-
Notifications
You must be signed in to change notification settings - Fork 17
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
Remora model finetuning #171
Comments
The resume feature requires a checkpoint model not a torch script model (optimized for inference, and lacking training status). You can recreate the checkpoint with the following snippet. I will try to add this to the core API to make this a bit simpler.
|
Thanks for clarifying! I've managed to produce the checkpoint file to use in the remora training command. Yet, I am running into this other error now. Any idea from where the issue is coming from?
Reference command line:
|
Training from a checkpoint file requires that the same data input size be used. Setting the |
I have double checked, both the pre-trained model and the training dataset have the same |
Could you post the exact command and error message here to help resolve the issue? |
here the full command line I used:
here the full error message arising after the last command:
|
I've made some minor changes around this logic in the latest version. Could you upgrade and report if this is resolved in the latest version? |
Update using the new released
|
Hello,
I am trying to finetune this model to call 6mA (
dna_r10.4.1_e8.2_5khz_400bps_sup_v4.2.0_6ma_v2.pt
), but independently from how many layers I try to freeze (I've also tried with 0) I run into this error. Any idea on which could be the issue?For reference, I am using
remora v. 3.1.0
and this is the command:The text was updated successfully, but these errors were encountered: