Skip to content

LeViT safetensors load is broken by conversion code that wasn't deactivated #2412

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Jan 16, 2025
Merged

LeViT safetensors load is broken by conversion code that wasn't deactivated #2412

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 12 additions & 11 deletions timm/models/levit.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -763,17 +763,18 @@ def checkpoint_filter_fn(state_dict, model):
# filter out attn biases, should not have been persistent
state_dict = {k: v for k, v in state_dict.items() if 'attention_bias_idxs' not in k}

D = model.state_dict()
out_dict = {}
for ka, kb, va, vb in zip(D.keys(), state_dict.keys(), D.values(), state_dict.values()):
if va.ndim == 4 and vb.ndim == 2:
vb = vb[:, :, None, None]
if va.shape != vb.shape:
# head or first-conv shapes may change for fine-tune
assert 'head' in ka or 'stem.conv1.linear' in ka
out_dict[ka] = vb

return out_dict
# NOTE: old weight conversion code, disabled
# D = model.state_dict()
# out_dict = {}
# for ka, kb, va, vb in zip(D.keys(), state_dict.keys(), D.values(), state_dict.values()):
# if va.ndim == 4 and vb.ndim == 2:
# vb = vb[:, :, None, None]
# if va.shape != vb.shape:
# # head or first-conv shapes may change for fine-tune
# assert 'head' in ka or 'stem.conv1.linear' in ka
# out_dict[ka] = vb

return state_dict


model_cfgs = dict(
Expand Down
Loading