TensorRT-LLM feat: Update cutlass

Mar 23 '25 11:03 Funatiq

/bot run

Mar 23 '25 11:03 Funatiq

PR_Github #188 [ run ] triggered by Bot

Mar 23 '25 11:03 niukuo

PR_Github #188 [ run ] completed with state FAILURE

Mar 23 '25 11:03 niukuo

@Funatiq Please try 1 hour later. We are performing some maintenance work.

Mar 23 '25 11:03 chzblych

/bot run

Mar 23 '25 13:03 juney-nvidia

/bot run

Mar 23 '25 22:03 juney-nvidia

/bot help

Mar 23 '25 22:03 juney-nvidia

PR_Github #197 [ run ] triggered by Bot

Mar 23 '25 22:03 niukuo

PR_Github #197 [ run ] completed with state FAILURE

Mar 23 '25 22:03 niukuo

/bot run

Mar 24 '25 06:03 Funatiq

PR_Github #248 [ run ] triggered by Bot

Mar 24 '25 06:03 niukuo

PR_Github #248 [ run ] completed with state SUCCESS /LLM/main/L0_MergeRequest_PR pipeline #244 completed with status: 'FAILURE'

Mar 24 '25 07:03 niukuo

/bot run

Mar 24 '25 08:03 Funatiq

/bot run

Mar 24 '25 15:03 Funatiq

PR_Github #319 [ run ] triggered by Bot

Mar 24 '25 16:03 niukuo

PR_Github #319 [ run ] completed with state SUCCESS /LLM/main/L0_MergeRequest_PR pipeline #302 completed with status: 'FAILURE'

Mar 24 '25 18:03 niukuo

/bot run

Mar 25 '25 07:03 Funatiq

PR_Github #394 [ run ] triggered by Bot

Mar 25 '25 07:03 niukuo

PR_Github #394 [ run ] completed with state SUCCESS /LLM/main/L0_MergeRequest_PR pipeline #350 completed with status: 'FAILURE'

Mar 25 '25 09:03 niukuo

/bot run --disable-fail-fast

Mar 25 '25 09:03 Funatiq

PR_Github #424 [ run ] triggered by Bot

Mar 25 '25 09:03 niukuo

Now that we've upgraded to CUTLASS 3.8.0, there's fp4/fp6 definitions so we probably can merge contents in cpp/tensorrt_llm/kernels/internal_cutlass_kernels/src/internal_cutlass_type_conversion.h into cpp/tensorrt_llm/kernels/cutlass_kernels/cutlass_type_conversion.h and remove some release check exemption entry for this file.

We can do it in another PR if you prefer.

Mar 25 '25 10:03 tongyuantongyu

Now that we've upgraded to CUTLASS 3.8.0, there's fp4/fp6 definitions so we probably can merge contents in cpp/tensorrt_llm/kernels/internal_cutlass_kernels/src/internal_cutlass_type_conversion.h into cpp/tensorrt_llm/kernels/cutlass_kernels/cutlass_type_conversion.h and remove some release check exemption entry for this file.

We can do it in another PR if you prefer.

I would prefer to get this PR merged first. It can also help with trtllm-gen integration.

Mar 25 '25 10:03 Funatiq

PR_Github #424 [ run ] completed with state FAILURE /LLM/main/L0_MergeRequest_PR pipeline #365 completed with status: 'FAILURE'

Mar 25 '25 13:03 niukuo

/bot run --stage-list "H100_PCIe-5"

Mar 25 '25 18:03 Funatiq

PR_Github #461 [ run ] triggered by Bot

Mar 25 '25 18:03 niukuo

PR_Github #461 [ run ] completed with state SUCCESS /LLM/main/L0_MergeRequest_PR pipeline #395 (Partly Tested) completed with status: 'SUCCESS'

Mar 25 '25 20:03 niukuo

/bot reuse-pipeline

Mar 26 '25 12:03 Funatiq

PR_Github #589 [ reuse-pipeline ] triggered by Bot

Mar 26 '25 12:03 niukuo

PR_Github #589 [ reuse-pipeline ] completed with state SUCCESS Reusing PR_Github #461 (Partly Tested) for commit fc85c76

Mar 26 '25 13:03 niukuo