dnrops
/
transformers
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

updated logging and saving metrics (#10436) 

* updated logging and saving metrics

* space removal
This commit is contained in:
Bhadresh Savani 2021-02-27 23:23:44 +05:30 committed by GitHub
parent f52a15897b
commit aca6288ff4
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
12 changed files with 70 additions and 190 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
examples/language-modeling/run_clm.py
View File

@ -375,16 +375,11 @@ def main():
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload
output_train_file = os.path.join(training_args.output_dir, "train_results.txt")
if trainer.is_world_process_zero():
with open(output_train_file, "w") as writer:
logger.info("***** Train results *****")
for key, value in sorted(train_result.metrics.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
metrics = train_result.metrics
# Need to save the state, since Trainer.save_model saves only the tokenizer with the model
trainer.state.save_to_json(os.path.join(training_args.output_dir, "trainer_state.json"))
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
# Evaluation
results = {}
@ -396,13 +391,8 @@ def main():
perplexity = math.exp(eval_output["eval_loss"])
results["perplexity"] = perplexity
output_eval_file = os.path.join(training_args.output_dir, "eval_results_clm.txt")
if trainer.is_world_process_zero():
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results *****")
for key, value in sorted(results.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
trainer.log_metrics("eval", results)
trainer.save_metrics("eval", results)
return results

23
examples/language-modeling/run_mlm.py
View File

@ -411,17 +411,11 @@ def main():
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload
metrics = train_result.metrics
output_train_file = os.path.join(training_args.output_dir, "train_results.txt")
if trainer.is_world_process_zero():
with open(output_train_file, "w") as writer:
logger.info("***** Train results *****")
for key, value in sorted(train_result.metrics.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
# Need to save the state, since Trainer.save_model saves only the tokenizer with the model
trainer.state.save_to_json(os.path.join(training_args.output_dir, "trainer_state.json"))
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
# Evaluation
results = {}
@ -433,13 +427,8 @@ def main():
perplexity = math.exp(eval_output["eval_loss"])
results["perplexity"] = perplexity
output_eval_file = os.path.join(training_args.output_dir, "eval_results_mlm.txt")
if trainer.is_world_process_zero():
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results *****")
for key, value in sorted(results.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
trainer.log_metrics("eval", results)
trainer.save_metrics("eval", results)
return results

23
examples/language-modeling/run_plm.py
View File

@ -392,17 +392,11 @@ def main():
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload
metrics = train_result.metrics
output_train_file = os.path.join(training_args.output_dir, "train_results.txt")
if trainer.is_world_process_zero():
with open(output_train_file, "w") as writer:
logger.info("***** Train results *****")
for key, value in sorted(train_result.metrics.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
# Need to save the state, since Trainer.save_model saves only the tokenizer with the model
trainer.state.save_to_json(os.path.join(training_args.output_dir, "trainer_state.json"))
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
# Evaluation
results = {}
@ -414,13 +408,8 @@ def main():
perplexity = math.exp(eval_output["eval_loss"])
results["perplexity"] = perplexity
output_eval_file = os.path.join(training_args.output_dir, "eval_results_plm.txt")
if trainer.is_world_process_zero():
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results *****")
for key, value in sorted(results.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
trainer.log_metrics("eval", results)
trainer.save_metrics("eval", results)
return results

25
examples/multiple-choice/run_swag.py
View File

@ -227,6 +227,8 @@ def main():
# Set the verbosity to info of the Transformers logger (on main process only):
if is_main_process(training_args.local_rank):
transformers.utils.logging.set_verbosity_info()
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
logger.info("Training/evaluation parameters %s", training_args)
# Set seed before initializing model.
@ -367,17 +369,11 @@ def main():
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload
metrics = train_result.metrics
output_train_file = os.path.join(training_args.output_dir, "train_results.txt")
if trainer.is_world_process_zero():
with open(output_train_file, "w") as writer:
logger.info("***** Train results *****")
for key, value in sorted(train_result.metrics.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
# Need to save the state, since Trainer.save_model saves only the tokenizer with the model
trainer.state.save_to_json(os.path.join(training_args.output_dir, "trainer_state.json"))
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
# Evaluation
results = {}
@ -386,13 +382,8 @@ def main():
results = trainer.evaluate()
output_eval_file = os.path.join(training_args.output_dir, "eval_results_swag.txt")
if trainer.is_world_process_zero():
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results *****")
for key, value in sorted(results.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
trainer.log_metrics("eval", results)
trainer.save_metrics("eval", results)
return results

10
examples/multiple-choice/run_tf_multiple_choice.py
View File

@ -206,14 +206,10 @@ def main():
result = trainer.evaluate()
output_eval_file = os.path.join(training_args.output_dir, "eval_results.txt")
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results *****")
for key, value in result.items():
logger.info(" %s = %s", key, value)
writer.write("%s = %s\n" % (key, value))
trainer.log_metrics("eval", results)
trainer.save_metrics("eval", results)
results.update(result)
results.update(result)
return results

25
examples/question-answering/run_qa.py
View File

@ -201,6 +201,8 @@ def main():
# Set the verbosity to info of the Transformers logger (on main process only):
if is_main_process(training_args.local_rank):
transformers.utils.logging.set_verbosity_info()
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
logger.info("Training/evaluation parameters %s", training_args)
# Set seed before initializing model.
@ -479,16 +481,10 @@ def main():
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload
output_train_file = os.path.join(training_args.output_dir, "train_results.txt")
if trainer.is_world_process_zero():
with open(output_train_file, "w") as writer:
logger.info("***** Train results *****")
for key, value in sorted(train_result.metrics.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
# Need to save the state, since Trainer.save_model saves only the tokenizer with the model
trainer.state.save_to_json(os.path.join(training_args.output_dir, "trainer_state.json"))
metrics = train_result.metrics
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
# Evaluation
results = {}
@ -496,13 +492,8 @@ def main():
logger.info("*** Evaluate ***")
results = trainer.evaluate()
output_eval_file = os.path.join(training_args.output_dir, "eval_results.txt")
if trainer.is_world_process_zero():
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results *****")
for key, value in sorted(results.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
trainer.log_metrics("eval", results)
trainer.save_metrics("eval", results)
return results

25
examples/question-answering/run_qa_beam_search.py
View File

@ -200,6 +200,8 @@ def main():
# Set the verbosity to info of the Transformers logger (on main process only):
if is_main_process(training_args.local_rank):
transformers.utils.logging.set_verbosity_info()
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
logger.info("Training/evaluation parameters %s", training_args)
# Set seed before initializing model.
@ -518,16 +520,10 @@ def main():
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload
output_train_file = os.path.join(training_args.output_dir, "train_results.txt")
if trainer.is_world_process_zero():
with open(output_train_file, "w") as writer:
logger.info("***** Train results *****")
for key, value in sorted(train_result.metrics.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
# Need to save the state, since Trainer.save_model saves only the tokenizer with the model
trainer.state.save_to_json(os.path.join(training_args.output_dir, "trainer_state.json"))
metrics = train_result.metrics
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
# Evaluation
results = {}
@ -535,13 +531,8 @@ def main():
logger.info("*** Evaluate ***")
results = trainer.evaluate()
output_eval_file = os.path.join(training_args.output_dir, "eval_results.txt")
if trainer.is_world_process_zero():
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results *****")
for key, value in sorted(results.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
trainer.log_metrics("eval", results)
trainer.save_metrics("eval", results)
return results

22
examples/text-classification/run_glue.py
View File

@ -417,16 +417,9 @@ def main():
trainer.save_model() # Saves the tokenizer too for easy upload
output_train_file = os.path.join(training_args.output_dir, "train_results.txt")
if trainer.is_world_process_zero():
with open(output_train_file, "w") as writer:
logger.info("***** Train results *****")
for key, value in sorted(metrics.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
# Need to save the state, since Trainer.save_model saves only the tokenizer with the model
trainer.state.save_to_json(os.path.join(training_args.output_dir, "trainer_state.json"))
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
# Evaluation
eval_results = {}
@ -443,13 +436,8 @@ def main():
for eval_dataset, task in zip(eval_datasets, tasks):
eval_result = trainer.evaluate(eval_dataset=eval_dataset)
output_eval_file = os.path.join(training_args.output_dir, f"eval_results_{task}.txt")
if trainer.is_world_process_zero():
with open(output_eval_file, "w") as writer:
logger.info(f"***** Eval results {task} *****")
for key, value in sorted(eval_result.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
trainer.log_metrics("eval", eval_result)
trainer.save_metrics("eval", eval_result)
eval_results.update(eval_result)

14
examples/text-classification/run_tf_glue.py
View File

@ -247,18 +247,10 @@ def main():
results = {}
if training_args.do_eval:
logger.info("*** Evaluate ***")
result = trainer.evaluate()
output_eval_file = os.path.join(training_args.output_dir, "eval_results.txt")
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results *****")
for key, value in result.items():
logger.info(" %s = %s", key, value)
writer.write("%s = %s\n" % (key, value))
results.update(result)
trainer.log_metrics("eval", result)
trainer.save_metrics("eval", result)
results.update(result)
return results

14
examples/text-classification/run_tf_text_classification.py
View File

@ -293,18 +293,10 @@ def main():
results = {}
if training_args.do_eval:
logger.info("*** Evaluate ***")
result = trainer.evaluate()
output_eval_file = os.path.join(training_args.output_dir, "eval_results.txt")
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results *****")
for key, value in result.items():
logger.info(" %s = %s", key, value)
writer.write("%s = %s\n" % (key, value))
results.update(result)
trainer.log_metrics("eval", result)
trainer.save_metrics("eval", result)
results.update(result)
return results

26
examples/text-classification/run_xnli.py
View File

@ -291,33 +291,19 @@ def main():
trainer.save_model() # Saves the tokenizer too for easy upload
output_train_file = os.path.join(training_args.output_dir, "train_results.txt")
if trainer.is_world_process_zero():
with open(output_train_file, "w") as writer:
logger.info("***** Train results *****")
for key, value in sorted(metrics.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
# Need to save the state, since Trainer.save_model saves only the tokenizer with the model
trainer.state.save_to_json(os.path.join(training_args.output_dir, "trainer_state.json"))
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
# Evaluation
eval_results = {}
if training_args.do_eval:
logger.info("*** Evaluate ***")
eval_result = trainer.evaluate(eval_dataset=eval_dataset)
output_eval_file = os.path.join(training_args.output_dir, "eval_results_xnli.txt")
if trainer.is_world_process_zero():
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results xnli *****")
for key, value in sorted(eval_result.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
trainer.log_metrics("eval", eval_result)
trainer.save_metrics("eval", eval_result)
eval_results.update(eval_result)
return eval_results

31
examples/token-classification/run_ner.py
View File

@ -387,18 +387,12 @@ def main():
else:
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
metrics = train_result.metrics
trainer.save_model() # Saves the tokenizer too for easy upload
output_train_file = os.path.join(training_args.output_dir, "train_results.txt")
if trainer.is_world_process_zero():
with open(output_train_file, "w") as writer:
logger.info("***** Train results *****")
for key, value in sorted(train_result.metrics.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
# Need to save the state, since Trainer.save_model saves only the tokenizer with the model
trainer.state.save_to_json(os.path.join(training_args.output_dir, "trainer_state.json"))
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
# Evaluation
results = {}
@ -407,13 +401,8 @@ def main():
results = trainer.evaluate()
output_eval_file = os.path.join(training_args.output_dir, "eval_results_ner.txt")
if trainer.is_world_process_zero():
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results *****")
for key, value in results.items():
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
trainer.log_metrics("eval", results)
trainer.save_metrics("eval", results)
# Predict
if training_args.do_predict:
@ -429,12 +418,8 @@ def main():
for prediction, label in zip(predictions, labels)
]
output_test_results_file = os.path.join(training_args.output_dir, "test_results.txt")
if trainer.is_world_process_zero():
with open(output_test_results_file, "w") as writer:
for key, value in sorted(metrics.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
trainer.log_metrics("test", metrics)
trainer.save_metrics("test", metrics)
# Save predictions
output_test_predictions_file = os.path.join(training_args.output_dir, "test_predictions.txt")