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

Add tests for no_trainer and fix existing examples (#16656)

Browse Source 
* Fixed some bugs involving saving during epochs
* Added tests mimicking the existing examples tests
* Added in json exporting to all `no_trainer` examples for consistency
This commit is contained in:
Zachary Mueller
2022-04-08 10:03:56 -04:00
committed by GitHub
parent ab229663b5
commit d57da99237
11 changed files with 414 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
examples/pytorch/text-classification/run_glue_no_trainer.py
View File

@@ -14,6 +14,7 @@
# limitations under the License.
""" Finetuning a 🤗 Transformers model for sequence classification on GLUE."""
import argparse
import json
import logging
import math
import os
@@ -150,7 +151,6 @@ def parse_args():
"--hub_model_id", type=str, help="The name of the repository to keep in sync with the local `output_dir`."
)
parser.add_argument("--hub_token", type=str, help="The token to use to push to the Model Hub.")
parser.add_argument("--hub_token", type=str, help="The token to use to push to the Model Hub.")
parser.add_argument(
"--checkpointing_steps",
type=str,
@@ -488,7 +488,10 @@ def main():
if isinstance(checkpointing_steps, int):
if completed_steps % checkpointing_steps == 0:
accelerator.save_state(f"step_{completed_steps}")
output_dir = f"step_{completed_steps}"
if args.output_dir is not None:
output_dir = os.path.join(args.output_dir, output_dir)
accelerator.save_state(output_dir)
if completed_steps >= args.max_train_steps:
break
@@ -526,7 +529,10 @@ def main():
)
if args.checkpointing_steps == "epoch":
accelerator.save_state(f"epoch_{epoch}")
output_dir = f"epoch_{epoch}"
if args.output_dir is not None:
output_dir = os.path.join(args.output_dir, output_dir)
accelerator.save_state(output_dir)
if args.output_dir is not None:
accelerator.wait_for_everyone()
@@ -557,6 +563,10 @@ def main():
eval_metric = metric.compute()
logger.info(f"mnli-mm: {eval_metric}")
if args.output_dir is not None:
with open(os.path.join(args.output_dir, "all_results.json"), "w") as f:
json.dump({"eval_accuracy": eval_metric["accuracy"]}, f)
if __name__ == "__main__":
main()