TIO-IKIM / CellViT

CellViT: Vision Transformers for Precise Cell Segmentation and Classification
https://doi.org/10.1016/j.media.2024.103143
Other
217 stars 33 forks source link

sweep init failed for training #4

Closed randy-tsukemen closed 1 year ago

randy-tsukemen commented 1 year ago

Thank you for sharing great work!

While I tried to reproduce the training part, the train_cellvit.yaml is hard for me to understand the format. Do I need to fill them with the sweep format with max and min like below?

'parameters': 
    {
        'x': {'max': 0.1, 'min': 0.01},
        'y': {'values': [1, 3, 7]},
     }

I also tried to figure out how to use the insert_parameter_tag function since it looks like we need to parameters tag to push to wandb but currently the function is not used in the code.

If possible, could you kindly provide the filled train_cellvit.yaml?

FabianHoerst commented 1 year ago

I removed the function insert_paramezer_tag, as it was not used. Tags have to be given inside the .yaml configuration file. I added a new configuration file train_cellvit_sweep.yaml for wandb sweeps, hopefully explaining the use of sweeping parameters. Unfortunately, we cannot share all configuration files as the work is still under peer review.