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

How to get inference_config parameters while finetune on my own dataset? #3

Open
wsxwd opened this issue Mar 6, 2025 · 1 comment

Comments

@wsxwd
Copy link

wsxwd commented Mar 6, 2025

No description provided.

@GongyeLiu
Copy link
Collaborator

Hi, @wsxwd

The inference_config is primarily used to normalize the reward values. This configuration is optional, the inference code should works fine without it.

VideoAlign/inference.py

Lines 67 to 74 in f9a9654

def _norm(self, reward):
if self.inference_config is None:
return reward
else:
reward['VQ'] = (reward['VQ'] - self.inference_config['VQ_mean']) / self.inference_config['VQ_std']
reward['MQ'] = (reward['MQ'] - self.inference_config['MQ_mean']) / self.inference_config['MQ_std']
reward['TA'] = (reward['TA'] - self.inference_config['TA_mean']) / self.inference_config['TA_std']
return reward

In practice, we perform inference on all the samples in the validation set to compute the mean and standard deviation in inference_config. You might consider applying a similar strategy for your dataset.

I hope this helps!

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

No branches or pull requests

2 participants