[Misc] Remove no-op self-assignments across vLLM (#55988)

Signed-off-by: shaolila <shaolila@buaa.edu.cn>
Co-authored-by: shaolila <shaolila@buaa.edu.cn>
This commit is contained in:
Adababy
2026-09-16 02:17:07 -07:00
committed by GitHub
co-authored by shaolila
parent 5f203baedb
commit ceb87de065
4 changed files with 9 additions and 12 deletions
@@ -615,14 +615,15 @@ class RoutedExperts(PluggableLayer):
# compressed-tensors checkpoints with packed weights are stored flipped
# TODO (mgoin): check self.quant_method.quant_config.quant_format
# against known CompressionFormat enum values that have this quality
if quant_method_name in (
"CompressedTensorsWNA16MoEMethod",
"CompressedTensorsW4A16FlydslMoEMethod",
if (
quant_method_name
in (
"CompressedTensorsWNA16MoEMethod",
"CompressedTensorsW4A16FlydslMoEMethod",
)
and is_transposed
):
if is_transposed:
loaded_weight = loaded_weight.t().contiguous()
else:
loaded_weight = loaded_weight
loaded_weight = loaded_weight.t().contiguous()
if shard_id not in ("w1", "w2", "w3"):
raise ValueError(f"shard_id must be ['w1','w2','w3'] but got {shard_id}.")