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.

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

Error:Exception: MoE JIT is designed to work on sample size = 800, while receiving sample size = 1600 (> 800) #96

Open
Satan012 opened this issue Feb 18, 2022 · 2 comments
Labels
question Further information is requested

Comments

@Satan012
Copy link

No description provided.

@ghostplant
Copy link
Contributor

It is expected to set an upper bound of sample size when creating tutel.moe_layer(..).
If not set by default, the upper bound will be automatically set to be the batch size from the first time of data forwarding.

To set the upper bound manual in your case, please add this line after the creation of tutel.moe_layer(..):

class MyModel(Module):

  def __init__(self):
    ...
    self._moe_layer = tutel_moe.moe_layer(..)
    self._moe_layer.expected_sample_size = 1600  # Add this line to set the upper bound manually
    ...

@ghostplant ghostplant added the question Further information is requested label Feb 18, 2022
@ghostplant
Copy link
Contributor

ghostplant commented Feb 26, 2022

In v0.1.5, expected_sample_size will be automatic decided by MoE Layer, so that you no longer need to manually set self._moe_layer.expected_sample_size = 1600.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
question Further information is requested
Projects
None yet
Development

No branches or pull requests

2 participants