Support for loading in FP8 mode#138
Open
Yahweasel wants to merge 1 commit intoByteDance-Seed:mainfrom
Open
Conversation
Loads the FP16 model with on-the-fly quantization. Because most math isn't supported in FP8, the weights are changed dynamically as needed, so a few checks are added to perform the upconversion.
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Unfortunately, the 7900XTX and other consumer-grade AMD GPUs don't really support bitsandbytes. Or dfloat11. But, FP8 support works fine. With this patch, I can load and run BAGEL on a 7900XTX. Unfortunately, it still takes about 17GB at max, so it's too much for a 16GB card, but it does work, and it's a heck of a lot faster than CPU.
Use --mode 4 to load in FP8 mode. Loads the FP16 model with on-the-fly quantization. Because most math isn't supported in FP8, the weights are changed dynamically as needed, so a few checks are added to perform the upconversion.