You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi!
When I run the evaluation part of HumanML3D with the prompt python -m eval.eval_humanml --model_path ./save/humanml_trans_enc_512/model000475000.pt --device 0, there's a NAN Error.
Whether it's Replication 0 or Replication 1, they all show the same problem.
And it should be a number, not NaN, as the screenshot attached shows, which is from the official evaluation file eval_humanml_trans_enc_512_000475000_gscale2.5_wo_mm.log in ./save/humanml_trans_enc_512
It's really puzzling.
I am looking forward to your help, and thank you in advance!
The text was updated successfully, but these errors were encountered:
And with the screenshot of the detailed error message.
I really want to know what caused Nan error to occur, not only in the official checkpoint but also in the own ckpt after train_MDM
CMY-CTO
changed the title
nan error in eval_HumamML3D
nan error in eval_T2M_HumamML3D
Feb 2, 2024
Hi!


When I run the evaluation part of HumanML3D with the prompt
python -m eval.eval_humanml --model_path ./save/humanml_trans_enc_512/model000475000.pt --device 0
, there's aNAN Error
.Whether it's Replication 0 or Replication 1, they all show the same problem.
And it should be a number, not NaN, as the screenshot attached shows, which is from the official evaluation file

eval_humanml_trans_enc_512_000475000_gscale2.5_wo_mm.log
in./save/humanml_trans_enc_512
It's really puzzling.
I am looking forward to your help, and thank you in advance!
The text was updated successfully, but these errors were encountered: