[Bugfix] Wait for the linear bias before layerwise online processing (#49805)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Harry Mellor
2026-07-25 17:56:32 +00:00
committed by GitHub
co-authored by Claude Opus 5
parent 26d725c334
commit 9321aff536
4 changed files with 67 additions and 10 deletions
@@ -14,7 +14,7 @@ from vllm.model_executor.layers.quantization.base_config import QuantizeMethodBa
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from .meta import (
SKIP_TENSORS,
SKIP_LOAD_TENSORS,
capture_layer_to_meta,
get_numel_loaded,
materialize_layer,
@@ -140,7 +140,7 @@ def _wrap_parameters_weight_loader(layer: torch.nn.Module) -> None:
"""Wrap each parameter's weight loader."""
# Note that nested wrapping will occur for shared tensors
for name, tensor in get_layer_tensors(layer).items():
if name in SKIP_TENSORS:
if name in SKIP_LOAD_TENSORS:
continue
if _get_weight_loader(tensor).__name__ != "online_process_loader":
tensor.weight_loader = make_online_process_loader(layer, name)
@@ -20,20 +20,23 @@ __all__ = [
"get_numel_loaded",
]
# Modules whose tensors are never moved to, or materialized from, the meta device.
SKIP_MODULES: set[str] = {"HadamardTransform"}
SKIP_TENSORS: set[str] = {
# Tensors never loaded by a weight loader, so the layerwise trigger ignores them.
SKIP_LOAD_TENSORS: set[str] = {
"_expert_map",
"expert_mask",
"expert_global_to_physical",
"expert_physical_to_global",
"expert_local_to_global",
"e_score_correction_bias",
# Built after create_weights(), so it is not tracked by the layerwise-reload
# trigger and would be re-materialized into uninitialized memory. Skip it.
"bias",
}
# Tensors which are never moved to, or materialized from, the meta device.
# `bias` is built after create_weights(), so it is never on meta to begin with.
SKIP_TENSORS: set[str] = SKIP_LOAD_TENSORS | {"bias"}
def to_meta_tensor(tensor: torch.Tensor) -> torch.Tensor:
"""Convert a tensor to a meta tensor while preserving class and attributes."""
@@ -33,15 +33,15 @@ def get_layer_params_buffers(layer: torch.nn.Module) -> LayerTensors:
def get_layer_size(layer: torch.nn.Module) -> int:
"""Calculate total number of elements across loadable tensors in a layer.
Excludes SKIP_TENSORS (e.g. _expert_map) which are never moved to meta
device and never loaded via weight_loader during layerwise reload.
Excludes SKIP_LOAD_TENSORS (e.g. _expert_map) which are never loaded via
weight_loader during layerwise reload.
"""
from .meta import SKIP_TENSORS
from .meta import SKIP_LOAD_TENSORS
return sum(
tensor.numel()
for name, tensor in get_layer_tensors(layer).items()
if name not in SKIP_TENSORS
if name not in SKIP_LOAD_TENSORS
)