Skip to content

Fix noaux_tc cuda Error 700 in CUDAGraph and Add wfp8apf8 moe quant method #4115

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
Jiang-Jia-Jun merged 10 commits into from
Sep 22, 2025
Merged

Fix noaux_tc cuda Error 700 in CUDAGraph and Add wfp8apf8 moe quant method #4115

Jiang-Jia-Jun merged 10 commits into from
Sep 22, 2025

Conversation

@ckl117
Copy link
Collaborator

@ckl117 ckl117 commented Sep 15, 2025 •
edited
Loading

  • Fix noaux_tc cuda Error 700 in CUDAGraph(GLM45-Air)
    -- Add input renormalize to specify whether topk normalization is required in noaux_tc op.
    -- Support stable sort

  • Optimize per_token_quant_fp8 kernel performance improved by 50%

  • Support Wfp8Afp8MoEMethod(weight quant in channel-wise)

python -m fastdeploy.entrypoints.openai.api_server \
    --model ${model_path} \
    --max-model-len 32768 \
    --max-num-seqs 128 \
    --tensor-parallel-size 1 \
    --load_choices "default_v1" \
    --quantization wfp8afp8 \

@paddle-bot
Copy link

paddle-bot bot commented Sep 15, 2025

Thanks for your contribution!

@ckl117 ckl117 changed the title Optimize per_token_quant_fp8 op Optimize per_token_quant_fp8 OP and Support Wfp8Afp8MoEMethod Sep 18, 2025
@ckl117 ckl117 changed the title Optimize per_token_quant_fp8 OP and Support Wfp8Afp8MoEMethod Fix group_topk cuda Error 700 in CUDAGraph and Add wfp8apf8 moe quant method Sep 19, 2025
@ckl117 ckl117 changed the title Fix group_topk cuda Error 700 in CUDAGraph and Add wfp8apf8 moe quant method Fix noaux_tc cuda Error 700 in CUDAGraph and Add wfp8apf8 moe quant method Sep 19, 2025
qingqing01
qingqing01 approved these changes Sep 22, 2025
View reviewed changes
Copy link
Collaborator

@qingqing01 qingqing01 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

后续增强下单测对出现bug case 的覆盖

@Jiang-Jia-Jun Jiang-Jia-Jun merged commit f38b174 into PaddlePaddle:release/2.2 Sep 22, 2025
23 of 27 checks passed
@ckl117 ckl117 mentioned this pull request Sep 24, 2025
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@qingqing01 qingqing01 qingqing01 approved these changes

@zhoutianzi666 zhoutianzi666 zhoutianzi666 approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@ckl117 @qingqing01 @zhoutianzi666 @Jiang-Jia-Jun