Fix tokenizers caching #502
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
thomwolf
approved these changes
Aug 17, 2020
src/nlp/utils/py_utils.py
Outdated
| if _transformers_available: | ||
| import transformers as tr | ||
|
|
||
| if isinstance(obj, (tr.CTRLTokenizer, tr.GPT2Tokenizer, tr.OpenAIGPTTokenizer, tr.XLMTokenizer)): |
There was a problem hiding this comment.
This list might be a bit cumbersome to maintain in the future.
Should we just do a check that the cache attribute exists and is a dict maybe?
There was a problem hiding this comment.
I changed it so that it checks if cache exists and is a dict
| co_filename = "" if obj.co_filename.startswith("<") else obj.co_filename | ||
| co_firstlineno = 1 if obj.co_filename.startswith("<") else obj.co_firstlineno | ||
| co_filename = "" if obj.co_filename.startswith("<") or obj.co_name == "<lambda>" else obj.co_filename | ||
| co_firstlineno = 1 if obj.co_filename.startswith("<") or obj.co_name == "<lambda>" else obj.co_firstlineno |
thomwolf
approved these changes
Aug 19, 2020
This was referenced Jul 7, 2021
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Outdated suggestions cannot be applied.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.
I've found some cases where the caching didn't work properly for tokenizers:
unique_no_split_tokens's attribute is not the same across sessions (after loading a tokenizer) then the caching could be inconsistentTo fix that, this is what I did:
save_regexfunction for pickle that makes regex dumps deterministicunique_no_split_tokensdeterministic in Sort unique_no_split_tokens to make it deterministic transformers#6461I also added tests to make sure that tokenizers hashing works as expected.
In the future we should find a way to test if hashing also works across session (maybe using two CI jobs ? or by hardcoding a tokenizer's hash ?)