Enable modelopt gemma3 nvfp4/fp8, make workflow more robust (#22771)

Signed-off-by: Zhiyu Cheng <zhiyuc@nvidia.com>
Signed-off-by: Michael Goin <mgoin64@gmail.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
This commit is contained in:
Zhiyu
2025-09-19 22:40:33 +00:00
committed by GitHub
co-authored by Michael Goin
parent 711e912946
commit 431535b522
7 changed files with 82 additions and 22 deletions
+5 -2
View File
@@ -31,8 +31,11 @@ logger = init_logger(__name__)
def make_compiler(compilation_config: CompilationConfig) -> CompilerInterface:
if compilation_config.use_inductor:
if envs.VLLM_USE_STANDALONE_COMPILE and is_torch_equal_or_newer(
"2.8.0.dev"):
# Use standalone compile only if requested, version is new enough,
# and the symbol actually exists in this PyTorch build.
if (envs.VLLM_USE_STANDALONE_COMPILE
and is_torch_equal_or_newer("2.8.0.dev")
and hasattr(torch._inductor, "standalone_compile")):
logger.debug("Using InductorStandaloneAdaptor")
return InductorStandaloneAdaptor()
else: