[Bugfix] Fix Qwen4Exp PLE embedding rejecting INC (AutoRound) checkpoints (#59990)

Signed-off-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>
This commit is contained in:
Vadim Gimpelson
2026-10-05 01:52:06 +00:00
committed by GitHub
co-authored by Claude
parent 49e0f47978
commit 4f52fa35ef
2 changed files with 25 additions and 0 deletions
@@ -27,6 +27,7 @@ from vllm.model_executor.layers.quantization.compressed_tensors.compressed_tenso
CompressedTensorsConfig,
)
from vllm.model_executor.layers.quantization.fp8 import Fp8Config
from vllm.model_executor.layers.quantization.inc import INCConfig
from vllm.model_executor.layers.quantization.modelopt import (
ModelOptMixedPrecisionConfig,
ModelOptQuantConfigBase,
@@ -190,6 +191,11 @@ class Qwen4ExpPLEEmbeddingMethod(QuantizeMethodBase):
and quant_config.get_scheme_dict(None, layer_name=prefix) is None
):
return Qwen4ExpPLEUnquantizedEmbeddingMethod()
if (
isinstance(quant_config, INCConfig)
and not quant_config.config_parser.resolve(None, prefix).quantized
):
return Qwen4ExpPLEUnquantizedEmbeddingMethod()
if not isinstance(quant_config, Fp8Config):
raise NotImplementedError(
"Qwen4Exp PLE embedding does not support quantization config "