fix fHAS_FLASH_ATTN_V2_ROCM flag bug for DCU

server/text_generation_server/utils/flash_attn.py

@@ -45,7 +45,7 @@ if IS_CUDA_SYSTEM or IS_ROCM_SYSTEM:
                 "Use the official Docker image (ghcr.io/huggingface/text-generation-inference:latest) "
                 f"or install flash attention v2 with `cd server && make install install-flash-attention-v2{architecture_suffix}`"
             )
-        if not (is_sm8x or is_sm90):
+        if not (is_sm8x or is_sm90) and IS_CUDA_SYSTEM:
             raise ImportError(
                 f"GPU with CUDA capability {major} {minor} is not supported for "
                 "Flash Attention V2"