[Feature][Humming] Humming feature integration (#56685)

Signed-off-by: jinzhen.ljz <jinzhen.ljz@antgroup.com>
Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
Jinzhen Lin
2026-09-20 08:11:01 -07:00
committed by GitHub
co-authored by Codex mgoin
parent 27757dde02
commit 7d99c2c4fd
62 changed files with 2040 additions and 1086 deletions
@@ -24,7 +24,7 @@ class HummingLinearKernel(MPLinearKernel):
return True, None
def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
from vllm.model_executor.layers.quantization.utils.humming_utils import (
from vllm.model_executor.layers.quantization.utils.humming import (
convert_linear_layer_to_humming_standard,
get_humming_linear_compute_config,
prepare_humming_linear_layer_config,
@@ -57,13 +57,14 @@ class HummingLinearKernel(MPLinearKernel):
x: torch.Tensor,
bias: torch.Tensor | None = None,
) -> torch.Tensor:
from vllm.model_executor.layers.quantization.utils.humming_utils import (
from vllm.model_executor.layers.quantization.utils.humming import (
apply_humming_linear,
)
return apply_humming_linear(
layer,
x,
skip_bias_add=bias is None,
layer_config=self.layer_config,
compute_config=self.compute_config,
locks=self.locks,