Skip to content

Conversation

vllmellm
Copy link
Contributor

@vllmellm vllmellm commented May 30, 2025

This PR removes the unnecessary constraint and assertion of a specific max_model_len value from the ROCM_AITER_MLA attention backend on both VLLM v1 and v0 engines.

lm_eval results on DeepSeek-V2-Lite-Chat with default engine args on both VLLM v0 and v1 engines.

VLLM_USE_V1=1 VLLM_ROCM_USE_AITER=1 VLLM_ROCM_USE_AITER_MOE=0 VLLM_ROCM_USE_AITER_RMSNORM=0 VLLM_ROCM_USE_AITER_LINEAR=0 lm_eval --model vllm --model_args pretrained=deepseek-ai/DeepSeek-V2-Lite-Chat,tensor_parallel_size=1,block_size=1 --trust_remote_code --tasks gsm8k --num_fewshot 5 --batch_size auto

Tasks Version Filter n-shot Metric Value Stderr
gsm8k 3 flexible-extract 5 exact_match 0.6657 ± 0.0130
strict-match 5 exact_match 0.6543 ± 0.0131

VLLM_USE_V1=0 VLLM_ROCM_USE_AITER=1 VLLM_ROCM_USE_AITER_MOE=0 VLLM_ROCM_USE_AITER_RMSNORM=0 VLLM_ROCM_USE_AITER_LINEAR=0 lm_eval --model vllm --model_args pretrained=deepseek-ai/DeepSeek-V2-Lite-Chat,tensor_parallel_size=1,block_size=1 --trust_remote_code --tasks gsm8k --num_fewshot 5 --batch_size auto

Tasks Version Filter n-shot Metric Value Stderr
gsm8k 3 flexible-extract 5 exact_match 0.5974 ± 0.0135
strict-match 5 exact_match 0.5876 ± 0.0136

Copy link

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

@mergify mergify bot added the v1 label May 30, 2025
@vllm-bot vllm-bot merged commit 77b6e74 into vllm-project:main May 30, 2025
9 of 12 checks passed
amitm02 pushed a commit to amitm02/vllm that referenced this pull request Jun 1, 2025
amitm02 pushed a commit to amitm02/vllm that referenced this pull request Jun 1, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants