"tests/vscode:/vscode.git/clone" did not exist on "2ce3d0ce360bf1e31e7ea3b2b0971a37bf5707c6"
Unverified Commit 74056039 authored by Wei Zhao's avatar Wei Zhao Committed by GitHub
Browse files

Fix minimax m2.5 nvfp4 kv scales weight loading (#37214)


Signed-off-by: default avatarwzhao18 <wzhao18.sz@gmail.com>
parent d7d51a7e
......@@ -439,6 +439,17 @@ class MiniMaxM2Model(nn.Module):
if is_pp_missing_parameter(name, self):
continue
# Remap qkv_proj.[kv]_scale to attn.[kv]_scale
if name.endswith((".k_scale", ".v_scale")):
remapped_name = maybe_remap_kv_scale_name(name, params_dict)
if remapped_name is not None and remapped_name in params_dict:
param = params_dict[remapped_name]
weight_loader = getattr(
param, "weight_loader", default_weight_loader
)
weight_loader(param, loaded_weight)
break
param = params_dict[name]
weight_loader = param.weight_loader
weight_loader(param, loaded_weight, shard_id)
......
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