Closed ishandutta0098 closed 1 year ago
Check out this pull request on
See visual diffs & provide feedback on Jupyter Notebooks.
Powered by ReviewNB
View / edit / reply to this conversation on ReviewNB
soumik12345 commented on 2022-12-21T18:05:33Z ----------------------------------------------------------------
Instead of downloading the dataset from dropbox, can you please keep them as an artifact under your WandB project and fetch it from there.
View / edit / reply to this conversation on ReviewNB
soumik12345 commented on 2022-12-21T18:05:34Z ----------------------------------------------------------------
Line #1. !pip install -q --upgrade torch_snippets
Please keep all the pip installations under a single cell, separate from the cells containing python code.
View / edit / reply to this conversation on ReviewNB
soumik12345 commented on 2022-12-21T18:05:35Z ----------------------------------------------------------------
Line #1. !mkdir cropped_faces
Please keep all the shell commands and pip installations under a single cell, separate from the cells containing python code.
👍
View / edit / reply to this conversation on ReviewNB
soumik12345 commented on 2022-12-21T18:05:36Z ----------------------------------------------------------------
Line #3. for i in range(len(images)):
Please add a tqdm.tqdm
here.
My newest trick is from tqdm.auto import tqdm
View / edit / reply to this conversation on ReviewNB
soumik12345 commented on 2022-12-21T18:05:37Z ----------------------------------------------------------------
wandb.log
cannot be called before wand.init
View / edit / reply to this conversation on ReviewNB
soumik12345 commented on 2022-12-21T18:05:37Z ----------------------------------------------------------------
The batch_size
should be added to WANDB_CONFIG
.
View / edit / reply to this conversation on ReviewNB
soumik12345 commented on 2022-12-21T18:05:38Z ----------------------------------------------------------------
Line #1. !pip install torch_summary
Please keep all the pip installations under a single cell, separate from the cells containing python code.
View / edit / reply to this conversation on ReviewNB
soumik12345 commented on 2022-12-21T18:05:39Z ----------------------------------------------------------------
Line #10. return error_real + error_fake
Why not log the real and the fake losses individually?
return discriminator_error_real, discriminator_error_fake
View / edit / reply to this conversation on ReviewNB
soumik12345 commented on 2022-12-21T18:05:40Z ----------------------------------------------------------------
The learning rates should be added to WANDB_CONFIG
as well,
View / edit / reply to this conversation on ReviewNB
soumik12345 commented on 2022-12-21T18:05:41Z ----------------------------------------------------------------
Line #5. for epoch in tqdm(range(25), total = 25):
The number of epochs should be added to WANDB_CONFIG
as well.
View / edit / reply to this conversation on ReviewNB
soumik12345 commented on 2022-12-21T18:05:42Z ----------------------------------------------------------------
Plotting images as a grid doesn't add any value to the example. You can remove this cell.
View / edit / reply to this conversation on ReviewNB
soumik12345 commented on 2022-12-21T18:05:43Z ----------------------------------------------------------------
Plotting images as a grid doesn't add any value to the example. You can remove this cell.
View / edit / reply to this conversation on ReviewNB
soumik12345 commented on 2022-12-21T18:05:44Z ----------------------------------------------------------------
Plotting images as a grid doesn't add any value to the example. You can remove this cell.
View / edit / reply to this conversation on ReviewNB
soumik12345 commented on 2022-12-21T18:05:45Z ----------------------------------------------------------------
Logging a grid of image to your workspace under a table doesn't showcase the recommended usage. Instead, please log a list of wandb.Image
, where each image is a single image generated by the generator. The result in that case would be an interactive media panel simiar to this.
Hi, @soumik12345 I am checking them and will update you soon.
Hi @soumik12345, as discussed I am closing this PR until further confirmation on the type of examples that can be contributed to the repository.
About the PR
Added an example notebook to Generate faces using DCGANs in PyTorch. A report for this is also available on the dashboard.
Colab Link Wandb Dashboard
Requesting Review
@soumik12345 @tcapelle