Skip to content

Conversation

jacquesqiao
Copy link
Member

@jacquesqiao jacquesqiao commented May 15, 2018

fix: #10648

# trainer.save_params(save_dirname)
# https://github.com/PaddlePaddle/Paddle/pull/10445
trainer.save_inference_model(save_dirname)
trainer.save_params(save_dirname)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I noticed we switched from trainer.save_inference_model(save_dirname) back to trainer.save_params(save_dirname). What is the diff between them? We discussed this here. To me save_inference_model looks more reasonable.

Copy link
Member Author

@jacquesqiao jacquesqiao May 15, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

#10648
@wangkuiyi Think that we only need to save parameters and do not need to save inference program, inference program should be provided by a Python function.

num_epochs=1,
event_handler=event_handler,
reader=train_reader,
feed_order=['img', 'label'])
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

test_recognize_digits_conv fails on this line.

@daming-lu daming-lu mentioned this pull request May 16, 2018
@jacquesqiao
Copy link
Member Author

copy and merged by #10674

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

High-level API Trainer constructor cannot be like require both trainer_program and inference_program

2 participants