[Bug]: Colab error while training hypernetwork

[andrewontheb]

Is there an existing issue for this?

• I have searched the existing issues and checked the recent builds/commits

What happened?

I've been using fast-stable-diffusion notebook in google Colab. And training process is stopped at the step when an image should be saved to textual_inversion log. I’ve already tried to change checkpoint models, training settings and materials

Traceback (most recent call last):
File "/sd/stable-diffusion-webui/modules/hypernetworks/hypernetwork.py", line 701, in train_hypernetwork
p.sampler_name = sd_samplers.samplers[preview_sampler_index].name
TypeError: list indices must be integers or slices, not str

Steps to reproduce the problem

1. Go to ....
2. Press ....
3. ...
   4...

What should have happened?

Save a training progress image to log directory every N steps

Sysinfo

What browsers do you use to access the UI ?

Apple Safari

Console logs

Startup time: 38.3s (import torch: 15.0s, import gradio: 2.0s, setup paths: 4.0s, initialize shared: 0.5s, other imports: 3.3s, setup codeformer: 0.6s, setup gfpgan: 0.1s, list SD models: 0.2s, load scripts: 4.2s, create ui: 3.2s, gradio launch: 4.9s, add APIs: 0.2s).
9aba26abdfcd46073e0a1d42027a3a3bcc969f562d58a03637bf0a0ded6586c9
Loading weights [9aba26abdf] from /content/gdrive/MyDrive/sd/stable-diffusion-webui/models/Stable-diffusion/deliberate_v2.safetensors
Creating model from config: /content/gdrive/MyDrive/sd/stable-diffusion-webui/configs/v1-inference.yaml
Downloading (…)olve/main/vocab.json100% 961k/961k [00:00<00:00, 15.4MB/s]
Downloading (…)olve/main/merges.txt100% 525k/525k [00:00<00:00, 7.23MB/s]
Downloading (…)cial_tokens_map.json100% 389/389 [00:00<00:00, 1.92MB/s]
Downloading (…)okenizer_config.json100% 905/905 [00:00<00:00, 6.89MB/s]
Downloading (…)lve/main/config.json100% 4.52k/4.52k [00:00<00:00, 27.0MB/s]
Applying attention optimization: xformers... done.
Model loaded in 47.5s (calculate hash: 24.9s, load weights from disk: 1.3s, create model: 1.8s, apply weights to model: 15.5s, calculate empty prompt: 3.9s).
Preprocessing [Image 10/11]: 100% 11/11 [00:04<00:00,  2.63it/s]
Applying attention optimization: xformers... done.
*** Error completing request
*** Arguments: ('task(egtmx3f2ofk35e0)', '', '2e-5:800, 8e-6:1600, 5e-6', 1, 1, '/content/gdrive/MyDrive/sd/Processed', 'textual_inversion', 512, 512, False, 10000, 'disabled', '0.1', False, 0, 'once', False, 100, 100, 'custom.txt', True, 'a photo of a young woman ', '', 20, 'Euler a', 7, -1, 512, 512) {}
    Traceback (most recent call last):
      File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/call_queue.py", line 57, in f
        res = list(func(*args, **kwargs))
      File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/call_queue.py", line 36, in f
        res = func(*args, **kwargs)
      File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/hypernetworks/ui.py", line 25, in train_hypernetwork
        hypernetwork, filename = modules.hypernetworks.hypernetwork.train_hypernetwork(*args)
      File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/hypernetworks/hypernetwork.py", line 477, in train_hypernetwork
        textual_inversion.validate_train_inputs(hypernetwork_name, learn_rate, batch_size, gradient_step, data_root, template_file, template_filename, steps, save_hypernetwork_every, create_image_every, log_directory, name="hypernetwork")
      File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/textual_inversion/textual_inversion.py", line 366, in validate_train_inputs
        assert model_name, f"{name} not selected"
    AssertionError: hypernetwork not selected
Calculating sha256 for /content/gdrive/MyDrive/sd/stable-diffusion-webui/models/hypernetworks/Joki .pt: ee9e9198f019b24aad90f2d54ee7b78f2e9f557763f594e1bb9cd1521fbf6dc7
Training at rate of 2e-05 until step 800
Preparing dataset...
100% 11/11 [00:02<00:00,  4.34it/s]
Training hypernetwork [Epoch 9: 1/11]loss: 0.0650765:   1% 99/10000 [01:14<1:52:48,  1.46it/s] *** Exception in training hypernetwork
    Traceback (most recent call last):
      File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/hypernetworks/hypernetwork.py", line 701, in train_hypernetwork
        p.sampler_name = sd_samplers.samplers[preview_sampler_index].name
    TypeError: list indices must be integers or slices, not str

Additional information

No response

[aria1th]

@andrewontheb Thank you for the bug report! the bug seems to be fixed with patch .

Since we can't find the commit 9aba26abdfcd46073e0a1d42027a3a3bcc969f562d58a03637bf0a0ded6586c9, unfortunately I can't offer you the separate branch that you can test.

You can test the fixed option with the branch
https://github.com/AUTOMATIC1111/stable-diffusion-webui/tree/fix-preview-while-generation

by performing git switch fix-preview-while-generation in the repository path, maybe.

[catboxanon]

9aba26abdfcd46073e0a1d42027a3a3bcc969f562d58a03637bf0a0ded6586c9 is a model hash, not commit hash. Since no sysinfo is provided it's probably safe to assume they're running v1.6.0.

[illuculent]

Claims this is fixed and closed but I still get this issue. I have the latest update to auto1111. AUTOMATIC1111 deleted the fix-preview-while-generation branch 2 weeks ago. It appears the fix was merged and then deleted so I cannot so what the fix was.

[B-Hard]

Same here. Before an image gets saved in log directory it aborts the embedding or training of hypernetwork:

** Error training embedding Traceback (most recent call last): File "C:\AI\stable-diffusion-webui\modules\textual_inversion\textual_inversion.py", line 593, in train_embedding p.sampler_name = sd_samplers.samplers[preview_sampler_index].name TypeError: list indices must be integers or slices, not str

[B-Hard]

This fix solved the issue for me. de5bb4c but results are really bad and not useable or nearly recognizable as something close to references.

BUT if using Read parameters - text2img prompt checkbox. It still aborts the whole operation when an image should be generated... :(