why when evaluating a pretrained checkpoint, the test/mean_score in eval_log.json is not equal to the test_mean_score in the checkpoint

[Zack-C-bot]

the env_runner code seems same,but it produces two different results.

[lucywang720]

i ran into the same problem, very strange

[weleen]

I find similar results replied by the author in https://github.com/real-stanford/diffusion_policy/issues/14#issuecomment-1712724428