main
txt 9 lines 449 Bytes
Raw
1 # GFX950 model configurations for GSM8K evaluation
2 # TP8 models only
3 # DeepSeek-V4-Pro-NVFP4.yaml is temporarily disabled: repeated CI failures due to
4 # cold NFS reads of the 850GiB checkpoint exceeding startup_max_wait_seconds
5 # (see https://github.com/vllm-project/vllm/pull/47972). Re-enable once the
6 # startup timeout fix has landed and is verified green.
7 # DeepSeek-V4-Pro-NVFP4.yaml
8 DeepSeek-V4-Flash-NVFP4.yaml
9 Qwen3-30B-A3B-NVFP4-quark.yaml