Upgrade to Transformers 4.40 #1027
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
Co-authored-by: regisss <15324346+regisss@users.noreply.github.com>
regisss
requested review from
ssarkar2,
bhargaveede,
vivekgoe,
mandy-li,
libinta,
dvarshney-habana and
ZhaiFeiyue
as code owners
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
ssarkar2
reviewed
May 31, 2024
| eos_token_id = eos_token_id[0] if eos_token_id else None | ||
| if eos_token_id is None and self.generation_config.eos_token_id is not None: | ||
| eos_token_id = self.generation_config.eos_token_id | ||
| stopping_criteria.append(EosTokenCriteria(eos_token_id=eos_token_id)) |
There was a problem hiding this comment.
TODO: add a warning here:
something like:
if generation_config.static_shapes and (generation_config.pad_token_id == generation_config.eos_token_id):
logger.warning(
f"For Gaudi, we pad input_ids with pad_token_id (= {generation_config.pad_token_id}), which is equal to eos_token_id for this model, and EosTokenCriteria stopping criteria is requested, so this option (ignore_eos=False) isn't available. Try setting `--ignore_eos` to False."
)
We can add this in a separate PR later, just making a note/todo here.
| eos_token_id = eos_token_id[0] if eos_token_id else None | ||
| if eos_token_id is None and self.generation_config.eos_token_id is not None: | ||
| eos_token_id = self.generation_config.eos_token_id | ||
| stopping_criteria.append(EosTokenCriteria(eos_token_id=eos_token_id)) |
There was a problem hiding this comment.
same stopping criteria warning as mentioned in greedy()
| @@ -2355,6 +2530,7 @@ def expand_if_needed(tensor, new_size, value, dim=-1): | |||
| "transo_xl", | |||
| "xlnet", | |||
| "cpm", | |||
| "jamba", | |||
| eos_token_id = eos_token_id[0] if eos_token_id else None | ||
| if eos_token_id is None and self.generation_config.eos_token_id is not None: | ||
| eos_token_id = self.generation_config.eos_token_id | ||
| stopping_criteria.append(EosTokenCriteria(eos_token_id=eos_token_id)) |
There was a problem hiding this comment.
same comment as greedy()
| context_mask = 1 - (col_indices >= row_indices + diagonal).int().expand_as( | ||
| mask | ||
| ) # Expand to match mask shape | ||
| context_mask = (col_indices <= row_indices + diagonal).bool().expand_as(mask) # Expand to match mask shape |
There was a problem hiding this comment.
we have a .bool() here instead of .int(). We dont support bool, not sure if this will cause a fallback to cpu, etc
ssarkar2
approved these changes
May 31, 2024
libinta
approved these changes
May 31, 2024
libinta
reviewed
May 31, 2024
| fill_value=1, | ||
| dtype=torch.bool, | ||
| ) | ||
| self.register_buffer("causal_mask", torch.triu(causal_mask, diagonal=1), persistent=False) |
There was a problem hiding this comment.
check why this is removed
This was referenced Jun 4, 2024
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?
As per title.
Before submitting