Move get_mask_sizes from Cache to masking_utils and remove use of get_seq_length. #39142
Conversation
This file contains hidden or 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
…yeredCache (huggingface#38077) - Introduces CacheLayer and Cache base classes - Ports Static, Dynamic, Offloaded, Quantized, Hybrid, etc. to use layers - Implements method/attr dispatch across layers to reduce boilerplate - Adds CacheProcessor hooks for offloading, quantization, etc. - Updates and passes tests
manueldeprada
force-pushed
the
cache-move-mask-sizes-out
branch
from
6b6314d to
f09e0cd
Compare
|
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. |
manueldeprada
force-pushed
the
cache-move-mask-sizes-out
branch
4 times, most recently
from
c030aa2 to
b78affa
Compare
manueldeprada
force-pushed
the
cache-move-mask-sizes-out
branch
from
b78affa to
16a6624
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.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.
This PR depends on #39106
Look at the last commit, f09e0cd:
I think having the get_mask_sizes out of cache makes much more sense. There is only one extra change:
transformers/src/transformers/masking_utils.py
Line 643 in 6b6314d
It substitutes
past_seen_tokens=past_key_values.get_seq_length()(which depends on cache info that might be hard to cumpute, i.e., QuantizedCaches). What we would like to compute isbut that is not compatible with
torch.export.The new solution is
torch.exportfriendly and works both whencache_position = torch.tensor([ 0, 1, 2, 3, 4, 5, 6])(prefill phase) and whencache_position = torch.tensor([16]).