1*H100 with vLLM 0.19.0 Failed
#7
by JeffreySheng - opened
The model card announce that：
Software Integration:
Supported Runtime Engine(s):
vLLM
Supported Hardware Microarchitecture Compatibility:
NVIDIA Blackwell
...
Preferred Operating System(s):
Linux
Model Version(s):
The model version is v1.0 which NVFP4 quantized with nvidia-modelopt v0.42.0
...
Inference:
Engine: vLLM
Test Hardware: NVIDIA Hopper H100
Im very confused that you said supported hardware must based on NVIDIA Blackwell, but you inference the evaluation on NVIDIA Hopper H100 with vLLM? Thats not make sense.
And I run this with vLLM on 1*H100 is failed.
Is this an official mistaken announcement or something?