Add FP8 Blockwise GEMM Backend Flag --fp8-gemm-backend (#14379)

This commit is contained in:
b8zhong
2025-12-09 12:05:56 -08:00
committed by GitHub
parent 73df7a4e8d
commit 55504df2f7
10 changed files with 306 additions and 33 deletions
+22
View File
@@ -368,6 +368,15 @@ def _print_deprecated_env(new_name: str, old_name: str):
os.environ[new_name] = os.environ[old_name]
def _warn_deprecated_env_to_cli_flag(env_name: str, suggestion: str):
"""Warn when a deprecated environment variable is used.
This is for env vars that are deprecated in favor of CLI flags.
"""
if env_name in os.environ:
warnings.warn(f"Environment variable {env_name} is deprecated. {suggestion}")
def _convert_SGL_to_SGLANG():
_print_deprecated_env("SGLANG_LOG_GC", "SGLANG_GC_LOG")
_print_deprecated_env(
@@ -388,6 +397,19 @@ def _convert_SGL_to_SGLANG():
_convert_SGL_to_SGLANG()
_warn_deprecated_env_to_cli_flag(
"SGLANG_ENABLE_FLASHINFER_FP8_GEMM",
"It will be completely removed in 0.5.7. Please use '--fp8-gemm-backend=flashinfer_trtllm' instead.",
)
_warn_deprecated_env_to_cli_flag(
"SGLANG_ENABLE_FLASHINFER_GEMM",
"It will be completely removed in 0.5.7. Please use '--fp8-gemm-backend=flashinfer_trtllm' instead.",
)
_warn_deprecated_env_to_cli_flag(
"SGLANG_SUPPORT_CUTLASS_BLOCK_FP8",
"It will be completely removed in 0.5.7. Please use '--fp8-gemm-backend=cutlass' instead.",
)
def example_with_exit_stack():
# Use this style of context manager in unit test