Jun

After diving into the details, we found that the values output by selective_scan_fn are very large in magnitude, even though the inputs to the function have normal magnitudes. The output...

hi all, Finally, we found that disabling AMP and reducing the learning rate can solve this issue.

Hi @cloneofsimo , Thanks for your guidance very much. When you get the chance, would it be possible for you to check the following dataset class? I use a csv...

hi @hu-po , Thanks for sharing the fine-tuning code very much. Would it be possible for you to give guidance on how to prepare the customized dataset (e.g., data format...

Hi @weidai00 @tifat58 @qingshi1 @innat @emmanuel-contreras We provide a step-by-step tutorial on fine-tuning SAM on 2D and 3D medical image datasets. It requires less than 10G GPU memory. Hope that...

@WusterHappy We provide a step-by-step tutorial on fine-tuning SAM on 2D and 3D medical image datasets. It requires less than 10G GPU memory. Hope that it could be useful. https://github.com/bowang-lab/MedSAM#model-training-video-tutorial

@lgn211 @Xushuolin We provide a step-by-step tutorial on fine-tuning SAM on 2D and 3D medical image datasets. It requires less than 10G GPU memory. Hope that it could be useful....

@kirilllzaitsev , @exhyy , @trieunus , the same issue here. have you solved it?

Hi @exhyy , Thanks for your reply very much. I change the optimizer to `optimizer_state = torch.load(input_optimizer_file, map_location=map_location)` but still have the OOM issue. Would it be possible for you...

I also tried to set `split_batches=True` and adjust the batch size to 64 (two nodes and each node has 4 gpus). The training process iterates few steps and runs into...