FP8 with Tensor Reorg #760
Draft
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
…g ACT_X/PARAM_XL/etc...
…te/etc... still lots of work left though!
…tocasting + make main_stream default. New gpt2_forward() feels clean!
…re matmul work to do...
…rnorm low shared memory fallback + ...
… to fix (famous last words)
… support scaling for BF16...
…rnel (no format conversion, no elementwise)
… think it's correct)
…nd adamw (~70% due to sparse accesses/bank clashes)
…s find out... but not today!)
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Outdated suggestions cannot be applied.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.
Current draft of the massive changes for FP8 with "tensors" being passed around instead of pointers. Only putting this up now so others can give feedback on the overall design & style (not ready to be integrated yet).
It's simpler and more elegant than my previous version of FP8 by far in my opinion, and it solves a few other problems making things cleaner elsewhere. But it does change a lot of code...!