yu-rp / distribution-shift-iverson

Hi there 👋. This is Runpeng Yu.

---

I am a Ph.D. student majored in Data Science at NUS.

distribution-shift-iverson's Issues

Number of epochs

Hi,

Thanks for the good work.
I am trying to reproduce your results and I have 2 questions.

• How many epochs (or iterations) you used to train the networks?
• Is it possible for you to release your trained models?

Thanks

Issue in Transformation Step Code

Hi,

I am trying to replicate your experiments. I was able to successfully follow the Finetuning of Diffusion Model
step mentioned here. However, when I tried running the code provided for the Transformation step, I get the following error:

Traceback (most recent call last):
  File "scripts/augment.py", line 444, in <module>
    main()
  File "scripts/augment.py", line 355, in main
    print(outstr)
UnboundLocalError: local variable 'outstr' referenced before assignment

When I commented out: print(outstr) and changed the --dataset PACS_Augmentation to PACS, I got the following error:

File "scripts/augment.py", line 444, in <module>
    main()
  File "scripts/augment.py", line 353, in main
    dataset = onedomainPACS_withlabel(root = "../../Data/", test_envs = opt.domain_index, data_augmentation = False, image_size = 256)
NameError: name 'onedomainPACS_withlabel' is not defined

Therefore, I would greatly appreciate your clarification on how to run this step.

Thanks.
Sara A. Al-Emadi

Problem about cuda-out-of-memory

Hi,
I want to reproduce your results via your provided codes. But I was stuck in the fine-tuning section. No matter how I reduce the batch size and input image size, it still says cuda out of memory. I run the codes by your instructions on a nvidia3090 gpu with 24g memory. But it seems that before the program loads the images, all the memory has already been allocated. So although I reduce the batch size to 1 and input size to 64, the cuda out of memory problem is still there. Does the memory be allocated to the models? Do you have any idea on how to solve this problem? I guess 24g is enough to tune a diffusion model.