[Quantization] Support online quantization with partially pre-quantized checkpoints (#51392)

Signed-off-by: Felix Marty <Felix.Marty@amd.com>
Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: Codex <codex@openai.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
fxmarty-amd
2026-09-04 11:46:06 -07:00
committed by GitHub
co-authored by Codex mgoin
parent a11dfcff89
commit 5690b02c03
25 changed files with 846 additions and 175 deletions
+2 -1
View File
@@ -24,6 +24,7 @@ from vllm.model_executor.custom_op import PluggableLayer
from vllm.model_executor.determinism.batch_invariant import (
linear_batch_invariant,
)
from vllm.model_executor.layers.quantization import resolve_quant_method
from vllm.model_executor.layers.quantization.base_config import (
QuantizationConfig,
QuantizeMethodBase,
@@ -282,7 +283,7 @@ class LinearBase(PluggableLayer):
self.quant_method: QuantizeMethodBase
if quant_config is None:
self.quant_method = UnquantizedLinearMethod()
elif quant_method := quant_config.get_quant_method(self, prefix=prefix):
elif quant_method := resolve_quant_method(quant_config, self, prefix=prefix):
self.quant_method = quant_method
else:
raise ValueError("All linear layers should support quant method.")