[Model][Quantization] Add Ling-3.0-flash-fp8 support (#51265)

Signed-off-by: Isotr0py <Isotr0py@outlook.com>
Co-authored-by: Isotr0py <Isotr0py@outlook.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
zexplorerhj
2026-08-09 23:09:08 -07:00
committed by GitHub
co-authored by Isotr0py mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent 3b4c86e489
commit ba1cdcfcf0
6 changed files with 298 additions and 36 deletions
@@ -1,7 +1,7 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from typing import TYPE_CHECKING, Any
from typing import TYPE_CHECKING, Any, Literal
import torch
@@ -108,6 +108,9 @@ class Fp8Config(QuantizationConfig):
raise ValueError(f"Unsupported activation scheme {activation_scheme}")
self.activation_scheme = activation_scheme
self.ignored_layers = ignored_layers or []
self.ignored_layers_match_mode: Literal["exact", "substring", "suffix"] = (
"exact"
)
self.store_dtype = store_dtype
if weight_block_size is not None:
if not is_checkpoint_fp8_serialized:
@@ -177,6 +180,7 @@ class Fp8Config(QuantizationConfig):
prefix=prefix,
ignored_layers=self.ignored_layers,
fused_mapping=self.packed_modules_mapping,
match_mode=self.ignored_layers_match_mode,
):
return UnquantizedLinearMethod()
if not self.is_checkpoint_fp8_serialized:
@@ -196,6 +200,7 @@ class Fp8Config(QuantizationConfig):
prefix=prefix,
ignored_layers=self.ignored_layers,
fused_mapping=self.packed_modules_mapping,
match_mode=self.ignored_layers_match_mode,
):
return UnquantizedFusedMoEMethod(layer.moe_config)
if self.store_dtype == "mxfp4":