Unverified Commit 9f430c94 authored by Vadim Gimpelson's avatar Vadim Gimpelson Committed by GitHub
Browse files

[BUGFIX] Add missed remaping of the names of fp8 kv-scale (#32199)


Signed-off-by: default avatarVadim Gimpelson <vadim.gimpelson@gmail.com>
parent f8bd8394
...@@ -64,6 +64,7 @@ from vllm.model_executor.layers.vocab_parallel_embedding import ( ...@@ -64,6 +64,7 @@ from vllm.model_executor.layers.vocab_parallel_embedding import (
) )
from vllm.model_executor.model_loader.weight_utils import ( from vllm.model_executor.model_loader.weight_utils import (
default_weight_loader, default_weight_loader,
maybe_remap_kv_scale_name,
sharded_weight_loader, sharded_weight_loader,
) )
from vllm.model_executor.models.qwen2_moe import Qwen2MoeMLP as Qwen3NextMLP from vllm.model_executor.models.qwen2_moe import Qwen2MoeMLP as Qwen3NextMLP
...@@ -1065,6 +1066,12 @@ class Qwen3NextModel(nn.Module): ...@@ -1065,6 +1066,12 @@ class Qwen3NextModel(nn.Module):
if name.startswith("mtp."): if name.startswith("mtp."):
continue continue
# Remapping the name of FP8 kv-scale.
if name.endswith("scale"):
name = maybe_remap_kv_scale_name(name, params_dict)
if name is None:
continue
for param_name, weight_name, shard_id in stacked_params_mapping: for param_name, weight_name, shard_id in stacked_params_mapping:
if weight_name not in name: if weight_name not in name:
continue continue
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment