Skip to content

Conversation

@xinghai-sun
Copy link
Contributor

@xinghai-sun xinghai-sun commented Jun 15, 2017

  1. Add shuffle type of instance_shuffle and batch_shuffle_clipped to help investigate problems like Abnormal learning curve bumping at early batches of each epoch during DS2 training. #100.
  2. Enable min_batch_size to temporarily avoid "batch_size < trainer_count" error (A permanent fix should be done on Paddle).
  3. Add argument print.
@xinghai-sun xinghai-sun requested a review from qingqing01 June 15, 2017 09:13
type=distutils.util.strtobool,
help="Use sortagrad or not. (default: %(default)s)")
parser.add_argument(
"--shuffle_method",
Copy link
Collaborator

Choose a reason for hiding this comment

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

We should choose the best default mode after experiments.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Yes. Will do that after we get a conclusion.

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

Labels

None yet

2 participants