Add DeepGEMM to Dockerfile in vllm-base image (#21533)

Signed-off-by: Matthew Bonanni <mbonanni001@gmail.com>
Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
Matthew Bonanni
2025-07-31 18:01:55 -07:00
committed by GitHub
co-authored by mgoin
parent c3e0e9337e
commit e360316ab9
4 changed files with 46 additions and 7 deletions
+12
View File
@@ -17,6 +17,17 @@ from vllm.platforms import current_platform
from vllm.utils import has_deep_gemm
@functools.cache
def is_deep_gemm_supported() -> bool:
"""Return ``True`` if DeepGEMM is supported on the current platform.
Currently, only Hopper and Blackwell GPUs are supported.
"""
supported_arch = current_platform.is_cuda() and (
current_platform.is_device_capability(90)
or current_platform.is_device_capability(100))
return has_deep_gemm() and supported_arch
@functools.cache
def is_blackwell_deep_gemm_used() -> bool:
"""Return ``True`` if vLLM is configured to use DeepGEMM on a
@@ -142,4 +153,5 @@ __all__ = [
"fp8_m_grouped_gemm_nt_masked",
"per_block_cast_to_fp8",
"is_blackwell_deep_gemm_used",
"is_deep_gemm_supported",
]