{#BxJy@0ClR&(>`^pc{VQ;5D25hZkHwK??5rsE!s++0&m>rJ;)k>35On!VPslh{~U
z=Kk!&6SKf0T#94UrY}}1(jOXqphWl!D>dDzZ7mD?k>%~n2T|oZz{m0^zpvxZt`HvB
zT`2A@<ch4#tpZ{|%Kp0+{{}}yYFd<XExp|E%pn!un)&YrVA&Ssvv@(7nsH{_;SYcw
zRy(WGf=HWlw+XmJ=&I`ZJ<JIo0x~Wgg5UuniRD&<Ga!*>r|()$KqB76+2{<AgMsj1
z%(g*gekPw=M?o|`&uB>7;TY(O`Lc^|<XLGc+0V{vgli<nlGTy3n+#bNCM&OHj?ovZ
zi;C4!Gl0<XYogfPg9>fQLyMI{i>oDkj%hl_5JKg%_1{poD*U(_QFsP(VKQTxvG6<8
zeu5PD@H^=4*zgY8a1#w=0i9)86wZjM&lp$*^r~<%kl`*S^=GA`BCXVbqAFLyKLHtT
z^1Ax}8QRrdQuxIdRrtkU#?vpVELL(vBE+kSgg97s;SW(^b7TydqaHOwY~PU|-dq?5
z;Tw@La09eAHXH}Q!rK~rM8loHVVP}`+3t0a966aFy^;gb?o-;6JFXm$p*=+A+*Kg`
z0U*m9e^(Om0Gs|#Q3MHl#9#XG|BV!&aIuJsMJ_>EZ%yMj>D#B06N$B7$I&@Vd2S7U
zpT|d(v(|!^50d83ON%5H@zS#Y!e~v9LqQQN2O<jB_>+nlDod{S7SKX)iJlzTkmzsG
zLW!>^hdcKaXd)up@OML*qlTmsSC?|<h=&SRfF<5LEs7TUW(0}Ki*P{%`h_XG!7G5&
z36@|}z$E39#YRL9=B6Bn@O+p(A8;O1do|)6Y_Bwl;WaRO4TMR;Rl~MwIJ^@<c9)+V
zE7UF%9#r;bO<-=P6IwOmbcM2|sZ=S7)iyujmigoz?<M+T={U+(x^OL#RGg>mwmqu9
zl?^`ek7-0wB9hd2SlpjuuS4jxvO1dxrLrSCSoEO;KD;!Gu?L)$n^YwJvwwoiM@@Z;
zNL8t>%5BdXUt!;{A0URClMFuZwDjpl*zgdX%Yxx2EK8K}j)~S&Zr7ev%lgU>sj$j@
zOy!<JQcPHQAcoH%dHw~mNZ~i{_rAI0kYo8F2O^h>Tv6nr$VHPYnp`ml4>_1PCHnoh
zR6^Jn`paR);9BNS<~4~tBpTV;S@n&N{c6}SC$AFvjY9Wfm9XBE2*NV|{iORyOeQW7
zcohkabyFzm2qT&f98wA0ndBRro<aA1rFT!L>Ijo==-o2}S{Ld#!ekiMof?8{3LPZ~
zzv<&4$i9Kl_hdO?-@p*){B0B(c9b#m<ez(=;)~}(!J|w%|HX5vd!bXEq0Iy`oa!6`
zZ4JG5lsv}PeM6A&u~2LeV;9252-I)<k4-s$JT$(C$>Gl*A2RLzTcK4wOoocQ@9AMI
zs*jh4KImc6_>Y%I-TPR_R)jRi7zclB#SrMVeW8iR7z_W}KF7VYemosoc8tm7e>|Og
zFVuZKbR+_GUmpSuejob&7-QuJzaIi!I2$TB&R7i>&gS2{U~A}6g7B@^hTV(wt_p2G
z&R7h+tDN^j>q32z<<@N-f;<;WIzg6uZtoCedrxS}3C3*L-ZKPR8(KvWv^KQs1e2wD
z<CV~*6HKPzjaSAFy)QmVwrpMB5M=49P}@ny5PIY!qcbdBRWfvD>z$b|4?(WIes|`%
z$joc655bp)22V1XXlW>+m&rogLZf?`bn;e5KB?aMI<%^nF-P8a_cB)1hu?%Q^)hL~
Vhu@H8{xhM3%l!v6Z!>4O{{#C~h9&?2

delta 6398
zcmZWt3wRS%+Wt<ODKlwunoKg&CQZ{$+cdY{GA)q?Q`-zJEtdo+7o`>qp+)#>kwUw_
z=u(PSZM^}*lHH<6E0v3)a#>t~A889+L|1qHg;f_9$}fmoD<H0TA^s=evlp7@oSDq?
zoipdmIp6ob?{{|Y%{;X?(=@GYhMNO!^C1OyHju*&CzFSGjs|#Ux2X@c0t!9NB>3F`
z((Hapv;!)Iz@$>3l0_#(HLyejj+}JK9(3eOKIv}jV-TFd0cdjfmHp<}Gr}gVQ<8$t
z`sV><(lx=#zno&eV&cWIFi8*8`0kBBhnVCki)$tF<2XptJXKXzRxs414u{31p6s~X
z@`q@DWl`Q5Q2Lw7?6l+yPHKUu9`f{1zLBoXt)*y~?42Jk{tUcA?oimruiT~(UK+X$
zDjd~$@qkk38mYm(q3kZTR8zeeKgi%!Ooa}pe<&xrnAoL!Ww}{A2Ew?|IeA23j=#*H
zM|<CrwQ=LjZI|e<yRmMmI@~CR@k6va?3N^~V?g3%xh~F?aKEWYtq%AkA8V2~#Z5JO
zJJpU|)<fOJOd=#uHT)zs1ieZnUMT+@=W1<R4U)eUeBUxxmA)WOPf&*44lMdX)Bx3>
z41^ZX7Ve;Q%F0|%Gl+%YIUXmbK)nX6&kyOvOkh_r$-%HY+!(eyfiaLZZ*-5b_Cuh|
zAQs(t$t&a~%zio$szqZB0m&Y4_?G-tNk%{-1m*3@)K=?Nit?wxf_dO~0<~W~*;UK7
zje%-@g(^#Ogv0KjBxMGud7=@KfvaXx)09&ed7eYotiDevM#x<?5tyl-nF_|>_d>G2
z$#3^JIVPQ=L>Hhb&3hr}!5}(->Tl|y+bTe6@7DTi0=czw%~+fTBSG?|8r??fq3;Xo
z?*w!XZ%PbZ-9M4>ctKnMo)n-qO!{xev4gUdXye2H%zjEyQV8WmWj-_q0&JF=xB*qG
zoVKv)%@PsBI(eCB)?Lv+`L?F#6;wuSJ3p~3lMhmvt#eeoN=GgCDp7q2beVvuBksUY
zP^p_-o?1cLF#+c$DSZJCiti3vmh$SLBWPxuXqxykFXQu>%5c~;Cy89y-{>knh{pky
zQVQUHjoR)DI4EYL9?&X~vR`4?r!^1^+zuKObFPa+Q5(E(1Jt@hT~l3~ffvwkJp*>U
zQB5V_<BZZFvBL@MhicD#w4nBNPPcoLz?%r-r@&@uguMp4<Y){xZvM>}{4-;6tEgtO
zqg6$DcFm**5Q!Pqs41(*(~L@2Et3GYvTAKUald@D37<45?GlP-59+ie5(D`Y@fHpH
zjee&Kz+??6B|A2=YDseY5KX&%LDMDe`H3wLGNKnPkW_-s|3wmsTWYzoY?0{&?a%-S
z)C>L|;Pd5-G2Vi96HNjXCmXerROJW?mqI_bH+jC|sOjn{i(R!<Mucz=gYP#${w@RC
zqnDo?bGI^*h{N*mn1#+2iEU3iXIgrnBa1nEN-B0^-xpMl=@moY2%4;JBu=;N3OWQq
zwpw0E5294LjLArRFHISAur*2GzHfu`nn7-2Y-3g@;~!WE$eWpBv_Srlnc&)%C0tU6
z?bGaRmJz~uScR{vD8_p_O{{@6;{l!2N)0tqGt+3I<%z19$R|ImDxsGq)8Vk)WRlOR
zyt?m9F;x4w?ytx#EGyG{XUe$=PoTAX-$;-V3d>V9<CzAxS}L+j_)dA9#)7Q!YZ`NF
zsLPYCr_#yxPz%^%Em_)Nz#_UK3wLUnImu!7auTMw{0W2BF4-#_cx)~>@Fxavn@A+g
zRkmX~P+8;@(y0H+2qbnq$lH?8XQq>BwCIzZnMJsO1a!|wPO&Wpvi~f9Cs^<X`$&@L
zZ$*7^>|CRDaMpBVp5xCq<DW@q;*n=kOFp&Gzcd*PW$pM^sRW!!1XhZ!T7G#Jmq)Fq
zD;L$`yYsQLxN5hS7$t|%ia$-MI;%-zX@Qm+g5H8SaUW0*f*UKn5X2sXF5dIFe!+1)
zGb~OA?~BP|xj6<nUI*|!96^uA3~f5<syev5)+(gdmEs-Vw)vqcB&Gr)#dIvD#cGMQ
zrGkV%&1~C&e@qsaf-r@>Bqt_vlP??Cx3cgHxw*G9-#JgGVv7JZrcBSI`3UJrxEtdw
zGtwAVd=J>&V%hABFh{ku4J9uRlQ;|);;wPpaG|47#D#fym%KF5)%us)_;+tp+pFyX
z_KQ34p+e8RIAfD{M;xB(we|^rpu_fn=AP`rq>M$`#|HghBF)P3hqMxbu`qwotdyFw
zU*OO~9rpH(A?sWQL^UdTB#BMr@i*>_b|c%A-u^T$q3ib4=`2?~I4!xP0=kQ|HG#!0
z(fIgzAIL^6Ze77M6k&^?C_Z}wm{%yLtvLUalTL3;0_TNs)H=2~d4Ylw8^Ks0CIRj-
z>3-ezar+r6r=mSxEROYoSvzQpTSY*hCJ}DRQ&CBDWv!dN#^CuXC}jUw21VUz_yB{y
z%+}UOrL{#mClpdnb*)>JU(|9o!JU1~RAPZuS~x$4e_K`llF$o2$5IkkC!2Ur8Q#q-
zUWY5>bJ`s9Q3oT4Ug**QT1R$?kg9uCZZ5NsiN`_Nsmr~0dwCh%)`QP^@HjP_Ydfj(
zenp`Z>VPCwizi^{s#0pe%QC7?-1R7Xoi@G4&@ram;VtheUqaCDS+BCtfqx}<9#@LB
zu~C@6`JbV!5BqnkvR9;rTs^&ju46sJAC=$Kna!1V;8*hn2lnSG9X>9B8)V+do$hG>
z>bVkD{!Zt*d!)R4n^9H1yX?TqbP(2rE(qP$7o+pRGao=iyG(yd6aJtK*zx1JA)m7?
zF;5Nb6TYYiUhx2wpdS2GKDZZe?7mW^=ljxy2}<9z;=7D0cewnnlJ)W`HU*W)U935=
z-o|FL>{%Uu51pzrJtrSyGtp`JbJjd%mpA0{+{sj=0ly66d-OD+&9N6ita_Yq)8@~>
z@#`|g|7>Hl;w7+t(3d*N41_TFl!TZ7$uO8E8t{M~s@Tjdyg;6=FF+UNHTnts@xu!4
zS>L~_u&vLd?J_Dg?7gI4gUahR#MF?t14I$1tlqyFRiY5{mDz^?bX3c~bl7<V1jTVN
z;CNpH?q%qCkThuQ#dp^O*2ph&$#<`&4Ka@oi`_r=Y3atWyGjAlR0SydoY2q}2GIsY
z!Fc<S;C(+IiPL~PLNWf-I5FoJG#b%>MnhfjZ%~RKLEZgoDYrIr^CxlCOn<SQ&W}xN
zR%wX~1f(9bV)k?p=iQiQ0YuxNr45#p%PaT{fA4c*2~c)5k#U;bAA3g-<<~(6bR)c6
zRlkLzX24|A<%ANgT6_o~jafnH+60=8$FwoY6>RIojQkN_)k=g4aG@iPgBfv*@CLYN
z%z`V|pwkDPnNa^Wh$e9S83;~anFXnp;yp1%b!BPH4klp_G&uH_5;qoo<Ps1aVY}4e
z2-h`eEQBN6-A_~LW8VW!`9)JAEm+ndMQ?~%s;ri9y%xfAQx&~A&|1wnK|<O1QJPuh
zZ-M$d5E-&F5%ucTQy00!%`-s^#Xf&Ws;G|*-jW_S`wgJ@w$l1J5H-PO6L3#hOW99!
z(FbAkgD`S0Q1{{|RLY>Q<5imI66)=irzrQX*)HICovpqHI9GjUtmlm)CnTqMTiTT~
z=r}^*(XSMw^n9gg|BUq}P{*Ot*^>}G3L{5>IqHxhw`U;!t03-7#CzGs4IF+!zBf6w
z^}n=3|EDB$^?3qrrU{W=h-p(#4iG51Nj_c;?D>3*^|`+J95eDRQ18y0rHbk_c{&Z)
z@aN1;86v!DsA+W7-q0c0>`fhlKT8k}@YZC4<oD?jvqwPniD>Hf?6sIk#uf{ts{x?3
zqQ;+FYkE^VcGb?0fRF=OEh>OFCW?FvE!zE-)Wnz$odDcips^=~COHGn*Vo?#p{34u
zXiOy`H_=WpabmesSAR!r)7w8~&u~<G**-Bnw%CiPzKwAdHM4gdRpOrs%x?B_3J$0Z
z4$`gFAuZVrkcbbfseHW8%&t+3TfwVERCC`hA)q8!b|48HONbXwj?E1nS628jttZe;
z_OoQQ0#|Y%^)5vh5ypNUQhWc#uQ~_fa^TLAPav^EX{b4WKcnD$kSx61Rs!K{vzEGS
zQ4kprsukqx#<Bb)BRY(85{b$jnohur?`Qi?mb@)jn#M8%X(VCj?h-ziA2t0N)ym40
ze4J__a|vzP)q#J{Wc*1kO9`TB^12kCole54B+ow*bX&zqpyM{r0O}>SE0sA1M$}2T
z#NM_EkEd~g{Ew6XlI3!9jqZruG%*wG?6;|Edr)$CqkDIoD^cqTM}|vef166C#INTd
zY${DdMSHs9!D=dJ@lgXy@}6n|f5!8B@(qr#=xAKdPiHMrj24NTNDhF1(kwWeB<2IY
zoihY|(@4gG$J_ABX5!iSnu-0`tPZ&SO=8}^6wXa=Td9HbTT3$T;}#r8RG#w%&iVsB
zz(!w0q<CLMtbQCmm=3}6S@Lf!Iju0BVRSmeSrVH=<|^TaLa60`O2<hIe=T|TLLggk
z6A7JBj%~_f|HyZ&f<}DJqHJvROh^KsRimtSv+3D<a}s;62R~(^7^Uc7JCi7NDreb|
z<Ox9Z8}NJsJC@_`dFRpW1s~*ri0?F0U*pu)e65ejy|Ctc{*L_Yi|N_`5%E%Q8@N5M
z@!&-gwx#3q2HbQTNfpB)GZ6>9tj^}22<n({VfFdqb%3#Ga_*`<6#&Jk81+E4AR{IE
zz}m09A#^mICns{NNlc!)ejBwDEhI|%m7U-nkHCZE<7vq$?+F?s)|)&=U|-`ak|3l4
zDbEGg6W-(Uk7;H+(Oq8&B8^Bi;k(l0HabNvP3Qb74}_h`j937S1cE?8Ks=&wH1V5!
zL<Jo8R@^4`Zs{wdnC_Dl!?B1F?~6HnUND}PYJxb;;higQOrJ3K<-~gbFL>lN0ky|}
z75Hz)RoyRauh@myI)M?uhX=}WK`BWf?7?ch(@y(>4isl9E*+w}f5K^{9J4G=DF%?O
z_u=jb=xRyiODz>YWynQ>5N~>0Z|TaD?-pFi4fek42r1qEjPdEF3OaO|n~}FknNcS{
zFL15H`Ltc~ar?oMsxF^b@okp)Bv6x~%9K}Gkpse2T$}*4p%6{BCp$|JD`KI{Z6mxk
z2&fwWPG~+yJzVl_syHL&E+l9f4Yk#B!Kap*YpW=o<ysojdC#br3mq>JWh+VSJ~uA6
zfsa3x$fVEvRlZ!9G5d~>%K5)AV$%%-Yi$`u)<1syI-0Qo@WDJ~Fwk+(6Tu_ficH&z
z*xRz3SM!Jl(fO?gZ*|1zX0A;>o>9eLI!I(d*1yYuYtP-30rL3F6Rq4|IYvyoF+yf2
zJVQVPNpHr@DZ+$`z7A-J9s*KM9fJD(gypI-4DPT<Ql6o6Ibo4V6?daMOpXzvBV&3-
z#?){5CzZthAX8mU%RD8-<9BE^$x&QOdg8h!Bd(DQT1<enn>}|EmCY)G+<{C<;-|?^
zXO-%FsuG_{YzE*v?j=mkPM6PQ<)M`_wt5zO-=*8*OC)92-8Ym4hn-B_MWwwOFI6EW
zt|4;szyh$&ZutUHGYHT}`_awP)sMt_L_Ut3Gg%*q8sqAW(Oy{63(*>QzlPvYr>r0;
z*odsClA}ZL{!m2s0?o&*X$JXCYf>wtB1GY0n3nJ3qT^#_WPIFx;~`<_!1w!F5+Qmi
zRz@y?=2F)H0OtQkjeoCZ4`6RG(^UZ^llPEJeuFu7fI)I77oxm}HOF@ReP9gbBYftq
z2Jr}hP0-(@Kf;6m<UfKA`6k0J6nG_uCecC@E;O+kd0}=sw_ewN7dfn0Ka@!2&{es$
zc#y-p<=<r&&U`DS?6m21-XIb6Bn6`#hMo?Cb$?9uHU6j&mmq7d|2o-)`UiC6Scfdk
zW^(uuh@R2lGa7Y_N*8}?sC3m5Ut(vb-Z%(S*8fS)v<W4%ECh&euH@_ih%SRA%K+!m
zq4R1IiS`#4nxgAq$-3CWlG0NHhhIYA><yBWgV3Y=WO<EkGJQFNgia$)mQUK!71zx2
zceZ6oUkk@>C?!)Mfl5@`D_YhQdc;|BtNlLR*G>qSdeR9s#8{<~&fI(5z5t<v@;y!?
z+9$7c){p;R5d!!T8#ww;i6b1cavvYD3_U<lqNX~xJ!jH1N7wHF#;^@$2>9n_%u9x@
zIq<Fx#vW}?lP#_(*@wNJ*V3#TXMCuPUMuduRy_L&vN$dK$8ed%T`wZ57=872|EmwL
zbv?G$g-BJ98b_*<REpGiQpdc$)<q}E(~cif@SQ;;@bv8@PcWLvNhoj%vZmSVgJw=`
z<iMEBgcj8`u9#m?;N}$ka58`8u!0}-kv{8#w|4q4^7UcbWa%Pr9VDj{{D)7XTbpd&
z9kIPhXBal`9%UMi)V@h)8itRLvRsX9AQtZG=~0%>uOsh}*YTZSk1}mvt%xX(&;m#P
z+&YUR>muGGbOv{1-6+$Rp2#v{GHmG?Wx5vGc7#mh+MZFCQv;DRM`#CsYJixAlmBBS
zw;zpI`sf^P`_WO|UXDiQ_tBXOa(|?cwkq~L9(k#cPUrSL9(U`EyE-Cg`)C)})iKKS
z<labnKW*io-0Qlv*ZU`OBSJr|k2LnvTAn=J+F;GaUm9$SH6Ty7HaPKpZshU+ZLk&G
z-2K556o2-8L^hXLZf(5to%|b(Bjy2GXCP0vHux}7Hb7@{A6_W9)w21?$fW_AvliYQ
zhAbBUu7QlQ`APS!O^3JKXc~EYfYus@w~Vr1jC>s%<l^p8meu_c=TTZPtnMFW`XVx)
zSkM=dwxhI7v36Tz=TSP#uy)&oQKLuB9i^>?;h|BM(~*>8#B%zH^3i=}5(_uHb(H1s
z{>anE$Uukpk20N(3=#`E9r^MYZ9_jq^n-K;xp@Z3{I(89<`2?B?7n7@&Q`oO64^OO
Xr}M9kkk|aD|M5S3`TE&G`ULy`U(uUt

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1986_09.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1986_09.parquet
index 3e30ed0f9136b13ea14f9c7b5d5740d5c49e984d..a52e33ce96a546397c4656158574d4441ef5877c 100644
GIT binary patch
delta 3327
zcmZ8keRxyl-Tpo8dD3%|lPAe(&Pj6eQBK>0q$O?7X&WFeq3I#9EznfT>ZBBip<%E#
z3WI{Et#wSCN;fzIw3nN!6Rwq6w(2&gVigB^sT%_ZQ$Xs=>zMK}RJOeelkPp;Utas;
z_s4Uc=i|Qa`}c95Yp2}TPPxUl#A2BT{@4Y?zczmsf6$^Y3_K0+@wn(J8o&=&JJoPk
zndr8MHYw{tF@yi*#taP<*~{HFzvCB(xdXWFL1-Tx<?b?W+!%kqz_z6EX)5&tNG!ll
zAf}@P*4_j4kKy$^!MCAH(NT(dE2_T|hW*KIP=$|K-xU_p+&&s)a-9{Y^aXj0As?Ao
zWk6MfYpx8k<ZhPSYsc=u>n5nR*%A|IM&bylgXPE%LF8i_n!4)nGK^8PZzjt;>QeNp
z7C1agC|wsy*AjdHE%2QZxHcOttLe;`*8uWzC;02AR9KiYz9p>rCza|RUC5@oyJq9u
z;NPH~xjgt0xqE;G75^<cOlPyTLN5!gj8`~ij|c|k4AQB<4^d6w?j>WjRNJa8(E_N&
zyID0aXVF7cR^b|qR4|N(s9@UsPNvEp%4d~)@cc_uR(|3{<_f6E*U>yscU~<zPzQr9
zwzILRPG}aY)<R;CYFwXRPUk(#L79W5%n&id$k9U&Xa2Ovpud=_XCu|%_aq)o3{mND
z?>O73uJ4joGE0QLR^dxN`eCuZx?)Fh$nP>#-2&0Amfgn<xk}XiW#bkqR|AR&$(LY3
zV710D!GX;rk)vAW=9yG;voZ|KtokL~K?d5%kd@W}NBdpmoCy@XFmc#_F8-b&xxf3g
z_!HV(<Z>dTH6810+h8k2Hl`LC8TW%n=|fM08W<3j6Ttr-wdFTCln9_vajQc)XEi%v
z&gg9zNv$LEeeW|2qC(~Ak-NyV4m4wQ9kh<D0Qu_I0Ni-94gfoK0y)}A&T1zaf7zdW
zeHKvW!>><~-?^kV2r&kv1Ct69JIP}_rM8ldn#UONd|q<;Gf0p8L^$jQwd#m3w7SXw
z$_l`Xd~>wBjP(^ydl}3sYABBg$kB`5;m0BI7?o-yXM9|jIT3*%`IVb%wv$uZQKMtw
zvu<ctHiLf?G>eln;&UN1rsyQ<^2fv;3DQ-2aKu=s)WNVBxJyRpAsO1YLjv>Ym%>u|
z*9WN9%&5J4Oj!opW=%E;)wz;XFZ7#`axb)mAydzdTN~emw{4*GLSq9k{m2mAz@Xxb
zve-n9{8`&<nyH*5kv$YMY}=gtC4N%L07#j>($`JmV(EfR<rUIw%iJgQItTaRMS3za
zbbPaR+*CWTlA~j-ZJAqA4@mwG>wDwf2tT>whp;GGV8h;5-c|acaZLe9dZAq~+o2V`
z*da|*Ozci_jvXw>{9TyK#b-$&VgH|tlvMyT5<Abo0*dbe3)G&Uq|{oQg;G@RU8W*I
zcf16VJ&4(b)|TGS`>W#_ga%3@-{HIK@hrJ>{H&h2_uqa`1{=8-c^uJ8&)=z38C^q>
z9>5XCd%+56|Jp;lsj^n#G7Sn>X}#`sW^2_}T`E#fKqbE^rPHc9c9N{KS@n{a)$?;l
zeu^<DHV3m5(&tw}NG!D52%;kkspub&GT=u!U6dM|7bQMRtetQ|<ZZM(4x!l_=kZJ=
z4iN&%L5N;}=PcmAR8<>NmQKrv&kTBnQEv1smCeRKDM+joz<mEi^Z=mpi?0bT2ee)M
zkTFPGp=MD4u@Fi;i7EMM@<|C5OXHB&v~&TK?lN&rjS?y60=rDg4e<N>u6VfbGaWI0
zts5XaSl3-B)Z2xx7c;Ds;&hmKpTn!y6tW2wVXwc)_uvePCiiQH%?r_MnvHnSC9R4$
z2gF%qftglAYIHeE+RbPgq)WeLCU2J7IOO>5QzJJkKH{Km^nH*&yME!gu(KEyljagy
zRaH_<L)L4i6*&@mnM#Mao)ST#nK1Xbt7>(al+nyLxZW_?K<iYG5&sD?;iPq%$!{5`
z6aQU7sWIfH2D<9dU+CJHip`*qzY3Jvz-@P-JEQMdBu`5{k^Xh)SqMdz8P30_ce?RQ
zQXlr?BXl+|lL|X#>0nzhP*UtGoTXDM9ClGl^YNyl2j_rvZ^fg=WNCa?a3Trh|IFeI
zM%p`i2h}6`f6O3qOoU~ON!Y@MA$`>cOW!MkzQnE4ep*ih5;2k=mM|7BX+#$3Rkk*$
zro~A&G76L0I!kCEDpPVS@h{F{CpU{*?U4T?y|x(VX9HU8IyP8Nh0^&NNc=_U;?Ast
zwX*@c#hj6w#|OS?R9>W#Z^x&IjmNaOtw!8o&kPPe!A!1sC3PD>N)(D)Xvl?uZClk<
znY>BMF*Wdxm;BX>C@QF8lkf~Pc<V}?eA%B!y~>r^*#~zPMvelWQ~Cx^eq_mfF041w
zYPwaldE>OK9RJURoflupFXlnfp=?9_iU6F6<5p=atpWcEPgE`<TLrt`co%Q%itfjx
z#VXz+DT_f*q0D(o?GYo|&#gW&S{@xj^&ccAq=S%-ZY$s}xDLHRZV1X{?L+G<uCp?y
z0tiR)(N2!eXXU_EO|pr_j~qt)RQRh#;JeSa+PbI#+`%rW?6OIF?LBQlwI^hw#4DV1
zkFt6iv8_ST;hF(`n^jwbD!0tS=CfHjLZSL|Jfp;?zv15aSw<-ZoJ>5|ApkCEJPzFf
zD$x`78C(vs!sz>*)g>rda5=ECdjss3mOy&%7JIcCQ|0Ob;U)tLDx$h-N!%oujmUcb
zEj@?D@6+Os^kGD)A>VAHr%tU_gE4g(Zl}^JImnU29IFQP{&A3_)rJOgf@6caCqTv1
z>TH5)<n%Ok1ywne)_a<f*$pb1>CtOTPoa?N_mBy@Kc`R1eU`Y*iX|x)IfV3<0UY{=
ze3i`x)mZl>rE<E*@OLU@uqME)X4``4kTh43ruAUz#cRdQ-|BlqyC}%U-r?<Fi(z_w
zuU7785Jkp7#oBYtP|ttQhb@UIMFCzowDq5xw0=iPOybQ&ihG*5)2*$Qz}#dE(XdrX
zP2Y>ZnCvLb-9A0Z@v6xZjAP_yHl=;~6JO(xa`$Ox9L?0YkvpKxbe1*59Tw#UfNHYK
zNi}wnr&xWi`rfoch`cym>kTFsN6BGSPM$4R8bJ3OC(N+;T}VP>t}x_P_Gu40D;*^^
zmyy4lxy%^!fj$yWZjp|IDt=FU+xd7=#Ok|;DIL+)xMcgJAWjO(10V%RKtPD<Y$qqQ
z1Fi_a-K~2MRF)b8{3zWN?&6j@#hiBCwcNknHO0XJ4!}ohTx^-%sseNTcoahX{?riY
zjY$W&PP<QRM6~v*sF41E`F8H0z3P#GYbdo{`?pwy9?`1Yv+1vUX;qGqZ)m-4kMn;$
z`i$A@uy&@pE0woS63Ga)Nb~(c8*w|)AGMd<vV|?TcD6NDc6C_GaZB8--Ei+lo3tGs
z54EkNZ>u)$@!2PiApTSFs|E4Fo!WRA|NIx8g$3VSd=gDO1XDH8uRDT%wD>vDGyJM9
zkQpB%|6^2IbVb3yrL9y`oj7ib2RS?v*Gr`Za4EHLO=QLNeqFU_p)qO!u#9~`r$efb
z;c^S$s4niy`zXp^mpG=4m4s{;?ZY*Yn+dpO>QlV;)7+{(gLhB@{~O)Tos16%xx#||
zwPqy=cy&co<7TAn!xr7s<!*)U{=SD%5>0J4_ifjDX4K9q^wPmph{Ik?TT`K|oZ_A_
z(YpVPUUnxsXtl@21x&)NJlR(~bXt37##~gPVJSy_A^EDcyQEfBr%gz6P@nd>bk^{^
z2UPBT?Y+`~^O>3WlS4QAVlW|uN3QPeux7K$x3i>16UyeHmhW_wZR;pQdKBnUs0XD7
zexakxaDA-p#swr?n(qRE{#pdU<<b8TfG+@mz_s}vo^|oyax$6K-T4hGYHDP@Ksaa-
yesvfLKdaKuT$}&jhlck&b==@GetZN8!)|@Y@C&aWH{4!SFtqCh_i4ksw*LppQcr*Y

delta 3404
zcmZ8jdvp}l8UJoFvzgh)-QCH~WOw$F%r3j^5H`t7k_A??u$k=2Lv|yvK?niXH4BMh
z)gY;W5@V~8g0Xd~AlhnLsU`<#sZ|s)6_uV+tU?8OSkWld2W_fUrD%!0c>0H@bI$!^
z&bfE)H{b8^{mvhApFidnn-dEZ9{6X^B7U@KD!<Q4UJN_~@b+%eRno@^4l!ktd(>5+
zS|Ipr;}Qla{5W^b%+?XaE&{IOc4!%Xg6lG_xjBBk$bN0(&#2UoA(5gVMr><2bghN@
z0lJ6B^fl;{*b@|cFh(wjz({gAXu=0(--l+%a6=3z_;a&(k-Xpm7Qb&YtC|+>{IIXT
z1m9@Jy$;$PdeH>6c6;J^RFQZIwEl7wgfQ}%XU%S}qnFS$n&b1E*-b8$d@5DuQRPgX
zoT<h1E;P?~Sm5T_8AZ!vpYj?&S?Yvf9hHg*7mWw)ou5*vj^V4!sgCw(biV&1XkpLw
zzfW!jjs8bp@<Aq-s})w7A<cS)!;Yxkpq@gc74$vS>_|suXBE|a!@5`z)Y1=`wStmE
z_fR>NyTV9?BlI8@&R7m*WrtkIsfF;^GgMA__|Mr7K})`jW`g#>$0fV!px<F`Yn)vt
zGznE*km#oxdkTw~f(O*;Xkw8Cq81oBxc>*)H|HD3jrkgLv>Jk*#HPd`m5Hp{ZEn+g
z+NI^}wZh{z;Q}8!T^g*Ovbj_ax(roUK&;RD&})W#CF=N3<2ou|1F8tgy)Z9yy^h&%
zbTv-oskAa@GBsz8x((Q=_1AKHSZKi;Z44<mw$_DDm_VhkPP`g?Cw{_^9O*bIKB9NC
z&hH&$_WuI3&^l553-GU@=0Z=I8U-{g_LZq8Y!)X>AHIf0QhKP#cbsJrC1-|i!cUf=
zis4%zJ+usz@hbt`^ljuaaFS+m#ZLT|ev%CaUibA{L2ZI(j^R<4)C@9fK%@w%IPm};
z;3+MQZ`LJa)O)`C<Y$l>dR+LuTNp8lFZtvZRR&O(0X^SWtFJcJ_x;Sv(u7EI0bx*L
zXT00CLSi$OYQ)EU+;x^j6vFt|Zm!9J4|9y1Q-oJoMIdlS_afBDv}tR^C*9DZ-Uh+%
z!yNHsMSKRxJ5|y{f_{UzLV`@y7CLGyR_kD!1-MU)(2cW<Z?goJ;r|GW8GW+JBCqCy
zRd=7xw&Rs1kh7~HogxLk=F66+R>1}nQrE-W2xRNI9eTUTSppjO3O@KteVu8t`UZ|}
zq1bKq+mgRI+oNUyq-?MMWs^8xI;&6x6*t+lw+kzs{lB8;lcHHBzFL3NRNL3aF%9YF
z>=mi?Qt*TNRq+l)Kl+oO!u;4J8|{7ZIQDx%Ys?lQ0a^rbKpO3Cl|bDKxWtK%nX_MI
z&k0kxc)29wk=IMq8vrU2KN@=mRNrk@s2zKY(rWEiO4YbG*eNrZ*1ZtDAF&UjF8|GZ
zaB_SR>O=l$FW(WNrz$PG-z1E^BeQl9ZRB?Fbd*Rg*rtYzuEA(G&{5WVmVr#LYyTh>
zNDCh@pmG7*Mc$dK)zSnt9Kjw<&gGPhrtN$UciL@4>WVg|3_U{ApxVpW#gG|W0<!pJ
zsU0IG+C;_vjQZ3d$`drgb}veNoah>LLiA5)X&mHfYo_sRG!9V=>OP2_h2L5s_(@er
zRu^B!k3QAU3q#!Sn^Z0r|7THRx$tlE9iy>bfF^zXoM0}4=8xZI4bl&xW_}2nK~6kM
zQ_3&!`{h(ahL*ji#q+>_pNZ41kg$gf-Dgt2f?%-sT@N?Fw$}4qZh%~WT}QDHaR~EE
zS+kVlmRg|TGkRT4gzUnU3U9E)x28g32)z$juJ%1=QGW~MPm+(G{pO2X)Ck=D2$k-P
z?xFE@jQWmVjh%g#saUfxnykfI&3K*#ErE<bXTf{@b`F)@{n*f@DerPnH~d#nKD&7K
zi14FQREn=GXEaU2r3~b}7DiQ~@(WZ(=9ZNU62nHgAG@j+M{s~)uj7_S@U0AJcsmaM
z1wZR#NcZt`EYykTr%+l0a#MZnb?6LJ+n~`FQ1NMj(q?gYyU_KqSFDm}ZUjqT@81D(
zbctc?1TntdpGduQkUqfV3JUf+XfqRT4u{H0eP33Rjtl!-)Z#*XcFA4SL0Ug$lQHR!
z-y0rH0v@-BU$N5m;YC!p82mYl@E=83!kUC0a|ANuK3IIB1bP!!Nh1tl2aX!?KsjsW
zl15~ees8V~YZ>t^H!=!u={;7tFQQONL*i@B>cC%=rX5gtgXt=z3)29ta2@WSL|m#+
z1Bo-j_qbEHK-V;&-D2LzP2)o=jq3AM@=x(Ev7M(Gab1nL(UI-%f0#Yp`C{rCfRu<!
z=Q5CwfVugG_R8cXMrqJOuXyokFQTZhMw^5u*#4`QljbYIL~0M`cbM;bpg8(E(9`{+
zJbuucy&&`$87-3*S#SL78uiGx19xWpI(`8UDk)_EZBPZ^iaBnfp0$O73q3KV1m7+=
zG7W7D*0je)Xgt>@ZWPo7muZt7qqJ^O)_2%^Vq{Wm5Y@k%7?t)xCN@ySz2n;d2Yg;o
zPwQuFQ)i|t^BRC~pb#V8QOGHwv2d&!<>x_``^?Luif;JdLnim&tB9Y7d^aZyJe{$(
zQ+-LlLyNQr(st=dM|X2r>z3`5IL=Ay)y{7Xb8CD(rTPJ_F>C3t#x1p)3q;yc1F4H=
z)%dr2R>j|8$sRhH_+6_2ba~?u2pXtFcf7*jD#KZ$?|{uEs5x+z(Z-Hm7?G}p%=X(I
z)mnqrr&J5yu%Mz6s=XG^G6_~AvW*=g)E0ljh>tOr5=xVOGmM@(ty&8=XiMl8Dx=|1
z4*!8OYhmIxqa35nnuU*Y<}gVPsPr;u--l}C^vrG#Yl@sveRmnzhd@J<i7jb_S+W-N
z;Flc1JejAnp0e6)v?Mh|_akC3z@g8TadR%LHFW%4t-S0={Pk)8I>YP=b8|Q&OEXmI
zGE8i}_<8A^8+-q#e<Uczo3`Bt_6C~ixk(R{%@RGVfof>U&w+Y=qmQ;GCR7!8VgC=m
zSgYSrR?Y`_ONr{f+_{tKbS1EtSVOE@-(4o+5xd%Q`Rja5Fve}rPnON0o;Gs7(L>Ha
zyyxdesK#5IRAW0HGLy;cJ7@zLKYv;3jV2dIDG@XY?<yU?0Q80N1ki~E#~}%g`C{3t
z_UId(m1UCMW#n&W&#?wF@`obHb<z>g#7+7k=hl+xHs4y-|4Y5gr8wRb#P<aCPLM)4
zEFeU+wcw-ruU%38em99CD3IC-^d@F@q@7#h6j$nBxt0cRbxm-viv#ds)nyJ4s|v+}
z^f1W$NNNy>F_9`)>vxEa$f&<0s`$>3Wg)lEQMEbb8cf~)fV&E9)~nsqn1x<OQyTDM
zz2NpJXFO!`Y%jC5r8+9rgGaGsgj%Hg-taJliv09M2UmrY?~4Brn8qIB>*5jp1-D|w
zlWc9xvn$(Mxk>cgxL5z${ZPx=D%#$-tcdWH58wq1F0yqv2$2;{P;I`A<-PYzO0_`a
zjYR}%0!s}d!byKeJ-<7E`19g-)pP24eRqI=`l9FRqE6+gof++ii5l2ILdIHT->&j7
z@ofU6597Cu>Oz7(7`VAMYIdEt!yXTC^iZ6b!(5>K)YYBQWkA$JV2@3(F)M&|=P4#5
zYd(r2T7F#w`h{P8Qyi>I9MX4|%l0#lZ8eae4D{TIbM*Fe`5U(MFQNqgSrQPB#@7k?
z;-Zl%i<$&_#guxTt4KK>w33vVzY01w^xlh-XyRT=@4b3=MeWoH4--zw9C&HQmLfSS
z#XVtSNTQAX&7Ejvv~D{WG6{`59x5F?svoMDfr2_M<*9#3zA5@mQW{n3qtbMAi+(|R
z+px<+BG7UDgg@kbY%=}N{g-+hVALKN8sFY(%jMM7Ib5j=fte`zgVw-6YXFf|L{>3b
z6j}71)_~#S&gS>dBH^PZ7YNLiQ2>{RI2Z6m1c}h6O&;Ft;=$!)vnx6Zw=S!xQTQTZ
zpH+D7RU|xBMb3QM^#6ypZh7p8!DT#s011z}$&RgiUOr-2SW-0p(>?B!hU51C0`x{=
A2LJ#7

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1995_11.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-1995_11.parquet
index a9951d4bfeff3a1ebeacbaf79c3662904bc08fb3..dff9d8895adf167e09da6ecaff2fdd4e629b5b13 100644
GIT binary patch
delta 1926
zcmZvcdr%Wc9LIOV9ladMk~=O3Iq(rMA>k#Q5DE%oawLF`31X!cv=&iEtfhY_wzUr`
z9r4leHMB0$Q3SzSJ5@m|t&a+h#X*r4XX?<0s7!IxTA$dFqOlw6KTSHb-@EzF=lk2e
z-|o(>J7isR$SM`s3sdw!|LCSlzkg|hz9*cRU_ArC@m488tSZ7ljsadR*r^4uRa4^5
zlZ!bk*X7*|c5ViXmjLvUwzIM^2_;)IKXJ(!z`>|WD>GlWqYL4!A9NmZ9SfFc0a*(Q
zMQECq`5nJ$J&=n*y%wl{+!G8+V@DdflW=N9nMU3SW*^q8y_W?t%$EhakIGA|YFRS2
z)z-Lz)5zVx+wL2N_D-7KEj@}~WwbeFsH_{9%6a(<ZUK1&pa~i7EsB?eWMpt~@cZVg
z!RUTy>BCN7cJTD@)k@PaRj@FdQYVE2{$PlpC?0o~hyR!nkA}jzo%z#7qszHP-F!!=
zS|>Z9?ZcNr^f9eEZe=iluh0XwEh>3F;7-~j;RxCua`kW7<ZF$Vvq#3M*UJ`wR*sPh
zGtpfxC0L&D3?1O_nB{P=dOECcNi`R%i<1DobL_d=hVaO$2u)liQyL^F%SvWd%&WLS
zStDG*08P=Ecp&qQEb7h@udC5THLNf$NE?-PaIqE`!^B;=WO5{gwrn+j!|>U<1<Oqp
zTQ%x)j&$hYfLsme?bJ)<*_4aK=!e;HA+(-srG(W?iD!9q*jSJyg&L%pRk5ZfZydN5
z2h>lc&kXk5v}3DhB2NehfkVt3PkEM0YhsGZV=AiQKolrG3ibQ=M;zHU`*vVahfvnC
z^0y5-@rQ9<3pi^5^cQG%JpYtK9a?FQ2Dp(f5qMC~B=YzRF-n$a^?9J66HaM=7CpJ~
zboad~O0kIZLx7cI;JwBs``-1{^unT_6VP0%#*?L_NVg0+xvb3M;U8!y@*GXprIlIA
zQpE=v#-pV1H#O*-R+>Bm4TXJON{%7AMc0yn)B<I3qn`@NTK$ETqMT32=z70Hx-Y)d
zmsv~YH@p#VZy>v5|J+<P4xzC+#7r?oP!<O|tfdFhHDqALJ}u9W0WusMa?3Ml<7TR4
zR!cnp76X)uoDeV`9rkEq6eY#@D~OS7v-v%tkkc%)Dk+&s6U72PdsV*iv?zwi^C?64
zf+)R_vPbdNaVSPTl+GxM10|a%kAnWxsm%Gc?FJK^UVF)qUgfbyC@F4PaZf@AZIs)B
z-pv4$>=RugFb26+F!Dm*b%M!jYQ${<zbhFD<I!&tjN%tX!e5+%s9~XfQ6IXPPPavW
zX_@8%oYCH<QxwZAt8KLo&iFLmcTqDWs3YABrHYLV>SZCdVbg3>$9U6#hE9{~fx8G8
z6Zsn-HArh4BuJzp3L>J2Xrf@E7l;@lb$f%v-Zn)x{h<>3ml}Y`OtArAh$24_pjQC^
z%vbQ<scdxEm8Je*5MP?g#t4J^pg*a@f}N1MzW*W9pOG5@q#0NHNb}<WvJHRhBilBr
z{MTv1cYUO3M*ykEu?|A&zX~9O_*DlRryo2{NKg2GI&H%{9BiV{b|Rpt7Y{ht39uIj
zDXbaR<2Z#i>g&A$xyN4on!?5lj|0&W+(2r@BiB;=yQn{#iW?O+4Bu8*ok*O2pyT)F
zf!G{YClKc!sKL{7*jS-vIDib}8Xp;cV)L&V#+^B=Sr|U?qCe@!qdwAqJ%H@TX--1+
zKM5e~@S9FHT3EL!!+*9O+~Q=7um_)UvazrW-*K`gvCHf6&-Hc`;CoK83t2AGBys+M
zW?bqbrOkH&$lbLw|1CX0N=4%QO9$$n2d?|(CeA-_3Ttx7`JTclx#T13@SI#Wn!Mig
gxeK`2=PuwgKG%fr<dSxpFqg-M!M1i&9$Uu!1MA-A2mk;8

delta 1960
zcmZvcYfKbZ6vyvfXK;4dz3hxLv+hXMUEFmSd@;gWs1Rlc7O-(wQE4OffoP*v_@LIP
zjiF+EMzNN(R}>L6SS0qLqL9`Xky@c>#E@t-B}I`^p;YUm)ymqt`pMlU_kSlRzjMyL
z=gi67eA2b;q-*31H78yKV()#OSigLn*k&e1gvbC;e+oXBia8rA99IOXSj1hUE9X#(
z2MibiZG=Y+dA=-VuHMC+8(0eHr2{IURrHCmsj%9CoUY`ZUS%Qh09}4za)F9k&Bxa?
zed9eDq09n`5#&hFJ}vn)U%DMAbHP3%&|f<e0rE#ZE9cI`w4ywNvJ1>PCh7w>1hO1A
z1iHAL${Js@aO6>U^$O0Qw17ZEa9J9v2kFD|gShn;w`aP}wS)P1UiPAXpgaR8I?1<9
z3viIEBo5yBv7#w}?nmW6ss6{2%+rs%3hj@n8H-b2FcZvxuZ<Km&F@R`^P3pyL=^m>
zNuBn{wv1cS!Z$|gO-fz{bbktLCyn~pH4y;5Pj|XM(<uu8cU~O<pP@aGU4JR|U}?0H
z`f!wfyW#|>=q-6#GV0~xBa|_((6{_=4#f;Mz6<x&Bsg;Pb0+}0a^%Ib{pJzHCPQo?
zlRrezR^-hpnpbq4at-%J0F+>|^FZOtS=5vw{i;V-^{~jgDDhv#1V1za>o93=CQ07a
zU$e*Y9mBg#i$1a!Z8zwD_Dq2u4k#snE~jQMOMQ_+yyz#Xwn$nejg)umnLPh0TjWTP
zA`i95*<*|ChXS#nCl=^m%4HTcGx3LwGf`$F2Z2XYMpOP(@`jk){FtH=*clDxeh<Yu
z{sBk2rj`SzK9X`3EL&?aNk5GWI6<ou&_~h!IR0-AH5%oa2H-=?c;H9PlgT!K7%j!W
z@e0t;sSNo@3jI!XbIU`jSaVA6MFJPcz=zd`)Y|PO^y1w1acH*7;7`%w<!*~fDJXFI
z`91?h_D7RUi3QGr1gX!!__aj-jsaaX%9E#~UxuB^Cx5|qz*Lijpg=i&=t>Sb>sk(_
zX;;Q2UD|L#zAvq7PcEaf4;IF$2T7OIuN>wP2uUUcrrL*7P7gX}r0<|kWMQQiBhQxk
zEz#M>{Dv4!i?`keG4g~td}S2m3@cn(e6nMrl*Om6%(ga5QlzqgvX~bMqLosk`Sr0V
zMt^S#qiG)Gv{QZ!-7`7VvU-ahYGt=9Q;IWP!?k#yqWLDHS~ul$qE$&?vO3W#0qYR&
zYDQTcoEkW3^A@R8;P*~KhS8`)hSB`I5%5-qAQ@PwE@?+sr_d*Chn>^BfU~M6Oq%AL
z<#Mkx!RcSb1?MqIeilXL&AOK$RWs-+3#lDj=AgYyAQ2eoL}efF<pS$?zH3v3ytP7x
zL^`4&L=+KC6hSnUh#}H9RLJaro&LN}Cw1c8X{-o)a7`L(8`iUYsv7`HG<iONr~?2v
zdhx9^Hb&^JgQ27nb1ET~?f*lvUXkC!NF6Q?lDZdRqynD~l8Rlr&~s|>A3;)kAdFPu
zF&;vyz78Y3IM2hzioItDX*0i3P$S;wVaE%NXTyRz@NEw}26lX@u?|>;$7-xqtO|s=
zmAu6BG&W8uc^QVz^u%K>ogL<wB<iF#vs8I~sMLK|X8{r+p><WYVpBRR;69Bt3018L
zq5Kk#N@wlTrS34Y^Kk;c5DcB4&T<m5LPMVqx?ek#69<7<VGj6QI_nUhpM5)&boS!&
z>8yxR25S;Kd&7AB;A@^mxKKYB#!K;v4A$l@Elmo|eEqIJT4XIEuncEbt}R-ydRgI;
z*KO*|m7dF_N!M|620IqE;r<NPF0=*wp(#4B-AhQvg&Cow4$mh<tot*JJl%@hy`)#i
zyyU`9x6TZW?8euEIlK3Uk;_=0N#@})o{&jyRf%&mSsVGR3tBgEb<nzrn}b#@?$0E{
TQu~zTcyrvOLk+epwt)K&bFuZ$

diff --git a/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-2001_07.parquet b/orderly/data/test_data/extracted_ord_test_data_dont_trust_labelling/extracted_ords/uspto-grants-2001_07.parquet
index 1a9f4dc4dedba9e45420020010220a424803b7fc..d6d4bb71eac5401b2f7724afd0b65e7a8a08f077 100644
GIT binary patch
delta 12074
zcmZWv3wTq-ww@{3B|A-~$s~KX*=>{Vv`Le;Nt@j!Z37KWLPH;r(iU21X`zG`izU(m
z<)Kh2FL@(dh&-jd<RKyt0a4K+h=2$6a6myt5D`&CMLg<3>z(Sk-*>Nk->f8)wKADK
zkG1~w&wRaZcK5p3aYH>7b_Os{&ZU{fYa`N`v!$dK$w&d{-XHfbWry<uKp!fk-Xw2#
znkhrpyq%6jqxz$MI!U(G2`S@`k|HbTF;t{;d<GSAy~6akAt`xoxW#y-M+Bs3MIoh6
zmrIwR>DdZM(QfrliSUe2*7_UMgq|WuN`6uKX5)>9Lk!(*n|fP!E~ViI>h!4(sq_xR
z7L2YGH1LdreSjr_2HM=c2Q?^6U(pg@OV3_NO5R@T91nz%5s+<MkB({JM^%+xm_l8%
zEsF5|0Xr=qaP!OZ7raTpxLo3Cp{0SdN?{XlPK0nY@HtmYX*D&xx9Q=Mq2uzF!y!^{
z+fd|=;T&2?N1%aX3^DSmE)ZCoN;NJhN+|4)*<8U6_!-XgK&9s^bhDi}RqJf*&x&z|
z7gHo<meltwXjrT+SLr!A@^z{c_As|Cspu0K|D_~kN89BRyG0=tQbCgf&;$7_A(N(l
z1m13Kj|{kFKv)8_bu}?S%e!PUfDg;~N6N!ht>QGCo30!vPX++6AE#m#qxdKkuOk0P
zGiZqlLaqgK1ENrVls1l<h31JL<HU}O+!`kkOuc`lZ8vlv!3s(efmZE$yiZ9U$or>|
z00aea)Ei{cp%Fmn2haf;l-ZEK`v=QEH@#qJ-wB0KuL*TnKDfEA4Ac3Dm9ZL~ZC5^e
zGf{AWMc>%&^xvMWbJ-+;X7_n)EckV2*Enhhq_y!q;w-f$IZcNR9_>Fum!ga|Nog8>
zD+Nzh*1P(vEYvvusm@~*quVaNtWNJp_NhjDpN~*!g>rCS0bq^b){h<7hd&}IF7{>u
zwphnE;z=wXFYEFslL8i>55KQ(=4gX9k=<tlWzI5a?DSj}&H?nQLJ~+~<5S3#&bI&`
z8z)c`Sc3}vriT(>Z{h$yhT3g7nO->k;$-0y>cDbV*rx*a4VE-{ZDn&&_cV#c$04y*
zQzE(N)qN?xp-I5$u@axX%Zb)#BzCiDyc|ss?`jM~=4%v??JGr*QW|1|141Dc@-e30
zV}v-cP65hOd5z0B5N*+VR-%1!G%*RaCwpkeQ~WZzz_cZkQc67I@n#z4!(>;ZSjTV%
zza1jrHQlCRwEebC4obWLy(5#9<)dwJ<%!gXF0;6nNi?n~Vb9B@KHpGOB`Y>kd)$2D
zCNrQH*Z_n((K`-yTnRf+yvnffp!gHhKex-_je}p0>Ye419v~!w@}}9uKGFL;z+01V
zzv`ZvK?}k};5aL=fVR+)-UdKrp)+DLyH*?750B0O7WT;~zbNApWr7<RKLw?_g<@Px
zoMc~$YoF+fDa;C5K^gm85{j$_0q0IIAq&Ddmr;f^D%JkNU(i!VXLF+PYwvQjsEI4l
z(7KW0ftVDSBwmS$Gk>3l)QDZB0d`6yEvggP7#GOmp8?re<HAB7rIXZU^WURmVnVC~
zZV)SD3*eLDGqL6U->Y{G0Q^+{r%Fm79?5u-Z!W1DFkx!cf-@kr3%IYRQcI=9);iR!
zl>1t0hjn-97Wsv0Xpx}zL|!|u;j9e_ENWCyU&u?->L%Q*<g?Ka*6TL*fi^K1v|B^=
zP%y-eu}aN>EYu>C1hbJpMOx<bA#XvZ*%xvKiQZ{32c0k$z2!`I`YmM+OE8#C7Ag8y
zDhi8{oX#5-yorD{U|lLL=z-@Rt2SEJ?tLEkM$Tl&(vdZe;hq?vyBXVEO6^@eX>X}>
zZxn2l1OoW34cO7s9=u#^;xdd8{Ut#Gy<o#XrJIA<L31$6Qsyhm#=oT_e}82#7|aUc
z0pc;PSaque1Uy!W&*zMcu$cg2MNOQg{JC_H4`qcz!O#+1LuuU~NUq=IbDbys>}~n>
zN;E;*nF1BKG`{Nz*N>xsu3<aMmI5<&^vB<4>x4n5T1hrSjUT|-{ybnjiKmFm<FxuC
zR{XJmUsF&UskGt>{NxZ=RgnS^6(<!eK2|Nh9hU-Ei+AH<)$1D3T|Fo+T<N~)WRLT}
zZki)%<J0Qj9xKoH`O1u6SteIKkb$IzH>9<`0<^S*djjL{kf6)~!Qx*mD;nL6w;Izr
zm!VuIuw$fSk~#-9gq+DhxY)b08V<s@(B~eqGm=aSb$L|z09>S~dR*KRZ;-uIiL6re
zeFAz{JRfh0`_;?V3FucTb`=62dO2tNc)UADn&q>g^<s1am!xe1cAE`=(AYFU71=sK
zC-Dz~qU?V@Hk``eew317xi~nXR{feow;j3pxc1H9)Pw5`2)RELk0qEil1jG{c=thJ
zC7{~f3uuZ-trG7i<aXS0gUaS=wsL?WI@zR22dMj0RpUtEMPNLNGc+~N$a+qAoF_cO
z6TJ^qJdpcT)+tQ=URZ^iJb<5H;TTERko`FY8Wr0IpvxJz-+@Z*x5wZj)3dnKK_#I+
z#-7BHXA?(m-(!7%qfB|f)1MU{3e1Y|#Pj8*#=+tqoiXl>Y9Sx+GfM_X@uyn{qX-()
zL*zox96}$7-{?$;%Lf9zOCn501v1@wQLis$57hS@s9$rSUN{%QE>3TLY0+h))$oxx
zN1qUTs=nt`{e)BXou|<E=8B{BVDt4Ujd#%i@nwA(l!{d1G<Z{-l~@Yri)RuyLX9|{
zHxx9_LG4JIRhN#+rR=vFw1~a@2K#X}zN?`qykd%^%!l`DpS)MJ#sqL=%z`1{_*OvO
zbnO{N55!g)QQ{fi1Z%|ac{6MmwT3z7zB9Qbxg-=sM;4NWEn}n+02s>hDBk@~Mi>Qv
znvH%GRO8>oZi5a^60aFNnngpXWcOGo_yJmgs2o&<bYnzEQfef%1X$W07aNm`p+(%6
zbiuoM5$fbm8R4grN`a4%<d+?=0%NiJX9>l)jgk19v1nEbDmQ3^EV3&diVl#CHmE_7
zt~v|-uIA=UWCw{SllMTiILT;%--_#u>+HdqBskMGf*MdO-S76TV1SSVdK6%j0`QTz
zA}F*1)n>O8z-unA6sjw-#YU5v{koQ$=f8=i#51uGaZ-DJfGRL}#I2@R;99XRWjCx7
zrDiiUigt4WY!KVbF1TOZYhD2NirJPjI9!}(sfKTfpICT|JKZBGRvCN!SNwP^+O9xS
zF*>yXek=}6O@gb$DXD|#X`2Y#DJ*IG=HSUx9(JUxBm+spv*q7ZBzr^JnF_b#Ic7*x
zSf~YXVb?bFu@uFiOPR8nen5-D^xh>*Tj@y85S>wG_#>K*YmWpluWmtmG-$lmiwOEg
zr#e#jwV>AIOA!tLv_aaG1{zxU;V;h=`uCy_+i5^O6}S{72%w`EjATjtjd)(v<qRPS
z&|@i!p#}>)P{paF1kgC$z_})}!~)2us>;dW2}VFSkP7avLP&;-2*$!6(PWk2p9c!1
zKsPBx_fE8XmV$tz1fvJlRjE?w#T06Kx9+BjY)Q(_IdX4Kay~!bZwKBET|~l+<>z2R
zM#4q3#<Z#fs9sV@SI4SU-p#=O9nk&Ex!VW^04G*~s>?tX?(IMea3_FzC8&7-RT=b?
zMr_cd)a4pg3hD6NKgXwva*$_Y?^@U;GPFMqjY1MjAmlbvdAgXW{Ci{_P{yg(ROrO$
zxsZMUCUg5<c7FyF&glBcd!%536a-#?C657D3V_?VS<wmB8X6kSh8QL6=WLImG2gta
zj=LH5j{z{r;82i3G*(<+<k|_)vx)Ps)vr!bZD2gG+Av2vVKsCd*W!WEaADOzxQ+v-
za;y4t_<<aSXw*v3VcXW3=%o4j>qrp;1KgKsT!aHzXfD!$i}h^HT-86Cb^*uRxNppv
zD{P_^1UcPcPx9i&Dcy!-m_beBmSBg=(k?6pHZza}%HSm~zn#5R#cU{vjbN(Bx}s3W
zJ4(99KLxlw68vQd(N^Gdzxxt4SE5#R51H{Kf|2i%WuTb(nXGU+fp<Ca%@@$V@gRuK
zXBlfRmb#Ckl06VUTDVIwl&D7!s=5>*Tn4~~24|r4d1TC~r%{^?KPjFGk?4rlq!DPd
zRt-1uZ?TkG=&1M;@}5Osj4L1V_9OtTO<Mr96|G!(Gg{Sk!VVhikky3}`k~)*<vyP;
z8@B&2i5`O<VM&<$JP%|cNrvS`kY#vm5qe>SiydS%%->;@@(%U`#;{^WXO-OIYsrG4
zyajB!552z1Fs>iJ%=<O)KMGNbnriVh+$Q*dY#Z+Y35HdQ-+&nfnmR8UB?<+;st0A*
zd>mca4<5s^oX}RYFbgo7y)yvoi7~JxY-|a>EN;n2QtUHI(}X(Y9|`geZQ?&OjKnYe
zlA-HZGup5`76e^?<RLptV5RHTRJ1S)_u=MmGg?7_JE|I64904Ka1d<Y2#UGOocpmz
zDnVB)mke$WosSlV1Mk&7UKQv`)sbRK)t!kz#ZvT!%JRbY*YRDe@L@0VwL77oeal+>
z>F|7)lcsfkvg<SPB6VM!6t}%IR3L81OzL=}JM^{0!Sij>K)w%eh_kc_eIa>rn1`mw
z;9B5oYBgLx&YuUhhA-QV&r)h!k|2x)Fid3;l&U&odS+eoyVt*BC(3b$6_uHcW^NS&
zA729^RPQn!@S?=qCPagT;Z%k@I}T>KK42D)juO85bAB7SLkb0JvZ=6;0;p}MT`aZ7
zXMf&-F3>6^lZ{_X;3pzYiq2HYzt66fYh`$IJ+$y^EeEMclpPf>1cQ$Sg<#0tAg;G_
z>5JkJLvv4OqOw|BeEwwv_es#uC1b@*Pb&e%KR2RfwWyy<AvLbi{UKhrGup`=?O#bt
zDgbXGy=0#v*(f_vcIG>7mKZH~;|C-WZwIeRymov*hZh-u099x?jn32ff;dq~gj#XA
zPziU6p9>jLK?`fIg6`pO<LAWqEMMB*W??lzsuWfgyX#-wJCG2K-jxBL=RXObb1a7O
z__z}9wXywFY>&7&OE>smQ}CP=5Nsk@&4-Udl1p({jgp`p-Kdll6rls1$e4l;jR&>3
zW|+hmK;x3SFP(XXZBn6Xl`@N9TtSt4y;Twn67+jqyq}c|-xsafX_5jn*Da39F6E3#
zAfJ833jH>(%ozX^{oLeg%W&Ysx5VyjzbDPh@8oO?8iwJSBb+(X#%K~1LQ5m~Fja!m
z)w`Tsm-s`ZPcXgeO${{ky!SLIw)dVc-oe$RbNBPOMPiYIXKjyzQK&8kjIwS-{dS>N
zvn0b;k}GyN(hO~rCFZhIKLImqFVz*t6W_Y`KFORpE?zg|W%8iTDDhp#2KB}xsIFvj
z65v0v+<`Jltsm8jk2|%nR9x-Mf`#H4XNKuLyK`p*OyS?;(l%RH)02Q7otuMGqp(7T
z;>;)~3hSfL0a2Huv*L-xySaI4G%kvLCJP4W2ChJ1<h+#;K#q5+Yo5q#azKGN*qr8p
z?oY+$ocOpS)t5?3n;HPgQ0Eu7T~;^s@BN5B(z@K@=A8aU*D(SK(G+FjPiVZt_PU#$
zr21dR1rL3hecnhyTdN3j<KeHOF4qS`;@DgV>@Pl>TdV7P6DAu*$%yN>q|=eciv&9_
zv%_+VW?M9d!=frL54MV>d5);5Hz`@yr)cwWv`1W$=YTe`J1<^7=|wz@!vX_>>~!&V
zUJBHSD%W&aA})6E(e`59v|N&G<5$$(!_Z3cn9E=o>VqRe@b)|oil%8C<(}FoGyy6O
z_)T0Y-xn0Ki^Lw6E;_uyzZ>YDPi35VnauE`nAR^DW{E@lC1*H;)EZW=owd08zNEHP
zF_13t=J%m$ZPy`fredmYZ>4!54faBj0?@*q7Ps~*hhg!jerX-M#^O1Q)Mw6D{2}pM
z@CQ8lU~HGlWHrEnB?IFLy#N=10Djt!nqs#XqFrA8opu!<2+*4W3a`4yM8cNP{mIC~
z1fHPx|L)%0hXf4U#<_Uz+~OF&ejfgpn<NGQJU6$yX~;TY(e3@Vp~X0*F^aHCWN>+@
zOMI@szVE>i?n;D^c&dLt?3%qOrRL1dvGV}?gaNh2vIQc}U#t9}x#!Pv^@C=1GVPr&
zzM5~2`eZPHlW|dv8oW@vn=il<BJZB6`p*q?m8KGqiXqe1O}np)&%5&r5(&Hoz83QO
z6nTpzmEP<mtxiThmF4@)wPn^iOW9vWYCvn^L-ruaRD1a+Ab}2vm^i=^K}`{d3~<K3
zn(S+_kRdB0u?e;_BdL8-k_ywaffTM2pBs=8Va+~xcEH9+%{y=RWsaVJ|CcP>cJOh3
zx;&nu;Ye>55WWP)5b&bd1j^31;FE#{FnaN=3VU2K<6`LE=*vPgao^q#fOeW-29%p0
zc`&+i4(@1wKjQ8pKrQO=fWOKW4>+~lT^^-kRMPYD(0Hi3A63}a*|%6i+1oQBsU$e8
z>1n_uG4!T%;C&oac>%-vX&_7h)X(l)eG|U+o(40A0-+APm`)mYpQq99CiF5tc2KqO
zSnqr>PkcQPP|vtE5tal|-83_z`<sI+o5sjGgfw7uHvB^RN#fDnBW>J~Hp@}!)xs^*
zTz-91JD{wEQ=5{3Gc^MK8?Be2Z=>+NsHQwX&GF_%Ff2nCmb#NhP)z-C$;>IB{XU?k
z^XsS}iJ=otfTl+P+4)q3_cq{jO)Z?~Hk^p|$rh8GE2+0v#)BQkYeG!#Umv>#j{gHt
znS2{X^O01mts~hfgTihwAr4SO`JE*3Wnw5AjiYD28reNow@NzvFqMx!Vx4s$uWsy1
zP<W#^rEn@h<EWk?wWQo5uS)1t!kz&5tF#4xb4G;mfh|`?;yTP^3??{2?AsQo(T!n`
zY(VqX=rNl#D`>`V%#+v!^i~u4F$KNzI>^?|IgWOWMH>@TQjq1`9tGS~IoF^jt90pT
zG(dwIia;5v;!s2qWMPPW9&847-8zl#b#Y&tz9R(_I!cvcEb>lNQf%kFU2WVONys7<
z3PIJrI4a+}9tvbgli(OX3%a^29Q}klN9i32ss<Cz_fMFiY03ahiz*F0s^{+sSAy;Z
z5Sdb3Eh|wPi;@HQt9TMR?(TDCRZ`=?u)9m5xFu0d?}Hu=Y~sWrljRk6Ecj-c=OA8+
z&^li8=LQkbiWKxf#)&B)gsyHjY`uj2{hPJ}RDN6<2;|fKbR+Z8l;ff3WT&Z*$A*r@
z3XO!MlXCw#3_>o4xMy;lzqpL#cGP%elp81cpX(&q1s9499nv;<*=1(no8Lf<X}HPI
zRP&3UDth<CAn&7+zb2oz3n!Ea=Xo;GGopM-YDb!$Vgq9r!nLk^u!E2)yDS>7olqqK
z<1WwZ!Z#B4e)X<n6D|Q;pJ?xWfFA#tddK~qgZPneG=yD6s?^Hq)NJoPNUswngAz3;
zT#lbl<-ZRdDHlpW;Z~f5Rjc8A|Fu*?=>D}-R5OGs;F_aBse^%6e0j=5s27u_itw^{
zWU5ay+K-Dn-N|Hs1S&z}T-2FcaYjV#DY4qu2ymqjl4Q$18xvmpKd5~g|0B8}4ryNo
zw}~IOXGMSpQ8_I|{nhEd%(`j+*NNOeSTb!*WRdd*ARP}zGetrNaTGniDC~>cJrxK~
z_U;R$PL$UYCK`1pkz2;*)1DBu%Y%i?UZsuf9&PWIp{hKsqGh&IbzLG90p@~dfvGc7
zNCq>uiIZmA@MElnfM{2KG-D(_fMVN~;!Crw&U17Zqs;K-`!eumm7FBk76E}0m;iLB
zC01pV3^l7{5}$>Uy;7l-EpP(kVBmLM6xAIT&6PMMakpg_)~{~r1~9d_Zni`m+L0H#
z5RtBt5Jio<dJY`V2R*l7{VmwuB6fAe!Lj1&9sMLr^{7z%jZ|aB*g1u8k61saJ+V4b
za1HFKhNRrBhSq8jjbfFtJ_3^S13s>gAl{$jp&w)+8iP<<x9INF!X|M<XBM;3iZ)uC
zCIL!Bd!xv5qGq@Y6U42ZjfziPXCxQZBwiI~yY7n2lbf+;P?N8BACR(zIHN26uz32(
zX9?kWDIf;t&XSy~`0||i#ax3dtGVaTGB!(8&NIL-#H@MiV6u3A-etH?+&RB6TMUxv
z0mAj42`6gD-_%I!ge<lzjeTcSvX77<!JxU!XTekR60->zm_=AVyDLTH7uXg10~0nx
z_FM*}T)Hf_E-;6O-q&Gq?Si3_eq|&~e!Jj#xa8pKg(IQFJ%e+L_ntCHKb&bS14Wk>
zit&pwCiLV+JY3pxBet7%r1tHRN03q{uUgW(lJfh`%<qdvD-5^XJ$Fj)-YMnolv4M5
zDkEoBMhdg31&#<xY`=l1q@Kgz#lzs(TR_P8?{+X2?D$N4ZBdS-#A7NEe_oVTwnf>~
z3Y@J#2!Maz!GuO2`-T)3aos1ce4!+V5XXuEn-Oh`>^1o_TU~wtQpx_<E&|~Sr#I{%
zfrVsAMh(HbtIEU~i**gLd6N79w{sl+v1;c2IZXjTvTy3L_X?y~!;Y*fJ@B%?_?2!y
zQ>6p$YZUbeOhb+Wi7%i2m0)iaGlPj+cLww(Gx@vsK`IkIMIs?c)yZ`vFOAUYD#c$G
z=fF3`%q7-=Iir|YNzgc4yKW)qszSVq9H_C0kGj^yP-Xo3b!ENH;2f!ca<(>xs^@F9
zYQfXnUvyx()L%PPH(<h(g<{tdH|!LzElDHDQNA=ScK<3#Q5~cSZCh7SG#|$I+%S`}
z<AzwaGzDG~r!DnI4{t@|l|YFfC~@XM!y)nArExLKn+k6bG_H`o$O574zeA4F#h#^B
z`{v!M{PQya-Kr?@tfy66=Xj_huIcWaXkbAl0|;Ym_96EGRJ4vWuLA<QFOFKq!7<{r
zWdmWg_}a1+@QgTkd2d28fB9(mZ}H~x>Vok;ZtW73i2O!G;tZV&8d6s{{bYDIP-Mrn
zlZSk%6npnKr2u2!UVdsG@64Y`ZRF23d?6_tivBG=v0^+qfB0gB2Xdlmr7h;(^h94k
zAWQ(^;?ZY8WvDkN!Y9Q?S9+M<qs79juTw%ruV-r9<sWBivWsu5jE{I9il45Gr{}ge
zV)8GG7mE>3XGwPidXE(honrpe9+KR3Jk8V7YvewQ&yE&}yPmFq_r%|xcEPvBoK^i(
z)9SZxkrKCKn7$R1=@yygW}hWsTHjAnme%zP*S^>m3($#^TDM`bxN%jP#(oNaoeFBv
zJEz%-e6%;0rc7s7iSclkm>}+rS@#Maq9m<(S&ad$R-$Qz!Zd(d#p_}kNdXkk6vp@-
zB`Io~2R^?`)szFybAb8OSS`FXxA~cc&Cft05k&eBkr1JXL=x#sL`sAvA|nz-L{3CO
z1QCfQq9mdsf{Cb)G(V#m5~um?dpUD{r{w=eh~r1j?$dlYwZlbn=J~8tu5)T7mDcS8
znDN4tKiz(@e8~7Tfe?WFp=F=r8_`^(84Zh>n$!6LdV)}7>?b_>qXeoKtwLk<bb|F1
zH=@><Yucu)A5i1@dB_USbzB%t(aymTE=47Tx<dJ87|H6&8e-5a8(E>O43P!bVGqFd
zC?glQ$Z<n8W@Ea}d4lLl6tLjoI@Pppruz;>Q#U4hq1M?N<$0f+Oui~1VWD$JqZGX+
z11$~fSW6AR4K9WnoSka91~U{l@U>CR3)%1Gz-N%>fFzV!Rv#E;TMFRo+DrT`#rGhH
z&nxf{o#HD)$LDhNo5DEVSWLJpZg?fT-mO~ciATreo%e3?nzH=bva~vW1iUi**)D{(
zR`83Yc3(a|CeUBw35v!%>lA@3c^isK=PwhAos|q^jRuK1K=1GT2R(p(k(rD?lefOZ
zE=1^yw8Bi3RS(MCX{x4efcr#>UWn-^EJu+vIo>1LDQ8oXqp2ksC{EpIgI=k{S9Z$5
z_iGwJxOQ9<V_ALb5!fFD&p6mCsl_SA8z98C>elhFmGV69>O*-xCLF*4FM{!wS!k{r
zB_tRp@}=yUA$bcut$3b`T1wJQsT#rqYe3;fdYtDlN!F{K)w=CPolkdvN4N;qQsKBi
z|08Ts!B5#eQq)1C6lLKc5SUo9U~mN9p;}i$4*HwRLCDaRtkvjVP6i=-OoQSI;AQ>{
z|2R#tOqAu;bpcDmKcq0~j*C)fJRm7cUN(W>h@x$CrD(piu#@jH-a?zwaVSQM6>91{
zos3q>Y+IwO7d<~%zr_EDB}YBlO5NK<05;GKL-423#OX%$*HKD(;QSKXAr*YTStkrs
zLo*#?oFiOeJs*)d)hQATR0=h^v(FE4_*A~a)Iptuj}2dp+U5KX^jkc=m;VU8myDv6
z-6yEQxF`xvrS4H`DjK~R*V#?`3Mp!Vt6fU1w#@^;4{Nk4xHch~9!RNd4mGvH<p8h<
z&f-7!>>&f0rhw{%RJ0Sp54yewmikDBO^n7RL#a|~7&&mcFD$PF9h8lt6t4rH>Epp3
z4J~r>rMG`VtLPen5J-r@#+$6fLXD)E?n%vOA!NfGn;4}q8kOD3Wp%9Aj^41qr{KG~
zH45FMvFMD-86V-XC=QfxpgD$HoXa%vO*dDOQ^}ftd}t|S=`6ON%(+4en}@Gz_yAPe
z^TM4LhGo?w@trt6TANPSZZEJkMo=#>4nur+uZ6M0CE?F3OlIGU>Iw)`wP7}u@x$Ta
zx>P2?6t5|1>qAg?<5<^hpjwMYN-YDN*+4hWKo`T7h8G()ks-ek-ki!9Iz+AO2>>^N
z$qm=XjsVD<I+MIUp-32EQ7g_R#9AWxXo{++ffgekQK}f^tc8tp+`|BiSnk6ZWL9FR
z_XIE}1D5^yPA3GD3r9}KfPx~j>oZ$=1W9zHOx=0HI_cg{1Qz$)KKSEk;Sq4^+3mK}
zi|FEXJ^Ql8YJ-O22&03HeM3BPPe(9KZY&6>G?pt5k4a;gj!!X4lj9o&#_!pQCO8MZ
zF$^QpEnsKA!zdgDbI@D;xsH0~1gv5d=s$KU$S>pcn6RhUrBFjQx9*6_n+FL9ZVVCj
z?(tMOHFaqN3GMtYP47jAtUBzWT{{)1nzfyVmH{P)Ao@tkWoXgS3AXIM9Cg@pn)R!6
z0K3;q(L94^8Mm3k+4*>77^O3D9i2HKfcliuyWBy57sl~}QDv-yV^7K9FlunOmc2h1
zs>#sb13|+xnOr9gpEWFTQ3gY0Hp#0FI#8FkFo1UGmBeQa3E^`x{^a$ahP%-unuQu;
zJzA<TF7dp^zNgmsYCSD!b@;hd-Qo1(^MG(4BAJ^-iLxyx0pAZj$21&8#o@#0j14A-
zZ>BTlxu+O77k#V1?Iv`#2tO!5BZ@(Qx<lBs&=C)^bGl9wH;1s)riKxCC}-1ywQfm^
zh1(VmS(!xj&t8cTBxg$9a;gN_EA|XS{&1I-%!nkp!J`_<(2YOQ$+)ME92>j5gkLXI
zf9pw&Mq{6f#Jdc2>qZ`Sjg#TF0RLX|yoN!H%vKLfWtqyL*%}f~jnYjDanA*$A#0ZI
zCp|Ap%kqbII-$gMh7y`13P10>%}L_PVg0p9so${#f|AUp>k53$^@hJYM001*{NW83
z*bl;P8)NM_8johgcmB?Xqj@|2JjFP!_lnU5De(80f!hrDL3l<rS~G@HG#a@?yqE5}
z%zZ=Cdc47bXXK)nvgtTo0?l^O$SlPvYEJ|cC$R*%EqUDNLGB`ex49&hK6F}R%tdzw
zJAa1Oe#ZE{-(8=I%2;xlMe!T~ra9cS@OL&QNwPkj{VB|5Fv&@&!}!jwwcML5Ssbl#
zQ1S9^6e3P^BCv0N)v$3QIUh+2*JUt?suRA3#dl)>6@$K0a7+{%8(yBl<dr`=inroJ
z{Umkd;Isw!D0!X-JvI3HiFR3%pokzS3N+b|@|`F_#x|7jor-C2m8Pi99LfqJ{V<fs
z0Y~R8Nzj4OG<i0;8=-3s$7V8t*n9c>LiD+u&j&tkgO%N30Y$>uVq6=Zoyq8iF3yMh
zNIb=p2~X~%YS0T&1ng|k(<SrejsIb;_51-yasC0Q1M~}M#}L#L-)oKgnck$ij3V;c
zqZZDQ{SVM9v?osphVNuDiH=GaGAYoxNi{)WIR%SfezzwTY)J)N4CkB;x=Z{h8o7$D
zLt$4henIo2&`<H`K-gnvm>f?Pt=PLmPDqy^w{r|PG6y?VH%B<fLETzJE`S)u2-k<h
z>xCZ^96V0hFiLm^l5aU>c$uBicXW(IZxs-xeF6jV40A~xXnvn6MSF4xzLf^e8R*}Q
zR5I~6cU0u6*g1heQe3IXD!kQQiD%Aotpy<KrV98R1-39fVGu;i${VWe)94)Pi3v^;
zSOKa?ga;?&0K*0k|2|p}Rnr04Kl<VC2RIgVe&pW;s&<Fd1O~}HmHcvBC0WTbln^cy
z7^lS2qzi;Q1SXxA^>6r`;z>{%TGkLj8Od4JgfTGw%Buj8Yiy_7C_4PEz~osrd*ON<
zoeDSeN%UEryI|RL=$L;-NSBc9i4DceR|DtQ5^62Z4%@OAmaeivgdQzGpM-~GG3L0}
z2N+%GjK)7ijt|6I$_UmNUJ7%aB>=g@E3+6Yi6zIgm|S=|{BstQ2*q%0HdCS5DS!Zd
zoMcp2BcScmD74*&O2f0WnOOA9aH9;T7^pAYhh@6H;ceNBMV&tqI~!mD8X8kk0D#$Y
z91RO!%O<z3+QUC*GkP?-#42lhwGX;SpdAL`WCvrj-*&JiNeRj5bl%Mfg|X`-2-+?<
ztjF0q@vV(`+#nC^{MfU*m>2YYJ-^pKsz93XTn95jnKKW!YsTsTwV2uoLf$vRzdM*_
zcp^OB$@GVN!`qxphLzmToRSDyHoXiy!^i@DzPSAXgcjT0K9*dp)HoNC7#Fo`D7rAM
z^Hpy4njA(a+dCFTMoWBUxH@dkVe09+`r8-WDGX`}ugD?c_UUkU4pW>}8xjVB#vj;@
zxHdDmxrr*n>&JL5kxZ|Jjm`DEqnb!Rl3G=G(T#twP(NGlfZh<1Pr^NOx=m;XvrY&4
zgGHIfO$472&ai*@Xf9J8b<aJ~!I!%~>Ek{W*5omnaB{dXk13U3810hu*);~u;B~ha
zZpdSZutZ{r#1hdG;fTZ$i6@dkM0aFEo<24w{>8a8dw6Yx7O>L(P5`v=<UT84dKqxj
z@$mbT^jy{bxBIeRoPfmQ_}YlS+idxyPdp>nGC&(kH;iu#O&Oh;DKHW2*;01ZJAK)c
z)5y<=&EJ3a<%MuTtKPtTc|r1b)o+`^(|T3EZKD3JI`&cc#a`91j~-IJwk&*;s5GxF
zi~ReqR)*u+^hwOh;}2<$eH*T9>uvAbhg2svgcrBzE$oR6efvAlmj1tSIl3`?rcH0v
z9NkEC|9{9fgk_WU$#6qBW3t{9vHPoV^<+KI?EdPZu|0e{ylgUQ<M8c=RL_RbPu8cz
zJp0XR`1j`5EZ7u&f3jW^PMo4w!Zl(06ur97*$e;uvX*?wlG;N*`)zKxbBf*=^V{6#
z9@4Ej_*{73R6S$P{A<>>9mVV~izCUj5zRx^%a3pW(>feIRo^#i=MQrBKRYGFc6pee
zO1kpYtzCbq2T#>28B%+wfi>aQsrnRV&DGrxsh$k)B`WsGje8za9X}s_ajKqO(@q}n
z`1yzR_kRzY+Vx57{ofzfotPgU-L5x8koURm`jkF%uHFx;rt3BT`@sJ?9LfIjR`2pA
znuq@Rm#%;RX&wH(U9XBEwTCQ@?fv;Li@a%iWell3Z1L-VEvEcw@$17DU+??nuQvAe
zT9Dd9ZEU&rUyHAa1xsoVS$wr3temblGGDFu^&#EjaLIJN2`&z|PS=|wj?D|NpRP~t
zeea&GPl21lx2Nlq$eUsY(Y|>woHIku$Gmy)0cq!NOaJhp@MA>D9zOK2l70Wkjv4xs
G>i+>%#3v;H

delta 12108
zcmZX43s@6J*Z&M;F&mO08L|smAYlUu36MY*2oNEXL?Rar3JL;(B7((gs;GFWil}(M
z)m5XkqS$&ttG!sYwQ9w;YSp*Cy;!SlwOVVvv{h@{ueZ1KKiK#AKi~KIJadxWoU=RG
znKS47&TqD-W8S@vc~L`Mr8WjIj~39(`prSf%(-H67Q#pYc<Z&Oi^^8&<E}B-%SzpG
z?w$l=lB}tV4nZSxP^MOrZn6I<F+7kISh!q$sfpv0s0ogPOs@lymhFJs4VQXFK$=Dr
z;)1*ET8+B*ML?S7Mfc<&*KlREw;@UBEr6uu77W{FxLNlqL-!<3zok8sQ1?A`=G4d3
zxMY1NMwjyHcqW7W1xtK&w5ewos!<rfrX`+<z55_(+54#TJP;~_AlvXfI;4h|%S+ra
zfx2#87Ucd5wwZzN)>p$`cE<t3YKdziE%lvN3Z1|{3Bu99V}DNkP{p=7A~>6d+@E6f
z;qvk>;M<T$H7p(&lb?femS6+?Ec*qZ(($F**>-Hy*z0oGHBtIK36c_X;(0R~8mUcJ
zy56ijMzz6S=B_yrU6%3RNc=X`C70OD3aO9?8Wn&Z$fpU(H1#8J_h@=$z$pX53ZSW}
zQbqs`?~nofx{QBnSfG58*n+c?<pbsM008#WM2s<t3`LPD(jU#BB`OHHr!dz)6lI5M
zqNurOzW5oAZT*njXa|C^uPaT9{#hhgKv68vggc%MF3JYk9|<u)P=JhDw3EgmVg8Xo
z=nv5AG$^skygfgdKkD4AZ`ln6f1ed~NItl!rUX;Dh?TKQt#wB(IvOivfMwrW@8s-=
z*V?TTUz773Hst-byK5XZ3({gsctm`H7HeIa_x&G*8myAyB>YYSenweq&rzAFas1M@
zBNU_U5?>2XZjJS*M!R<fsWieca6tiJmEh2g9T?0XmJ}AcQvsW=<s0xs7LS*8xs-7}
zv&Vz~scYhBy(X63YXv3t5@=|1T@%g&v^q-Si(?}bBtyrwn}LUo5U2^PR)v1mK?$%m
za)2L0?a`k|&Y$_oRN)Ki_2sPaiVE0XFek~Y%bE&$W=PB)P9{!L7fJ5BwO@&Et7EV!
zQsS|7+0i<+#AY&%m!n4Uo?749rdB|<pA>~iX^6EM5DJNqk1+lgAw+==1sImds~!A6
z)Twc;MX$(FQyglJchMP3`BikDu``uYN?hadHX7!_ct>4|&;j7dDieTFnyZGSE{#0`
z2txooWpvvh0&e478iv{KTIHa~4bXcsNr@~MZHpQfOMU8y7dJ4mIV+0T3v#K)GZdA}
z3Jug=XH0CT3DAqJ0K(nq-3)e25u0c5N5j%qKv%2|0Y%-wnhD4fDtblyfyseeL@OH=
zvNith+s-Ls3CqA!VgtK5E2uxNP61|iL8w=haj`PN2@F?3v36k@)rss1y0iQ^a32ju
zC!|%6q~3{ujC}9j1jk;{9+96m+5$@0=i^XN1@PH-g9&L6MmY>Jq*ke3%l{QEGT0kp
zh2Q$T)ErLYiJI1q5cfwU4AcpYvw^??x=jf0bHR<V0Qm4#Ir>7*ewU5H5xY(eY?Mk`
zP$RJ64v@yL2kBVpz}9R^E2&B6AEHB|CNcwV5=$cU;5>0{<glFcwT=ORpYFY?qy!SI
z44?2#MU?|4Ob=ar7Wlh?^J*fsQd(%KK|M;jXJYm6o-S>>SEzty2|6#vXco1ewjzN=
z4Jztud2v$Bgj;2NI{Myn!^-}liLI}<S^PGCy`LLpk(zvIXrfF~pN`xK(h`pcIrCCY
z9>2Yw=<Q}xy&XoPckIb_uel_{Twk9~JQ{tFh(t^KX*4GsjMDhwG)fK3z2ia5c<PdK
z4|Kl(9d1!@#{yQ5wTZN#1FlDGn8Cbx-!9;(oXuc;hOA)>H*<jYR%A~xwQs}ZeZ}^D
zp|C;X^Wl3|U_+~1c)2*1OEL8Jm(&Yrw-x`8Y^qPMH`S+^OFSj%_}66Q&QaFa*Qfb$
zfAJ7ksJdAMd@hT`<FSVXS&aZ;Q5j{9d_EcELRtP$Fti9&Q5xrmk{kE=Ec*#@SX6lJ
zgHb?Rw*#e1feG7k@OSB2>mXF2BrB-O3*bC&HZZ)4C+7g0dw|4Jfup$z&47<$jqx~F
z%{+w;Sn#I;KB}O$P)UWQc-|0LUYY<96(tqS9yUxo8I_RyT-D#vi_S&4Qse%3$V)vF
z-CdpB#s+j>2MY7oI&aw7<2<k#=ZH*n5<DroqS--9i?}%$KR|*!6Vw;}YF^XeY`EQ!
z)V2zx+kp)uEfd!^sLpSX2f`<P9$9w){sUcektG#pRD=%;9X0?LD9UGuozZ&P%Vo$S
zMGs@pd*ZohW7N-Xwo*X9NwGa2c+hK^JI3RknbI_m83n}97%q-!1U9P`0RPw|KowY9
zK^qA-zJm0Ro*qHv?s$`uVX0UUQyu<hhIR*X@=-0@z^OlOFu?EpSUePCRD&|761ZQ1
z!dgJJI2Y3tlUOP~jL9<Ibb_*`3btf`J}lm-P6nvwRC!&cupb!S#3|~k^|IdMF8guO
zr`4Ejm2?#y;+_T#iX8*cm6W^hK_&O=GjN%46K>0(;!v=mH@fm%w75pg!a8xM))@76
zg^&yQ*+qjx`7=`nqaYenL*zoe$&W6Ix3xyZ<^qB4mIzZ(o=h7MqjbgW{@ULCwHx=>
z3g?5^<;hL2F1uo|=r4+MbTN@VwY@#H6MAafPN9dU(l=_swj0wL?xX(VYq}CB5n=2M
zcte~OTMXO8Q?Xm1Qmp6oc@6VX3zDYQB%@(c_Iov2#$I`c{j>t#S5p*TF-=n9!LMl+
z+%MQ@1UMvO@epwQML^wh>>5rF#1<Ms@f2@_<>Gg|2{wz2e!eVtHWw$C`0LRDajQNu
zB=l)efqxZ`>9ug8_=Vo3ZXZI$JI6x73(#UjWukJV9VOc05<}n$U~Zlv*2NV<gSa>D
zWB7vD8n1=7#m;!GWX>ekBc6!g3&X^T1~a@XZZvej(W1>56Yi@QrU2D8rxd{J4!0DB
zm!^t!Micu@HMhun3rk7hV1uHhwp<^Tr_U9)8sCB&#L9#{uu_zmOpq5XraV|9`b`eF
zPwX}=hTUSSxdaXq=bJ0wJK|+Ck5QYml46ySH-5v<Mxq@GBoRXs^WdlA;KVq%TAZ9X
zh@RR>pi6#H^X&sC5_#Afw^kU{muk)bLxr$6rEQ6D2cB=LJ8vHc;KHtL=u;_)v6qr%
zv%P>`)A27rTj&s1p_Z5F{|KX_T0#KKu369$1{%KTOYFg6iM9lO1E@9*ND$ruXp^)t
z3DiBsk2u&O<n)Da>lr{T^?ej72#@hZTC2p{faix^NfP1!J(e=-%dx-%Rg_9fR<_*<
z?CU~`%z%V`)i5${f&tKV<OFA_kW7XL5~zeP(qs<tj{}8bpq-eYeK*WGOF=MBjKPI!
z%2f&UatbxRS5xQ4PMKauQ#Q_)b!(!1YrEG5+}fblm|3gO!<dwqPtZE!x>lfiRV8hU
zX*a4=?rlKW0x6^RN6y(yGCDZ94wPR3)RdUJDy1+HsF-<DQ1u5?Wzq`_SRaKFm#b9?
z<bN-GjOxm%4Dx8~Ju|yRhW16Fkw{|p`5h)ITN@slOA}9^4AZZx#-Ni6A^iuK%I!Mn
z{1V2T)&3RQD+Lpz!1ppNdImTW0DOs?6&7QuqM^a04_Cte_T~s0^G)3~+|_`041jSP
zFd11x>5X>BZh+RuwqLJ(E<yDI<AN3XIpQ&kzV(O(=Y_!~<pW^{2To;`XL0x!IhsJD
zNd)ILZ=H=!m~I?HbOiKqUnOz9>BvlTK^Zt-$CfTsz2DdkGEU>*h_j!uu~Jaaaf4m)
z%b%sRo8n;#HG^A%GaTj?VL7mxfW%h<FLBu|?9FoKg`&tHredr;6t%vqqzk;$fYT+x
zU-=Pb0UqbOuj2SJG%387%zGRGy!Q#ZRLuTTmOqnVwY2D_kI~-opdOt|GgN<4?0f?i
z?S=5^{BFfiqOKlPekn+}0)Q12q@X}H8FK15<hSD4;#oh*QD|Kffhilp;THZMEDRS~
zOCLMiIdpy8upy@>17K;~4yYH=(zUn3R9z=+prHnt9Z1t3{gx&7cs%K_<%h}i7&MM0
zsq0!cNJAjSydR|L$1g*>M>_BzgT8&ILCR;a?=$)pv)an#X3xYl=+9ovnmp*>I{ldb
z{3`c7;C&jRxNvHstL`qz(j<|POo#s*q4+m2p`ymNPYe=;0)M6hB{=ptI{zAY28U$&
zx0-~xfZ67r1=!pOJuYI^MfekOOG=z#uR)q5)F5{y$kqGB_frh5m1QGB>5uM*0H&BA
z_KpvLkN~`OU^amSZ+CY(){WLLj|BCOKeCa9m6YT<4kx08Y4|Tq`%4B3@NsXb#ukFH
z8Xz11JGOvA?i0@WOb8XDEtE?JH~BAw2_t~}TCiIMdW~Arj7II*AXF$t@2Jc#?>L6<
z87;!6eE`>HhhFv_OX1ZKxehx`YrP~Yq~i8)tc%*w=FbuXsd2*(_xSHgGI+jO>dW=u
zO;P4%p&uke0<+OH8Qcs!jg#~@j`J5lwf?IX!zL;mdt-#L00yWufd#6w#+4mS|GaUB
zog~N4T2QgkVB%Jb$QDtR8U@oFA2Q2FhYEKdPjdtJhJw{wRhSG0^%82<Q8fIKjDKJK
z5{GU|DcYS33Tq09C*f<Y=wn)?WYY1QG5jQ?Owj7f_z&34a*Yg=sBU9hYerIysu%MU
z9gl*A4*{IWCd#}KpncIW2iqQn7@Av^ii)ePk-1m&+~s<GmyBJL>Y75Z?vD*<SvAUZ
zD5Qps+CO6X;YD;1DKY6-_ooycL45><qi0%M?nsMD0dJPLo7qLA@?<B;&OYE~MLUlo
z!2{5<Qc*2rMfZ;*;SyexxNUg97Psqx099x;jV{pmU*c0jEM&xGLK%EXyeg!GjyAKF
za_Ag!8b2>aq<Q*vHwha6QiyeFF|FNJw!ez~y-n)zmD~d6SZ=RUQaT-f8P=24=DLx=
zFChgq`d}KKlK|=)NqO_}(~uNZ+-C+!y$yX?Cdn&6``ZwofL|RCs&V;niN}Y=#PwV{
z`xe`%LKP}y8o{rED(l7+Nqs#*wnrLVzdsjbXfu4)a$Pd-k@#m?BK)UlN>7sHn7CtN
zMS3yE$AMh-papuZZkgQ&CV9DMD$FCu)Z7%0rh7vZ-284XadF*n+#)J6qPahl%1@}F
zzW4ldq*>2DSGbcaPv#zGbM2xhgJ%<;1*1?+1Q=!6g3`LtB$Fh?Q<N^w&PdYxr%FsE
zr+x+|)>5o3j3&`|-%q4+=D29>l!N5??ES<u8828i97dHz%i{q5h50)ulT>?A#ZpQp
zEi4eY8G4f9ysy*E9YELx`1!&hajiWK=8EsyQ;g?r_T51+fj_~;ZL@5kCj&1!Gao00
zVucJvnNVgZ)`g<|q9#*24oxcD!z~I&<3icBY0yX4ad`>@=PnBZa=bgdYG!I<1|${F
zsm!K<&MQm!V9jhkv2J_AF5(en!H}p#yv1w)Q_+EnOM{9V>i{Wc+Y6hogg55&9pcLx
zheLcmGshqt5#|6iO_~2Q8n3Xv<zy$TdYzE8Y$q-_$nG+b^f^iVcV;yFS=3~C-~_QI
zD+6YUVpes~Utfz4v!dYZqC7hrP7=M@8KIMIQL?V9Xwz}@qPRFa0~*Dn+0pW+_T%9k
z78p>^P7}Y*PJo;!cg%!tvE9LkSqil?vPd0>-wN*;j#i2X9eVv>53B_BcNcMgt~1Wy
zDChJBp%G9~z-#0Z`F@~~Z5MxYXv4%8ynBFlS0ZD_t7Q5IqOm_oH)1jQ6vWB>waQyB
z(!7ua`=Co6n8^M^?Cd`bisHlmNv%7^;`xl!W6D+hCGnc^hdg?JY?sPt(Zhj71EUGG
z0GE*(a+Mb~M()T*JKX$xEh<28o;wBPUvrKLffM}?ry>{Qn?t|$r*m5{$uHK;3vuhh
z!U%8FB7D(F5d24H7WOm_=>TT!j_>Ox8m2XblEqCz3@>$v9XYywuZ(n-AtVz|<n+h(
zdCL;2dTxzf1lT!xG$oSF5f!<cmG3w8K5p0EZ(^s??l$pIt||2LV2W`@g*K>ht9Un8
zfXBsX=XBMho9G%%#UK?!>XRPYc}3jr%*~;5OhJST@=OdNpVi|?8Ie>YBZnoq9#eIR
zrN&(Hw>n&*j`rK?NzK_uJ4H(RtVP`bvjl{|Xt8L39qtm_25_V}eqlg@1bFBZ2Tl#x
z5~4nNx*zk?N%+4hz^(%=-eh?+L%Y*}@C`8dfg44}Fg6ln7YJrRu>`ChFSWtB;=@u6
zbR3YC&Bv|RmPqasF#ck%3;f-hUEtJe=de%}qmo_-f*Eg6`OR(p3MG`SB_)K4Ymo*q
zaSYvf3@~vdT0cu}>-GbN_0m9?0H|M_xBJE1TV4r-p?xDhdkzqi0kzC|fkuBC(LsLf
zpbD~7h=!SSZy=!7i;~9HVDvV9puKU7td+`;7t~Z2nA7ThZgloB3-{k>=H6&FAEXxM
zZ#5j^HwjITQ-vkJSs=`d0@xFR;E!mN4BZRG4?;;-<Tq-*+aAQQ44q%#j2%fawQoy=
z@sAZFT7DvFhVP*2Ne-QG5(w3Oi<T;N-vxY@v59MSeFG<<mt{f`C?`czT;D~81BUCQ
zFJgD(HYi^U8s7(0D&I`ed<ZqgIw#&v4EKNuQGgoC?<Q3#6G73aj-LHyNY7~P8tI5b
zgpd1#wby{`nz64!g1dXkzgB>TQN2agq&*^^N@!QYULWCJ&rd<288VX8VJf3Hju5||
z;;V`p!@ltXS`dyVTBT|ACVX^}#3rB<jp$(lI&}=BYv+<RKNfYwsH7mxzC9E;sbO4g
zIPt}0qfw3;)fRvfRK}sd^pJ&qa$INvHf=zyJtD4b)@8(F!YQg!3<d59N{VfJ)ZNS-
zjYDRskPpgVj-qni&qE=bj2ax{WkFY$nWN`8)5Yt}5wTC{_(#I0_0GivNF_L$)*=&&
zcpttMO|sIruQ*c0h^e~PyJq}VlIs9ohERZ4KX!8gtxQ0F$hfr$z>hxPrr&l6dvY3g
z0OUI^_4#t?{@U_fH08Ko9&a}WyR7JNq|iXfG%5GXA>eliPuZgYM8-UYFc6fFgYNub
zA-!*+@+bSeC4@_*#v_B=FxmS;8>uv8+|!Z%*EDr*cDeY@RGxlEiNdm7spvWJ?$pFq
zla6A2qnE(Vj$E*l@E1GHYPZefDe)P)UB`rPCC>ff-A5)|0@mO#_fG)L{G58v`Ogdz
z1Ma8^D~eQ!LuXR++>apLAxs5D;UIrCekqZE=zn7v@u2({aT-=_fbHI^iG+pyd)igA
zgmU1Rt40Zffm>WRZ4!)LH(i96#P!oX>QP?oZF9y0gWXpK8Wy5?S*54NZ>J|jvPTI-
zq7RVb$~F(D($w#0>J0uz^nsYzvI=&IZ?>cb{W(Vbz9k|2@|k|j%QOCWQ26dZ!i<d}
zg|v}(P%7u>FlM08w`O|>gyI3Ad!_?nLEqq^uvmFDp^wqCO5~8S*|ci{)$Kw;YM;_d
zTwL>brIhn&A~nyhx+)P00Q0eHv2kvy5D#W;69><;;-^^y!OM>9FvcJ}L;7fui!0|@
z>}TjMMw#Nt^`zhzRB}>Cn*{`l{)`4%^i-tEDj8~0$s`^#BYUe<BU@|-hQYw=_)vT>
z&#b-_r6l39#LT+G8+!muEUcLa#N^iO$c24>jieW9+_m%IcrNI@4Qp@1mZ!wItx>R6
z+}zqfcBu}z%qTj{$cHgK%~jz39ME_NT;PD?d=RzSsTO}~%_p(dF~23YELN}&?5%*L
z-LHU_3J``Om609-hI4%$?k}78!F*TTFKLLzz~6i<sW%IhXMt;NKv>;}uv;bq5P^ST
zxUph-TUz913)*aHoD3)t?F%I_gqlURDeMdo=d?8_E<4UjJ_#p@DL>tDPrTo@4d)JO
z^nhNjuWjc5^_f=`w=LKtDe_$^5_1;Lm7FR4=8Sl3p<b5K)cd%NO%d-d)Whqdd{GCC
z6?ZMV0{4n-?fIHD^<*xAaN`%kW!muVDrt?7#&#vK?}}fxb1+W)z1>E*iR8r_C7C58
z%bi}l3obd}Us4Grj#->TJik;g{dl&a1Qc9aDE_=OrFBFWMfYx&{Qq`@93KiI&3`Cp
zhkmEApW(0GBa+!iBpoTRd_|v6Z3~*cEeKUB^f#TocZ%=dDdz4JQ$O`?fcrPV#Th}A
z*zzT!;(8B({fB@sGHCHXfROT^g+a~T`FT$7TwtFI?)w1c>wQanVVPOtbs4?l*=1=Z
zoyy)Rz&-^CKJY(xFyRx(dLqq10`$9|U00G#g(HQ4O$xJy^qD*^SXaLesd%rli{PyM
znRPo!r9vjlpe8tRU9niaTwABkmgM@lm&V}-<+HozH~Ii6z^N<l&mhI>x2KisfSUz|
zJKB9rxfVFDQ`A#13EA=_o?QAj0-;gZEGDYsY+qSN<?eX}QmJq$NrE6zE7g+Hv^Pdu
zBA#2G36F|FD=Y)kMlo-b#PE7`#}d$0j(8P8^KA6fj*Ssi3IBXYN#AX7rqugPx+a3E
z<*PN}f~)WUqCKmn-s+*+{u36s#HlNsBzeESB8h;$Yb%l>yVpqyY9KAG_SBT-wZrJ%
zFHNNF{8HpsCcuxx!j;~zVN=j}B~ao&lsI{y{#CJaWmLrS#{8QET`Q#fSs;}BX9AE(
z{9vWUwrP(l_rfedPf--Po~KpZ-0@IFLe~BHVZe;M1LR~mdXQrP%In}_3Did4iPlvd
z94QvA8VF0o)vMN!{Rr*qEK*5StR4*?iHBEL<kfq)4J%MA@){7yDs&d8OI%_1lHq-U
zBI~1tJmky8*zsCp0x<OJBdQkh_Uzfz7XDn_*OHQ<=#l7KGoEY_99rXotoZ91YsCGT
zu^yj5$N)l$qtEM={=U)%9~E=gx@^{>7X<Bpo~fbRJI>hH05&dBHRkor?&8~;lSe?u
zY|xPa&PicD<6hNId~Izs-P+QCN!QC4i<j1>Nw@p@wgdEYMb&dIQn{5q$J5iQ<Q}uf
zhT6rs&y~V^;s?(;2w@$%E=QJ7yJNeQ#3=nVao9Swym`4b5}+gRYMlDT;*@nIHuEWb
zFA-Fu6KB}cT(mEXri>>Kce)@ozwWd!AN0N-&b=Ql41XNWI$f*K4)KR|``|iptH_XP
zJ}4#;EPq4Hj~MVYDM6cEa96jgF|#j@F;@*0!pjSD*DuLk4~Ya32__;TLJ<ie(vOIg
z2u(yrB$SAph=K?r5=KNxL`4J>2|t{>UOgmAz2FBq^WJXB|4t-5hdW<Ue>%O@UkbF@
z8cCUDhntxo%;?Jt{D;cwA>)$-!U^(+SKY$jgmJV9RYXj8w7r5_30=m1$)liR7`|)+
zYS7U!miM^oYD0hHE@iE|%Jut@b*`^*K^R4o9YVMg4Iw-h$}_<bmM^K}(0nWLRaSwB
z>)Cwp`Bo%k;b-J{Tm{xdbX{;YqN}06j7Mr!GkPL>9#AxOb5cIk*r$cMJ|=sP$0a0j
zv~@RxqIYCqV%-)txr%=YE{AHInaF;tfGLWb_~xjl<?N4g;L(R=f;eO;sU0@Tx)Q)+
znoIm`#Y0e!KUCnMTE!jxEjeoaOJSI4C?rf4H?oZFbgI_6qR}yV+t0V+)g`&rB}p~>
zNch=^O<gLqvy@*ZwRv*!F@e5^Cn*|oEYk$y{WgS>`74AJXZsCgO?rvRN59tg0X=};
z&rHQ%$)}uWmm_pLDL)n2Ye9+As%m@*a9>K%D-rp_P%BN7Ejm(7ayBK|E?S*}wBZd_
z=$4v2C8x6Z{*40w2*))M<_)KwhB=`AY!-Vpu`t1K6ZqL_+72F0p<K^8LMYegq>ex6
zMv`drT(meG#l{#W@x^TYknAO{DR_y4T1kq=&~OM}-w5)z(BoW(`s(m1ZR?KWw&!{t
z5q5!1F3tF<{W4Bg!K>_kDOy0IL}mUU;G0ymcyJKzR&^AS9s8zXz^{KPUZd8&84vvU
zm>OyG;1&KX|13?hk)h_>H!^VOpteZLWc@k_l=0)sIpB5#8!t=k(STGf+35s*BPnfN
zEJaJD`E9Lym*F<*O2$(nG*}T%U7+KUD6{SiwS40G!SW6MAd+mtXv(x_3jnO6>xSU#
zVI&%(9IaGI4{R^89#+8*nzX{8aA=|<4D*Gntm`tFVO|lVrxK{qZNWc8;rDVCM(SPo
z%!uV^mYm;-{)(pe@t4s@@dznfdrne=ad9Y|PCcT+sW9}9sJ0&3lTT5L9kZm=2J0dK
zys%26f}3OF>4B8WniWp1aby5k0O#^wx%QG_wJKnEOd{Hg;D=ogfw`7u*tjq}WGGck
z4JR8WzX+?#0A-~p#WCP9UK{LE(<0X{z56rTKvz}n;K=geu;CWl&rDU)OwZ(|a}cr}
z92*y^Fc_3Q%2hRNz73r;!=><_+RX~BKN9^@WseSWnH8@WaiA%JTb0E$@{PCFk$uCe
z_dhk4u<|t487ML{@hn@3A4KtCnq<0qN1nAIh<cgH(CY$=%#01L4jeKwsr@d8mqM5r
z9r)hNcn6Lg)w2|>l((l?&RWj7<`8;w!t+6OdX+BYP1wll)J4s~1e7<7b<6{*&8Sjp
z9$-%g+VOh25KgSyU)M>-{+qyzL`JXNsBz2zuoFzJyD@eoKuLk_L?$Ytx>zyDUJV=O
zJBI^S#d4oTpd=-h1U^q>bgiLG%>}|7Y0t(2vvXlE{ya>03Y^-s!<zUB`eY{irrKhK
z`obVX1{u|H(IgHX!8Cd4KtR4oZg?clRGz&hO*5BKma)5o;nzqj$Ja4RlH+gk3=i4K
zMmQgx9F9>KMFAW82%|tK%tWVixCOP$Nm$M(Oz2M=RnM>D^q5ekH>6Na7Pa<>%AE}f
zK5p<69&TPDoSwKQ(3QlfTdP>>8E76*^aezirJSHaZ%?qM_oEKE&ahsU7GUS|QnW<x
zTE)G{;mlkt_E)H&g8NYc$KtIDl97D;9M?#cr6ly#qD<gJA;omJvmW5(QT$+39+|<h
zr{r)rHMmE^{xTSblM#Ld>h*zCu8oG9^s6f!lwMz+PKvHW8R#WVz7KWllq6&g@#FI{
z{@oitk8q;NG%HaXYSBuCVTJ1``%$>sQ|+3VRD-uCYTrsez6c0EL8Nf9v9j&&0=_@m
z&eXk$1_xFpGgg=oIGW51%X*K23(@xqJj;mQFTlNdsJamNsP72(<v-#=>6~_vk(*C=
zY2&z&xQw&vz-FgpqM6$j$hR=Dt-re^LOt1K>XB1Lz*f3{I4bOK$<`(y|1N+AsH;u`
z^cA9;P`~&(m#UJX8-AwaF{~x~xw&paz2}F2?@A0qW0!{Dm-RIrm4_VTWO%xd|5wwZ
zx<QQ0S__P&smgkj#V?#1rJd^MNUHZ+(zL(ocu`uCJJi-@Cvf;IB{T))f7N!E>lY0W
zdvkK)1C}6FQVMl_29G*UdVBobS=2tF?ql|1fUz<b^ma6w6W#VF8wlfV_*P)3l~Kpc
zsX&{@P>Kcv7mN4NU01koX<CQ3X5cwl=s=*&%0yUqWOHK&IST;Z<&czl&>6KM3q2TY
z{{>q58>06;xG^1-u;i^2#a7mk$;}R&wlZ;&&Sdu2!1q=rKF%_nZ|mC3onVPCOpAhw
zgL{yl1j$Ljw&QL6wn=1b#2QFXVPaLMJax<OM*u1UJy38=C>s@MNMW*vJwJ-K;KThT
zHDnKE67W#+Y!_NF_{PZ=MVz1rB6%G&+TP^bP>hTnSH!m|X25mof*O-Qtsd!zqgW1P
zwB44}XJFL~c{+JHK|3+<AcgU@{+!D%L0`%FT;SohTG(zgC=lK+#I;!`9fS{a-PSyG
zTvY1|L#bcQy8I4y1>Nv$b~|2`3tu5g6;G9scf6!k=;crXced*2qIP-1@67eyzW{0W
zzX0_*{W9tvf_kI-teO7C6Er6jkYiW4a31zmtLQD-l`S+H*n3L6NeZ2o2)tM0SO$f`
z04nH!{xhZP8S@#g+<}Y=bRlprm5FZ6Ew}-N?mjGo+C$OL(db}u*F3X<;5xT?w0Rso
zi;g0Y9}g_275jF|2@O-v?H$97%EWfnt&#R|P`eqCw>9+j!i^E|M*imnE00swjS|*F
z@+%iUrsx5nGKK5F`Y`~uu0e%s&JI)T)xM+DZc#xmT2MI{oz5c!`veB!Ii{i-(DVUS
zjP_>|bS$kmrJ&y$sCW{*zEhDG$1Vu`;leUSTK?^xGCX&#V>1A0w^YEBk!KCi69z%F
zHt?a1iA>w-;y*x(L#tXL8<c~89FVcN?XtHUl<xr}D)JL?KCNFZ10}|RW&CPu8F6wM
ziVZ{yj9rr4s3kA72~0At$f^5^;z_<LE2#^j3}g#z(ij;1+1mh3#N2yMgaa=MOm@MG
zZumULiEtYqN1xNW@>b1+8SQ6<WC>Zx*ig871F+wdP@8dPDpfft*cTtt06sMI3ShMW
zP=m}^X@v;+^U#-p_%y~8bz*?QfzGPEL*)4NNOK7RDI<$vmc0ldS73Y^V<AasLmHC>
z&jl`|F|lxC;6WNw>ewp)AAOum(X}9Gy&8&kdyvmcFbaKR-+JZamuJx72<>hS@e>pn
zVUXbjJ@vKokW4EJJe$s#6Z0x@b{)(^LnBJ_05F-4qlzJ#BD6NzZw0TX5Y~2Q3Xod-
z>1bBqLOP>EV~Z@Z=C^~<BZ3-naNv)0#%lX6gDr}SiAU$MZv`m~UGIW=>!l3KarV3D
zDGhl1AQx==+_k5W7j*qxf7U*&KoNne3}%ATu?V-Q$7%t!oZ1ci?vsIw8B7yA6>!?g
z+g`5*p0zV6meDzAS}d5@c@VgUlV!1|u;mX3&DMRv<}9pK+m`?|C$xJgx-_HhZLTfw
zsh!cvULA`<!X%y&Tow4s&eYOlYVUsROkmJ6fyPXd%GU=rWio|nwSK~)HT=M~Mm3wj
ztxl>0caCvgA_e6{HZsfgo@x^PRAOcRCr<o>nfk^29UxgLpeEpfIoC|6*tutXIbd0;
zp_8CeLPQP-tjS`Ah5qcEl)(>kei7_E9QZbiNrf$e=xnA~erdEr65Kro&EmE9mQ2lN
zh_FN=h(r?65aEbK5s4-eLqvOcYPK#ilj>YZvnK*;C+pbk_cjMbGja_BG?8@O_y+&9
z(W$96CWt*(%-%oUkKOhUr~xc7%CQ5Wi6(Ey0;Z2>H{J<+Ia$YP7rfVxy)=WE-1#E-
zzfF!`3`9)P>6zmfCI79uxHaJGQ(fFj{kLk}l|Xx+YTcD5RO?p<juMr6{pyhazUofk
z(G*=AbLaRInsxUBiOqfYyZ?l$V@sf}S!ZTDw)D%fpDX@9&$qTSu&r5VQLpVJy8rv=
zKLu_y>*C>0fsm;>W6*-T0qay9&n&nr{cpq7_X6Xll6$Pa_k`+UV8>Kl0`u_O4e-CO
zx*xbf?#td^-1&rR`^7-aG@XIne(_1&#YKUUL>F;!(bgxfyMJJ7;KOM;CeS@ir(($2
z6BgeD{xwaf<@f&}XMfr)A@ljo@f}YXeRjM1@of@>za3=LVkx`2Z}N%a3FAB0_Wa%F
z=;=BoOU|CKxVR<IGhN39p6@fbxMlAX25bM-XJBO4{_9EI>E6Jf({*v|>E0)G>lX*?
zEjoP=`L1fwB?NzT?We$r7M(h<sYRz^$=MS<o!cIGr$uLsIJf<Op3tq_|I6Q3D`)7G
z5#;Pii{JlaQA{k@mA^k}arD)|lo@2=j=uWq6IcIqy-&v^Wc)p-cP{l~Pp6PU{ptFX
zwnx_f_V<7JH|dEXXHVSiivT}UXM|q_3TEm|K`Yw>%`<iJeczjA>Js3?z=@f<IP&#%
zpLX3Ffu5PVaOHod^WQYvLi#1Y+Ggp(gO(nAtPNDo(jj)~!9Pe>|96eBuO6N|OZRU0
F{{y}`LD~QS


From 1ab7a3015776750b9237dd2db49cb7051d920a57 Mon Sep 17 00:00:00 2001
From: Daniel Wigh <dswigh@gmail.com>
Date: Sat, 20 Jan 2024 22:15:12 +0000
Subject: [PATCH 29/29] fix strict mypy

---
 orderly/extract/extractor.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/orderly/extract/extractor.py b/orderly/extract/extractor.py
index a205e83c..ba8e4179 100644
--- a/orderly/extract/extractor.py
+++ b/orderly/extract/extractor.py
@@ -1028,7 +1028,7 @@ def contains_transition_metal(agents: AGENTS) -> bool:
                     return True
             return False
 
-        def contains_charcoal(procedure_details) -> bool:
+        def contains_charcoal(procedure_details: str) -> bool:
             if "charcoal" in procedure_details.lower():
                 return True
             return False