-
Notifications
You must be signed in to change notification settings - Fork 6
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
AE latent space training code #22
Comments
I've managed to get the │ ╭──────────────────────────── locals ────────────────────────────╮ │
│ │ entity = 'gladia' │ │
│ │ path = 'gladia/rae/1fdme9ar' │ │
│ │ project = 'rae' │ │
│ │ run_id = '1fdme9ar' │ │
│ │ self = <wandb.apis.public.api.Api object at 0x7ffac1a7b0d0> │ │
│ ╰────────────────────────────────────────────────────────────────╯ │
CommError: Could not find run <Run gladia/rae/1fdme9ar (not found)> If this is the case, it would be helpful to understand how to best recreate these four 2D MNIST AE spaces in order to faithfully replicate these results. update: I found the old #!/bin/bash
# Reconstruction
python src/rae/run.py -m \
core.tags='[classification, absolute, fig:ae-rotations, small_cnn]' \
'nn/data/datasets=vision/mnist' \
'train.seed_index=0,1,2,3,4,5' \
nn/module=classifier \
nn/module/model=cnn \
train=classification \
nn.module.model.latent_dim=2 \
nn.data.anchors_num=500 \
"nn.module.model.hidden_dims=[2, 3, 4, 8]" \
"nn.module.optimizer.lr=5e-4" \
train.trainer.max_epochs=40 After updating the omegaconf.errors.InterpolationKeyError: Interpolation key 'nn.data.datasets.val_fixed_sample_idxs' not found So pleased to find the AE training code was present all along, but currently stuck on updating the configurations to get it to launch successfully. |
Hi @dribnet, thank you for your interest, and I apologize for the delay! To reproduce the experiments, all the files are uploaded to Google Drive using DVC, so you shouldn't need to retrain or run anything to reproduce any of them. You only need to have the In this case, the notebook assumes to have the files inside the "checkpoints" folder (indexed by the If you want to download all the files for the project, you can run Let us know if everything works fine (the Google Drive linked to DVC should be correctly shared in read-only mode). |
I am trying to reproduce figure 1 from the paper:
I've found the code in
fig:latent-rotation/visualize.ipynb
and am attempting to get it to work. IIUC - it appears to assume that pre-baked model checkpoints are downloaded with thedown.sh
helper script.If I've got this right - could you provide a pointer to the AE training of these MNIST checkpoints themselves? My interest is in recreating your results including model training so that I can do follow up experiments which vary the upstream parameters on the latent spaces.
The text was updated successfully, but these errors were encountered: