Add flash-attention patch for falcon-7b #3580
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
andreaskoepf
requested review from
theblackcat102,
sanagno,
dvruette and
yk
as code owners
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Outdated suggestions cannot be applied.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.
Enable the
use_flash_attentionconfiguration flag for Falcon models. Whenuse_flash_attentionis set totruethe FalconAttention.forwad() method is replaced with a variant that uses Tri Dao's flash_attention instead of pytorch'sscaled_dot_product_attentionfunction.At the moment the patch works only for falcon-7b but technically it will also work for falcon-40b with the right configuration. The falcon model situation is currently a bit messy: The Falcon model was recently added to Huggingface transformers (see PR transformers#24523) but the falcon models on the hugginface hub use still the code which is shipped together with the weights (a PR to change this was reverted). Falcon-7b and 40b use both slightly different code (which was unified in the HF transformers impl and can there be controlled via a configuration member called
new_decoder_architecturesee configuration_falcon.py#L65-L67). The HF Falcon impl uses different names in the configuration class, e.g. compare new configuration_falcon.py and old configuration_RW.pyHF Falcon implementation compatible model configurations can be found here:
7B: config.json
40B: config.json