Add MiniMax-M3 modelopt nvfp4 support (#46756)
Signed-off-by: Xin Li <xinli@nvidia.com> Signed-off-by: jasonlizhengjian <jasonlizhengjian@gmail.com> Co-authored-by: Xin Li <xinli@nvidia.com>
This commit is contained in:
@@ -18,6 +18,7 @@ from vllm.model_executor.layers.linear import UnquantizedLinearMethod
|
||||
from vllm.model_executor.layers.quantization.modelopt import (
|
||||
ModelOptFp8Config,
|
||||
ModelOptMixedPrecisionConfig,
|
||||
ModelOptMxFp8Config,
|
||||
ModelOptNvFp4Config,
|
||||
ModelOptNvFp4LinearMethod,
|
||||
)
|
||||
@@ -84,6 +85,11 @@ def _mixed_precision_config(quantized_layers: dict) -> ModelOptMixedPrecisionCon
|
||||
kv_cache_quant_algo=None,
|
||||
exclude_modules=[],
|
||||
),
|
||||
mxfp8_config=ModelOptMxFp8Config(
|
||||
is_checkpoint_mxfp8_serialized=True,
|
||||
kv_cache_quant_algo=None,
|
||||
exclude_modules=[],
|
||||
),
|
||||
)
|
||||
|
||||
|
||||
|
||||
Reference in New Issue
Block a user