Improve random state handling #801
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Squashed commits: [58d1612] Remove seed from config and put it in the "shared" for the ProcessingUnit
ClemDoum
force-pushed
the
task/improve-random-seed
branch
from
e29cc06 to
6ff0eea
Compare
Codecov Report
@@ Coverage Diff @@
## develop #801 +/- ##
===========================================
+ Coverage 88.42% 88.47% +0.04%
===========================================
Files 76 76
Lines 4571 4571
Branches 882 882
===========================================
+ Hits 4042 4044 +2
+ Misses 397 395 -2
Partials 132 132 |
ClemDoum
force-pushed
the
task/improve-random-seed
branch
from
f956a78 to
9e43dfd
Compare
adrienball
requested changes
May 16, 2019
| while True: | ||
| noise_length = int(random_state.normal(mean_length, std_length)) | ||
| i += 1 |
| @@ -35,8 +35,9 @@ def test_should_get_slots(self): | |||
| - make me [number_of_cups:snips/number](five) cups of tea | |||
| - please I want [number_of_cups](two) cups of tea""") | |||
| dataset = Dataset.from_yaml_files("en", [dataset_stream]).json | |||
| config = CRFSlotFillerConfig(random_seed=42) | |||
| config = CRFSlotFillerConfig() | |||
There was a problem hiding this comment.
The config is not needed anymore here.
| @@ -101,10 +104,11 @@ def test_should_get_sub_builtin_slots(self): | |||
| - find an activity from [start](6pm) to [end](8pm) | |||
| - Book me a trip from [start](this friday) to [end](next tuesday)""") | |||
| dataset = Dataset.from_yaml_files("en", [dataset_stream]).json | |||
| config = CRFSlotFillerConfig(random_seed=42) | |||
| config = CRFSlotFillerConfig() | |||
| @@ -65,9 +66,11 @@ def test_should_get_builtin_slots(self): | |||
| - Can you tell me the weather [datetime] please ? | |||
| - what is the weather forecast [datetime] in [location](paris)""") | |||
| dataset = Dataset.from_yaml_files("en", [dataset_stream]).json | |||
| config = CRFSlotFillerConfig(random_seed=42) | |||
| config = CRFSlotFillerConfig() | |||
There was a problem hiding this comment.
The config is not needed anymore here.
| @@ -356,9 +360,10 @@ def test_should_get_slots_after_deserialization(self): | |||
| - i want [number_of_cups] cups of tea please | |||
| - can you prepare [number_of_cups] cups of tea ?""") | |||
| dataset = Dataset.from_yaml_files("en", [dataset_stream]).json | |||
| config = CRFSlotFillerConfig(random_seed=42) | |||
| config = CRFSlotFillerConfig() | |||
| classifier_config = LogRegIntentClassifierConfig(random_seed=42) | ||
| slot_filler_config = CRFSlotFillerConfig(random_seed=42) | ||
| classifier_config = LogRegIntentClassifierConfig() | ||
| slot_filler_config = CRFSlotFillerConfig() | ||
| parser_config = ProbabilisticIntentParserConfig( |
| classifier_config = LogRegIntentClassifierConfig(random_seed=42) | ||
| slot_filler_config = CRFSlotFillerConfig(random_seed=42) | ||
| classifier_config = LogRegIntentClassifierConfig() | ||
| slot_filler_config = CRFSlotFillerConfig() | ||
| parser_config = ProbabilisticIntentParserConfig( |
| @@ -162,9 +169,12 @@ def test_should_get_intents(self): | |||
| utterances: | |||
| - yili yulu yele""") | |||
| dataset = Dataset.from_yaml_files("en", [dataset_stream]).json | |||
| classifier_config = LogRegIntentClassifierConfig(random_seed=42) | |||
| classifier_config = LogRegIntentClassifierConfig() | |||
| parser_config = ProbabilisticIntentParserConfig(classifier_config) | |||
| random_seed=seed1), | ||
| slot_filler_config=CRFSlotFillerConfig(random_seed=seed2) | ||
| intent_classifier_config=LogRegIntentClassifierConfig(), | ||
| slot_filler_config=CRFSlotFillerConfig() | ||
| ) |
docs/source/tutorial.rst
Outdated
| different outputs. | ||
|
|
||
| If you want to run training in a reproducible way you can pass a random seed to | ||
| your engine: |
There was a problem hiding this comment.
I prefer using a more impersonal form in the documentation, but that's just a suggestion. That would be something like:
Reproducible training and testing can be achieved by passing a
**random seed** to the engine:
adrienball
reviewed
May 16, 2019
docs/source/tutorial.rst
Outdated
| @@ -174,6 +174,26 @@ the dataset we generated earlier: | |||
|
|
|||
| engine.fit(dataset) | |||
|
|
|||
| Note that by default, the training of the engine is non-deterministic: if you | |||
| train your NLU twice on the same data and test it on the same input, you'll get | |||
| different outputs. | |||
There was a problem hiding this comment.
I would be a bit more optimistic in the formulation:
Note that, by default, the training of the NLU engine is a non-deterministic process:
training and testing multiple times on the same data may produce different outputs.
ae0633e to
45f4fd4
Compare
adrienball
approved these changes
May 20, 2019
Merged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description:
Currenlty
scikit-learnbug the intent classification training was not deterministicDone:
sklearn==0.21which contains a fix which makesSGDClassifiertraining deterministicChecklist: