vllm nightly currently not supporting Blackwell with this model

#3
by 1anH - opened

NotImplementedError: No compiled cutlass_scaled_mm for CUDA device capability: 120. Required capability: 90 or 100

https://github.com/vllm-project/vllm/issues/32109

Unsloth AI org

Oh hm someone on Reddit also reported this - I tagged the vLLM team

Sign up or log in to comment