Turn off autotune for scaled mm for fp8 dynamic quant in torchao #2116
+4
−0
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
Summary:
Currently we do autotune in torch.compile (max-autotune-no-cudagraph), that will call triton
kernel autotune for fp8 dynamic quant, but it increases compilation time significantly and
not much benefit for llama models, so we turn this off for now
Test Plan:
python3 -m sglang.bench_latency --model meta-llama/Meta-Llama-3-8B --batch-size 1 --input 128 --output 8 --json-model-override-args '{"architectures": ["TorchNativeLlamaForCausalLM"]}' --enable-torch-compile --torchao-config fp8dq-per_row
Reviewers:
Subscribers:
Tasks:
Tags:
jerryzh168
requested review from
merrymercy,
Ying1123,
hnyls2002,
zhyncs,
ispobock and
ByronHsu
as code owners
|
Note: this will turn off fp8 autotune everywhere actually, but it might be fine for now, we could revisit if this is an issue for other models |
merrymercy
reviewed
Nov 21, 2024
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Outdated suggestions cannot be applied.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.
Summary:
Currently we do autotune in torch.compile (max-autotune-no-cudagraph), that will call triton kernel autotune for fp8 dynamic quant, but it increases compilation time significantly and not much benefit for llama models, so we turn this off for now
Test Plan:
python3 -m sglang.bench_latency --model meta-llama/Meta-Llama-3-8B --batch-size 1 --input 128 --output 8 --json-model-override-args '{"architectures": ["TorchNativeLlamaForCausalLM"]}' --enable-torch-compile --torchao-config fp8dq-per_row
Reviewers:
Subscribers:
Tasks:
Tags: