SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Fix slicing for 0-dim param (#36580)

Browse Source 
* fix

* switch to ellipsis instead

* Add co-author
Co-authored-by: fxmarty-amd <fxmarty-amd@users.noreply.github.com>

* Add co-author second try
Co-authored-by: fxmarty-amd <felmarty@amd.com>
This commit is contained in:
Marc Sun
2025-03-13 12:16:13 +01:00
committed by GitHub
parent fbb18ce68b
commit bc3d5781e7
2 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/transformers/integrations/tensor_parallel.py
View File

@@ -531,7 +531,7 @@ def shard_and_distribute_module(
param, empty_param, param_type, param_casting_dtype, is_contiguous, rank, device_mesh param, empty_param, param_type, param_casting_dtype, is_contiguous, rank, device_mesh
) )
else: else:
param = param[:] param = param[...]
if is_contiguous: if is_contiguous:
param = param.contiguous() param = param.contiguous()

2
src/transformers/modeling_utils.py
View File

@@ -795,7 +795,7 @@ def _load_state_dict_into_meta_model(
device_mesh, device_mesh,
) )
else: else:
param = param[:] param = param[...]
if casting_dtype is not None: if casting_dtype is not None:
param = param.to(casting_dtype) param = param.to(casting_dtype)
if to_contiguous: if to_contiguous: