Skip to content

Unable to finetune #97

@clararodrigo

Description

@clararodrigo

Hello, I am trying to further finetune MedSAM2 as I have quite a niche imaging application. I have set the model to training mode by doing:
predictor.train(True) predictor.sam_mask_decoder.train()
However, the loss never goes down and when I print out the gradient of all weights in the model, all are None. Of course, I am doing:
optimiser.zero_grad() loss.backward() optimiser.step()

I would greatly appreciate any help to make my model learn, thanks in advance.

Clara

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions