The run started yesterday broke several times with the message:
ValueError: Unable to create tensor, you should probably activate truncation
and/or padding with 'padding=True' 'truncation=True' to have batched tensors
with the same length. Perhaps your features (`input_ids` in this case) have
excessive nesting (inputs type `list` where type `int` is expected).
During handling of the above exception, another exception occurred:
The same error can be simulated by passing an input > 2048 (context window of gating model gpt-neo).
This PR proposes the fix to this problem by adding padding=True and setting the eos token as default pad_token on model initialization (required, otherwise we get another exception)
This PR also proposes a fix for the reward model tags in wandb, as they were being logged incorrectly.
The run started yesterday broke several times with the message:
https://wandb.ai/opentensor-dev/openvalidators/runs/j1n1rbcl/logs?workspace=user-
The same error can be simulated by passing an input > 2048 (context window of gating model gpt-neo).
This PR proposes the fix to this problem by adding
padding=True
and setting the eos token as defaultpad_token
on model initialization (required, otherwise we get another exception)This PR also proposes a fix for the reward model tags in wandb, as they were being logged incorrectly.