Refactor Clip and OpenAi query encoders (#2014)

Closes #2013

docs/experiments-ance.md

@@ -194,4 +194,4 @@ Top100	accuracy: 0.8522
 + Results reproduced by [@ArthurChen189](https://github.com/ArthurChen189) on 2021-07-06 (commit [`c9f44b`](https://github.com/castorini/pyserini/commit/c9f44b2a24103fff4887cade831f9b7c2472b190))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2022-12-23 (commit [`0c495c`](https://github.com/castorini/pyserini/commit/0c495cf2999dda980eb1f85efa30a4323cef5855))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2023-01-10 (commit [`7dafc4`](https://github.com/castorini/pyserini/commit/7dafc4f918bd44ada3771a5c81692ab19cc2cae9))
-+ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-16 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))
++ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-07 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))

docs/experiments-bpr.md

@@ -62,4 +62,4 @@ Top100 accuracy: 0.8571
 + Results reproduced by [@HAKSOAT](https://github.com/HAKSOAT) on 2022-03-11 (commit [`779668`](https://github.com/castorini/pyserini/commit/77966851755163e36489544fb08f73171e98103f))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2022-12-24 (commit [`0c495c`](https://github.com/castorini/pyserini/commit/0c495cf2999dda980eb1f85efa30a4323cef5855))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2023-01-10 (commit [`7dafc4`](https://github.com/castorini/pyserini/commit/7dafc4f918bd44ada3771a5c81692ab19cc2cae9))
-+ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-16 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))
++ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-07 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))

docs/experiments-distilbert_kd.md

@@ -62,4 +62,4 @@ recall_1000             all     0.9553
 + Results reproduced by [@lintool](https://github.com/lintool) on 2021-04-26 (commit [`854c19`](https://github.com/castorini/pyserini/commit/854c1930ba00819245c0a9fbcf2090ce14db4db0))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2022-12-23 (commit [`0c495c`](https://github.com/castorini/pyserini/commit/0c495cf2999dda980eb1f85efa30a4323cef5855))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2023-01-10 (commit [`7dafc4`](https://github.com/castorini/pyserini/commit/7dafc4f918bd44ada3771a5c81692ab19cc2cae9))
-+ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-16 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))
++ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-07 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))

docs/experiments-distilbert_tasb.md

@@ -62,4 +62,4 @@ recall_1000             all     0.9771
 + Results reproduced by [@lintool](https://github.com/lintool) on 2021-05-28 (commit [`102ed2`](https://github.com/castorini/pyserini/commit/102ed2b2e8770978e4b3e09804913dcffb63c4a7))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2022-12-23 (commit [`0c495c`](https://github.com/castorini/pyserini/commit/0c495cf2999dda980eb1f85efa30a4323cef5855))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2023-01-10 (commit [`7dafc4`](https://github.com/castorini/pyserini/commit/7dafc4f918bd44ada3771a5c81692ab19cc2cae9))
-+ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-16 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))
++ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-07 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))

docs/experiments-dkrr.md

@@ -145,4 +145,4 @@ Running hybrid sparse-dense retrieval with DKKR and [GAR-T5](https://github.com/
 + Results reproduced by [@lintool](https://github.com/lintool) on 2021-02-12 (commit [`52a1e7`](https://github.com/castorini/pyserini/commit/52a1e7f241b7b833a3ec1d739e629c08417a324c))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2022-12-23 (commit [`90676b`](https://github.com/castorini/pyserini/commit/90676b351b47585084aa8136265d02a67ced3803))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2023-01-10 (commit [`7dafc4`](https://github.com/castorini/pyserini/commit/7dafc4f918bd44ada3771a5c81692ab19cc2cae9))
-+ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-16 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))
++ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-07 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))

docs/experiments-dpr.md

@@ -647,4 +647,4 @@ Top100	accuracy: 0.8837
 + Results reproduced by [@manveertamber](https://github.com/manveertamber) on 2022-01-22 (commit [`ef70c6`](https://github.com/castorini/pyserini/commit/ef70c63efd773e87afd9708338827342f4960540))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2022-12-25 (commit [`0c495c`](https://github.com/castorini/pyserini/commit/0c495cf2999dda980eb1f85efa30a4323cef5855))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2023-01-10 (commit [`7dafc4`](https://github.com/castorini/pyserini/commit/7dafc4f918bd44ada3771a5c81692ab19cc2cae9))
-+ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-16 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))
++ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-07 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))

docs/experiments-sbert.md

@@ -112,4 +112,4 @@ recall_1000             all     0.9659
 + Results reproduced by [@lintool](https://github.com/lintool) on 2021-04-26 (commit [`854c19`](https://github.com/castorini/pyserini/commit/854c1930ba00819245c0a9fbcf2090ce14db4db0))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2022-12-23 (commit [`0c495c`](https://github.com/castorini/pyserini/commit/0c495cf2999dda980eb1f85efa30a4323cef5855))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2023-01-10 (commit [`7dafc4`](https://github.com/castorini/pyserini/commit/7dafc4f918bd44ada3771a5c81692ab19cc2cae9))
-+ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-16 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))
++ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-07 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))

docs/experiments-tct_colbert-v2.md

@@ -1,4 +1,4 @@
-# Pyserini: TCT-ColBERTv2 for MS MARCO (V1) Collections
+# Pyserini: Reproducing TCT-ColBERTv2 for MS MARCO V1
 
 This guide provides instructions to reproduce the family of TCT-ColBERT-V2 dense retrieval models described in the following paper:
 
@@ -398,4 +398,4 @@ ndcg_cut_10             all     0.6094
 + Results reproduced by [@lintool](https://github.com/lintool) on 2022-12-25 (commit [`0c495c`](https://github.com/castorini/pyserini/commit/0c495cf2999dda980eb1f85efa30a4323cef5855))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2023-01-10 (commit [`7dafc4`](https://github.com/castorini/pyserini/commit/7dafc4f918bd44ada3771a5c81692ab19cc2cae9))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2023-05-06 (commit [`dcc0ba`](https://github.com/castorini/pyserini/commit/dcc0ba06585a08d7c78cbffac4217b57e170fc3a))
-+ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-16 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))
++ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-07 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))

docs/experiments-tct_colbert.md

@@ -1,4 +1,4 @@
-# Pyserini: TCT-ColBERT for MS MARCO (V1) Collections
+# Pyserini: Reproducing TCT-ColBERT for MS MARCO V1
 
 This guide provides instructions to reproduce the TCT-ColBERT dense retrieval model described in the following paper:
 
@@ -415,4 +415,4 @@ recall_100            	all	0.9083
 + Results reproduced by [@lintool](https://github.com/lintool) on 2022-12-24 (commit [`0c495c`](https://github.com/castorini/pyserini/commit/0c495cf2999dda980eb1f85efa30a4323cef5855))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2023-01-10 (commit [`7dafc4`](https://github.com/castorini/pyserini/commit/7dafc4f918bd44ada3771a5c81692ab19cc2cae9))
 + Results reproduced by [@lintool](https://github.com/lintool) on 2023-05-06 (commit [`dcc0ba`](https://github.com/castorini/pyserini/commit/dcc0ba06585a08d7c78cbffac4217b57e170fc3a))
-+ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-16 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))
++ Results reproduced by [@lintool](https://github.com/lintool) on 2024-10-07 (commit [`3f7609`](https://github.com/castorini/pyserini/commit/3f76099a73820afee12496c0354d52ca6a6175c2))

pyserini/2cr/msmarco-v1-passage.yaml

@@ -1256,7 +1256,7 @@ conditions:
     display: "OpenAI ada2: Faiss flat, cached queries"
     display-html: "OpenAI ada2: Faiss flat, cached queries"
     display-row: "[<a href=\"#\" data-mdb-toggle=\"tooltip\" title=\"Lin et al. (2023) Vector Search with OpenAI Embeddings: Lucene Is All You Need.\">11</a>]"
-    command: python -m pyserini.search.faiss --threads ${sparse_threads} --batch-size ${sparse_batch_size} --index msmarco-v1-passage.openai-ada2 --topics $topics --encoded-queries openai-ada2-$topics --output $output
+    command: python -m pyserini.search.faiss --threads ${dense_threads} --batch-size ${dense_batch_size} --index msmarco-v1-passage.openai-ada2 --topics $topics --encoded-queries openai-ada2-$topics --output $output
     topics:
       - topic_key: msmarco-passage-dev-subset
         eval_key: msmarco-passage-dev-subset
@@ -1279,7 +1279,7 @@ conditions:
     display: "HyDE-OpenAI ada2: Faiss flat, cached queries"
     display-html: "HyDE-OpenAI ada2: Faiss flat, cached queries"
     display-row: "[<a href=\"#\" data-mdb-toggle=\"tooltip\" title=\"Gao et al. (ACL&nbsp;2023) Precise Zero-Shot Dense Retrieval without Relevance Labels.\">12</a>]"
-    command: python -m pyserini.search.faiss --threads ${sparse_threads} --batch-size ${sparse_batch_size} --index msmarco-v1-passage.openai-ada2 --topics $topics --encoded-queries openai-ada2-$topics-hyde --output $output
+    command: python -m pyserini.search.faiss --threads ${dense_threads} --batch-size ${dense_batch_size} --index msmarco-v1-passage.openai-ada2 --topics $topics --encoded-queries openai-ada2-$topics-hyde --output $output
     topics:
       - topic_key: dl19-passage
         eval_key: dl19-passage

pyserini/2cr/msmarco.py

@@ -666,8 +666,7 @@ def run_conditions(args):
 
 if __name__ == '__main__':
     parser = argparse.ArgumentParser(description='Generate regression matrix for MS MARCO corpora.')
-    parser.add_argument('--collection', type=str,
-                        help='Collection = {v1-passage, v1-doc, v2-passage, v2-doc}.', required=True)
+    parser.add_argument('--collection', type=str, help='Collection = {v1-passage, v1-doc, v2-passage, v2-doc}.', required=True)
     # To list all conditions
     parser.add_argument('--list-conditions', action='store_true', default=False, help='List available conditions.')
     # For generating reports

pyserini/encode/__init__.py

@@ -20,16 +20,17 @@
 # Then import these...
 from ._aggretriever import AggretrieverDocumentEncoder, AggretrieverQueryEncoder
 from ._ance import AnceEncoder, AnceDocumentEncoder, AnceQueryEncoder
+from ._arctic import ArcticDocumentEncoder, ArcticQueryEncoder
 from ._auto import AutoQueryEncoder, AutoDocumentEncoder
 from ._bpr import BprQueryEncoder
 from ._cached_data import CachedDataQueryEncoder
+from ._clip import ClipDocumentEncoder, ClipTextEncoder, ClipImageEncoder, ClipQueryEncoder
 from ._cosdpr import CosDprEncoder, CosDprDocumentEncoder, CosDprQueryEncoder
 from ._dkrr import DkrrDprQueryEncoder
 from ._dpr import DprDocumentEncoder, DprQueryEncoder
-from ._openai import OpenAIDocumentEncoder, OpenAIQueryEncoder, OPENAI_API_RETRY_DELAY
+from ._openai import OpenAiDocumentEncoder, OpenAiQueryEncoder, OPENAI_API_RETRY_DELAY
 from ._slim import SlimQueryEncoder
 from ._splade import SpladeQueryEncoder
 from ._tct_colbert import TctColBertDocumentEncoder, TctColBertQueryEncoder
 from ._tok_freq import TokFreqQueryEncoder
 from ._unicoil import UniCoilEncoder, UniCoilDocumentEncoder, UniCoilQueryEncoder
-from ._arctic import ArcticDocumentEncoder, ArcticQueryEncoder

pyserini/encode/__main__.py

@@ -17,11 +17,12 @@
 import argparse
 import sys
 
-from pyserini.encode import DprDocumentEncoder, TctColBertDocumentEncoder, AnceDocumentEncoder, \
-    AggretrieverDocumentEncoder, AutoDocumentEncoder, CosDprDocumentEncoder, JsonlRepresentationWriter, \
-    JsonlCollectionIterator, UniCoilDocumentEncoder, OpenAIDocumentEncoder, ArcticDocumentEncoder, OPENAI_API_RETRY_DELAY
-from pyserini.encode._clip import ClipDocumentEncoder
-from pyserini.encode._faiss import FaissRepresentationWriter
+from pyserini.encode import AutoDocumentEncoder
+from pyserini.encode import ArcticDocumentEncoder, AggretrieverDocumentEncoder, AnceDocumentEncoder, \
+    ClipDocumentEncoder, CosDprDocumentEncoder, DprDocumentEncoder, TctColBertDocumentEncoder, UniCoilDocumentEncoder
+from pyserini.encode import OpenAiDocumentEncoder, OPENAI_API_RETRY_DELAY
+from pyserini.encode import JsonlRepresentationWriter, JsonlCollectionIterator
+from pyserini.encode.optional import FaissRepresentationWriter
 
 encoder_class_map = {
     "dpr": DprDocumentEncoder,
@@ -30,14 +31,15 @@
     "ance": AnceDocumentEncoder,
     "sentence-transformers": AutoDocumentEncoder,
     "unicoil": UniCoilDocumentEncoder,
-    "openai-api": OpenAIDocumentEncoder,
+    "openai-api": OpenAiDocumentEncoder,
     "cosdpr": CosDprDocumentEncoder,
     "auto": AutoDocumentEncoder,
     "clip": ClipDocumentEncoder,
     "contriever": AutoDocumentEncoder,
     "arctic": ArcticDocumentEncoder,
 }
 
+
 def init_encoder(encoder, encoder_class, device, pooling, l2_norm, prefix, multimodal):
     _encoder_class = encoder_class
 

pyserini/encode/_clip.py

@@ -131,3 +131,21 @@ def __init__(self, model_name, device='cuda:0', l2_norm=False, prefix=None, mult
 
     def encode(self, *args, **kwargs):
         return self.encoder.encode(*args, **kwargs)
+
+
+class ClipQueryEncoder(QueryEncoder):
+    """Encodes queries using a CLIP model, supporting both images and texts."""
+
+    def __init__(self, encoder_dir: str = None, encoded_query_dir: str = None, device: str = 'cuda:0',
+                 l2_norm: bool = False, prefix: str = None, multimodal: bool = False, **kwargs):
+        super().__init__(encoded_query_dir)
+        if encoder_dir:
+            self.device = device
+            self.encoder = ClipEncoder(encoder_dir, device, l2_norm, prefix, multimodal)
+            self.has_model = True
+
+        if not self.has_model and not self.has_encoded_query:
+            raise Exception('Neither query encoder model nor encoded queries provided. Please provide at least one')
+
+    def encode(self, query: str):
+        return self.encoder.encode(query).flatten()

pyserini/encode/_openai.py

@@ -30,6 +30,7 @@
 client = openai.OpenAI(api_key=api_key, organization=org_key)
 OPENAI_API_RETRY_DELAY = 5
 
+
 def retry_with_delay(func, delay: int = OPENAI_API_RETRY_DELAY, max_retries: int = 10, errors: tuple = (openai.RateLimitError)):
     def wrapper(*args, **kwargs):
         num_retries = 0
@@ -45,7 +46,8 @@ def wrapper(*args, **kwargs):
                 raise e
     return wrapper
 
-class OpenAIDocumentEncoder(DocumentEncoder):
+
+class OpenAiDocumentEncoder(DocumentEncoder):
     def __init__(self, model_name: str = 'text-embedding-ada-002', tokenizer_name: str = 'cl100k_base', **kwargs):
         self.model = model_name
         self.tokenizer = tiktoken.get_encoding(tokenizer_name)
@@ -62,15 +64,42 @@ def encode(self, texts: List[str], titles = None, max_length: int = 512, **kwarg
         inputs = [embedding[:max_length] for embedding in inputs]
         return self.get_embeddings(inputs)
 
-class OpenAIQueryEncoder(QueryEncoder):
-    def __init__(self, model_name: str = 'text-embedding-ada-002', tokenizer_name: str = 'cl100k_base', device = None):
-        self.model = model_name
-        self.tokenizer = tiktoken.get_encoding(tokenizer_name)
+# class OpenAIQueryEncoder(QueryEncoder):
+#     def __init__(self, model_name: str = 'text-embedding-ada-002', tokenizer_name: str = 'cl100k_base', device = None):
+#         self.model = model_name
+#         self.tokenizer = tiktoken.get_encoding(tokenizer_name)
+#
+#     @retry_with_delay
+#     def get_embedding(self, text: str):
+#         return np.array(client.embeddings.create(input=text, model=self.model)['data'][0]['embedding'])
+#
+#     def encode(self, text: str, max_length: int = 512, **kwargs):
+#         inputs = self.tokenizer.encode(text=text)[:max_length]
+#         return self.get_embedding(inputs)
+
+
+class OpenAiQueryEncoder(QueryEncoder):
+    def __init__(self, encoder_dir: str = None, encoded_query_dir: str = None,
+                 tokenizer_name: str = None, max_length: int = 512, **kwargs):
+        super().__init__(encoded_query_dir)
+        if encoder_dir:
+            api_key = '' if os.getenv("OPENAI_API_KEY") is None else os.getenv("OPENAI_API_KEY")
+            org_key = '' if os.getenv("OPENAI_ORG_KEY") is None else os.getenv("OPENAI_ORG_KEY")
+            self.client = openai.OpenAI(api_key=api_key, organization=org_key)
+            self.model = encoder_dir
+            self.tokenizer = tiktoken.get_encoding(tokenizer_name)
+            self.max_length = max_length
+            self.has_model = True
+        if (not self.has_model) and (not self.has_encoded_query):
+            raise Exception('Neither query encoder model nor encoded queries provided. Please provide at least one')
 
     @retry_with_delay
     def get_embedding(self, text: str):
-        return np.array(client.embeddings.create(input=text, model=self.model)['data'][0]['embedding'])
+        return np.array(self.client.embeddings.create(input=text, model=self.model)['data'][0]['embedding'])
 
-    def encode(self, text: str, max_length: int = 512, **kwargs):
-        inputs = self.tokenizer.encode(text=text)[:max_length]
-        return self.get_embedding(inputs)
+    def encode(self, query: str, **kwargs):
+        if self.has_model:
+            inputs = self.tokenizer.encode(text=query)[:self.max_length]
+            return self.get_embedding(inputs)
+        else:
+            return super().encode(query)

pyserini/encode/optional/__init__.py

@@ -0,0 +1,18 @@
+#
+# Pyserini: Reproducible IR research with sparse and dense representations
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+from ._pca import PcaEncoder
+from ._faiss import FaissRepresentationWriter