[Frontend] new online quantization frontend (#38138)

Signed-off-by: Vasiliy Kuznetsov <vasiliy@meta.com>
This commit is contained in:
Vasiliy Kuznetsov
2026-04-03 11:58:39 -04:00
committed by GitHub
parent 97f92c6b47
commit 7b1a7423be
13 changed files with 1205 additions and 0 deletions
+9
View File
@@ -112,6 +112,7 @@ from vllm.v1.sample.logits_processor import LogitsProcessor
from vllm.version import __version__ as VLLM_VERSION
if TYPE_CHECKING:
from vllm.config.quantization import OnlineQuantizationConfigArgs
from vllm.model_executor.layers.quantization import QuantizationMethods
from vllm.model_executor.model_loader import LoadFormats
from vllm.usage.usage_lib import UsageContext
@@ -483,6 +484,7 @@ class EngineArgs:
hf_overrides: HfOverrides = get_field(ModelConfig, "hf_overrides")
tokenizer_revision: str | None = ModelConfig.tokenizer_revision
quantization: QuantizationMethods | str | None = ModelConfig.quantization
quantization_config: "dict[str, Any] | OnlineQuantizationConfigArgs | None" = None
allow_deprecated_quantization: bool = ModelConfig.allow_deprecated_quantization
enforce_eager: bool = ModelConfig.enforce_eager
disable_custom_all_reduce: bool = ParallelConfig.disable_custom_all_reduce
@@ -661,6 +663,12 @@ class EngineArgs:
if isinstance(self.ir_op_priority, dict):
self.ir_op_priority = IrOpPriorityConfig(**self.ir_op_priority)
from vllm.config.quantization import resolve_online_quant_config
self.quantization_config = resolve_online_quant_config(
self.quantization, self.quantization_config
)
# Setup plugins
from vllm.plugins import load_general_plugins
@@ -1431,6 +1439,7 @@ class EngineArgs:
tokenizer_revision=self.tokenizer_revision,
max_model_len=self.max_model_len,
quantization=self.quantization,
quantization_config=self.quantization_config,
allow_deprecated_quantization=self.allow_deprecated_quantization,
enforce_eager=self.enforce_eager,
enable_return_routed_experts=self.enable_return_routed_experts,