Skip to content
Benchmarks / Performance

Avoid unfused Transpose in DeepSeekV3 EP256 MoE layer #90

Sign in to view logs

Avoid unfused Transpose in DeepSeekV3 EP256 MoE layer

Avoid unfused Transpose in DeepSeekV3 EP256 MoE layer #90

Run time

Learn about OS pricing on GitHub Actions
Job Run time
Benchmarks/vLLM=${{ matrix.vllm_branch }}, vLLM-Ascend=${{ matrix.vllm_ascend_branch }}
-1s
-1s