8000 Weight Tying raises errors when running distributed checkpointing · Issue #348 · Modalities/modalities · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

Weight Tying raises errors when running distributed checkpointing #348

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

Jump to bottom
Open
le1nux opened this issue Apr 13, 2025 · 0 comments
Open

Weight Tying raises errors when running distributed checkpointing #348

le1nux opened this issue Apr 13, 2025 · 0 comments

Comments

@le1nux
Copy link
Member
le1nux commented Apr 13, 2025

NOTE: weight tying does not seem to work with distributed checkpointing

E         File "/raid/s3/opengptx/max_lue/repositories/modalities/src/modalities/checkpointing/stateful/app_state.py", line 105, in load_state_dict
E           OptimizerStateRetriever.load_state_dict_(
E         File "/raid/s3/opengptx/max_lue/repositories/modalities/src/modalities/checkpointing/stateful/app_state.py", line 175, in load_state_dict_
E           set_optimizer_state_dict(
E         File "/raid/s3/opengptx/max_lue/venvs/modalities_311/lib/python3.11/site-packages/torch/distributed/checkpoint/state_dict.py", line 1246, in set_optimizer_state_dict
E           _load_optim_state_dict(model, optimizers, optim_state_dict, info)
E         File "/raid/s3/opengptx/max_lue/venvs/modalities_311/lib/python3.11/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
E           return func(*args, **kwargs)
E                  ^^^^^^^^^^^^^^^^^^^^^
E         File "/raid/s3/opengptx/max_lue/venvs/modalities_311/lib/python3.11/site-packages/torch/distributed/checkpoint/state_dict.py", line 891, in _load_optim_state_dict
E           optim_state_dict = _unflatten_optim_state_dict(
E                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
E         File "/raid/s3/opengptx/max_lue/venvs/modalities_311/lib/python3.11/site-packages/torch/distributed/checkpoint/state_dict.py", line 716, in _unflatten_optim_state_dict
E           cast(DictValueType, state[fqn])[state_name] = state_dict[
E                                                         ^^^^^^^^^^^
E       KeyError: 'state.transformer.lm_head.weight.step'

Originally posted by @le1nux in #336
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@le1nux
0