Adding bf16 support to CUDA #40
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
ikawrakow
force-pushed
the
ik/cuda_bf16
branch
from
ba91320 to
6bfd451
Compare
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Outdated suggestions cannot be applied.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.
Haha,
llama.cppseems to not supportbf16on CUDA?This PR adds it. It works fine on my
RTX-4080, but I have no idea if it will work on older GPUs (if I understood correctly it should, with reduced performance), ROCm, etc.Performance is the same as
f16for TG (TG-128 = 41.2 t/s for LLaMA-3.1-8B for both).PP is lower but quite decent for prompt processing (PP-512(
bf16) = 5250 t/s vs PP-512(f16) = 7250 t/s for LLaMA-3.1-8B). In any case, much better than running on the CPU forbf16models.