Skip to content
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

Fix progress callback deepcopy #32070

Merged
merged 4 commits into from
Jul 19, 2024
Merged

Fix progress callback deepcopy #32070

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
b0b4108
Replacing ProgressCallbacks deepcopy with a shallowcopy
fozziethebeat Jul 18, 2024
4406c93
Using items instead of entries
fozziethebeat Jul 18, 2024
bdf0525
code cleanup for copy in trainer callback
fozziethebeat Jul 18, 2024
3251bb5
Style fix for ProgressCallback
fozziethebeat Jul 18, 2024
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
16 changes: 9 additions & 7 deletions src/transformers/trainer_callback.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,6 @@
Callbacks to use with the Trainer class and customize the training loop.
"""

import copy
import dataclasses
import json
from dataclasses import dataclass
Expand Down Expand Up @@ -617,13 +616,16 @@ def on_predict(self, args, state, control, **kwargs):

def on_log(self, args, state, control, logs=None, **kwargs):
if state.is_world_process_zero and self.training_bar is not None:
# avoid modifying the logs object as it is shared between callbacks
logs = copy.deepcopy(logs)
_ = logs.pop("total_flos", None)
# make a shallow copy of logs so we can mutate the fields copied
# but avoid doing any value pickling.
shallow_logs = {}
for k, v in logs.items():
shallow_logs[k] = v
_ = shallow_logs.pop("total_flos", None)
# round numbers so that it looks better in console
if "epoch" in logs:
logs["epoch"] = round(logs["epoch"], 2)
self.training_bar.write(str(logs))
if "epoch" in shallow_logs:
shallow_logs["epoch"] = round(shallow_logs["epoch"], 2)
self.training_bar.write(str(shallow_logs))

def on_train_end(self, args, state, control, **kwargs):
if state.is_world_process_zero:
Expand Down