Update tiny model creation script #27674
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
ydshieh
commented
Nov 23, 2023
Comment on lines
+426
to
+429
| if "vocab_size" in model_tester_kwargs: | ||
| if "text_kwargs" in inspect.signature(model_tester_class.__init__).parameters.keys(): | ||
| vocab_size = model_tester_kwargs.pop("vocab_size") | ||
| model_tester_kwargs["text_kwargs"] = {"vocab_size": vocab_size} |
There was a problem hiding this comment.
This was original in convert_tokenizer with if config_class.__name__ in [...] .
Move that logic to here and without using a list of specific config classes --> make it more generic, so we don't need to update this script frequently.
ydshieh
commented
Nov 23, 2023
Comment on lines
+449
to
+467
| # make sure this is long enough (some model tester has `20` for this attr.) to pass `text-generation` | ||
| # pipeline tests. | ||
| max_positions = [] | ||
| for key in ["max_position_embeddings", "max_source_positions", "max_target_positions"]: | ||
| if getattr(config, key, 0) > 0: | ||
| max_positions.append(getattr(config, key)) | ||
| if getattr(config, "text_config", None) is not None: | ||
| if getattr(config.text_config, key, None) is not None: | ||
| max_positions.append(getattr(config.text_config, key)) | ||
| if len(max_positions) > 0: | ||
| max_position = max(200, min(max_positions)) | ||
| for key in ["max_position_embeddings", "max_source_positions", "max_target_positions"]: | ||
| if getattr(config, key, 0) > 0: | ||
| setattr(config, key, max_position) | ||
| if getattr(config, "text_config", None) is not None: | ||
| if getattr(config.text_config, key, None) is not None: | ||
| setattr(config.text_config, key, max_position) | ||
|
|
||
| return config |
There was a problem hiding this comment.
This is to avoid the failing in TextGenerationPipelineTests at
outputs = text_generator("This is a test" * 500, handle_long_generation="hole", max_new_tokens=20)(failing happens if a model tester has max_position_embeddings=20)
Also taking max_source_positions and max_target_positions into account (e.g. whisper)
ydshieh
commented
Nov 23, 2023
| elif ( | ||
| # `FuyuConfig` saves data under both FuyuConfig and its `text_config`. This is not good, but let's just update | ||
| # every involved fields to avoid potential failure. | ||
| if ( |
There was a problem hiding this comment.
elif --> if just make sure config.text_config is also updated (for fuyu)
|
The documentation is not available anymore as the PR was closed or 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.
What does this PR do?
A few fixes or improvements I found necessary while working on #27388.
See comments along the changes