stable-diffusion-webui Run run_ webui

Describe the bug After I installed it, it ran successfully, but after a few hours, it ran again: run_webui_mac.sh will report an error, stop at AttributeError: 'NoneType' object has no attribute 'keys'. I tried to delete /Users/yjy/. cache/huggingface, but errors still occur.

To Reproduce Steps to reproduce the behavior: I don't know why there are error

Desktop (please complete the following information):

OS: macOS

$ ./run_webui_mac.sh   
WARNING: overwriting environment variables set in the machine
overwriting variable PYTORCH_ENABLE_MPS_FALLBACK
To make your changes take effect please reactivate your environment
WARNING: overwriting environment variables set in the machine
overwriting variable PYTORCH_ENABLE_MPS_FALLBACK
Already up to date.
Warning: caught exception 'Torch not compiled with CUDA enabled', memory monitor disabled
LatentDiffusion: Running in eps-prediction mode
DiffusionWrapper has 859.52 M params.
making attention of type 'vanilla' with 512 in_channels
Working with z of shape (1, 4, 32, 32) = 4096 dimensions.
making attention of type 'vanilla' with 512 in_channels
Error verifying pickled file from /Users/yjy/.cache/huggingface/hub/models--openai--clip-vit-large-patch14/snapshots/8d052a0f05efbaefbc9e8786ba291cfdf93e5bff/pytorch_model.bin:
Traceback (most recent call last):
  File "/Users/yjy/Documents/Code/AI/stable-diffusion-webui/modules/safe.py", line 76, in load
    check_pt(filename)
  File "/Users/yjy/Documents/Code/AI/stable-diffusion-webui/modules/safe.py", line 60, in check_pt
    unpickler.load()
  File "/Users/yjy/Documents/Code/AI/stable-diffusion-webui/modules/safe.py", line 23, in persistent_load
    return torch.storage._TypedStorage()
AttributeError: module 'torch.storage' has no attribute '_TypedStorage'


The file may be malicious, so the program is not going to read it.
You can skip this check with --disable-safe-unpickle commandline argument.
Traceback (most recent call last):
  File "/Users/yjy/Documents/Code/AI/stable-diffusion-webui/webui.py", line 82, in <module>
    shared.sd_model = modules.sd_models.load_model()
  File "/Users/yjy/Documents/Code/AI/stable-diffusion-webui/modules/sd_models.py", line 174, in load_model
    sd_model = instantiate_from_config(sd_config.model)
  File "/Users/yjy/Documents/Code/AI/stable-diffusion-webui/repositories/stable-diffusion/ldm/util.py", line 85, in instantiate_from_config
    return get_obj_from_str(config["target"])(**config.get("params", dict()))
  File "/Users/yjy/Documents/Code/AI/stable-diffusion-webui/repositories/stable-diffusion/ldm/models/diffusion/ddpm.py", line 461, in __init__
    self.instantiate_cond_stage(cond_stage_config)
  File "/Users/yjy/Documents/Code/AI/stable-diffusion-webui/repositories/stable-diffusion/ldm/models/diffusion/ddpm.py", line 519, in instantiate_cond_stage
    model = instantiate_from_config(config)
  File "/Users/yjy/Documents/Code/AI/stable-diffusion-webui/repositories/stable-diffusion/ldm/util.py", line 85, in instantiate_from_config
    return get_obj_from_str(config["target"])(**config.get("params", dict()))
  File "/Users/yjy/Documents/Code/AI/stable-diffusion-webui/repositories/stable-diffusion/ldm/modules/encoders/modules.py", line 142, in __init__
    self.transformer = CLIPTextModel.from_pretrained(version)
  File "/Users/yjy/Documents/Code/miniforge3/envs/web-ui/lib/python3.10/site-packages/transformers/modeling_utils.py", line 2138, in from_pretrained
    loaded_state_dict_keys = [k for k in state_dict.keys()]
AttributeError: 'NoneType' object has no attribute 'keys'

What is the solution to this problem?

Oct 09 '22 18:10 JavinYang

Same.

I managed to solve this problem by uninstalling the nightly versions of Torch and Torchvision and installing the stable ones. if you do that, the webui will load, but as you attempt to generate an image, you'll get back to the dear old error:

RuntimeError: view size is not compatible with input tensor's size and stride (at least one dimension spans across two contiguous subspaces). Use .reshape(...) instead.

That error can be fixed only with the nightly versions of Torch and Torchvision (installed by the macOS script). So you are back to square one.

My solution, for now, is to keep the nightly versions and revert to a previous commit:

git reset --hard e00b4df7c6f0a13941d6f6ea425eebdaa2bc9318

This commit is safe, at least with my configuration. Once you do this, you can launch the webui as usual. Just don't issue a git pull command.

Oct 09 '22 18:10 system1system2

Any better solution for this, maybe a specific torchvision & torch nightly version combo?

Oct 09 '22 19:10 StableInquest

Just ran into this also after doing a lot of manual installations to get to this point.

Oct 09 '22 23:10 realsama

Same here. It's working yesterday though, I ran the setup script again but still the same.

Edit: Temporary solved by rolling back to a previous nightly commit of torch, as mentioned here, Since the run_webui_mac.sh script will update torch automatically upon launch, I just commented out the git pull command in the script, and it worked just as before.

Oct 10 '22 00:10 Linus0080

This problem has been fixed. It should be safe to uncomment git pull in the script for anyone who did so.

Oct 10 '22 05:10 brkirch

just tried to start the webui, but that brings up this error (did run the setup script too same outcome)

thomas@MacBook-Pro-von-Thomas stable-diffusion-webui % ./run_webui_mac.sh 
To make your changes take effect please reactivate your environment
WARNING: overwriting environment variables set in the machine
overwriting variable PYTORCH_ENABLE_MPS_FALLBACK
Already up to date.
Traceback (most recent call last):
  File "/Users/thomas/SD_Test/stable-diffusion-webui/webui.py", line 8, in <module>
    from fastapi.middleware.gzip import GZipMiddleware
ModuleNotFoundError: No module named 'fastapi'

Oct 10 '22 06:10 NeoTrace82

did that now I have the same error as StableInquest

user@MacBook-Pro-von-user stable-diffusion-webui % ./run_webui_mac.sh
To make your changes take effect please reactivate your environment
WARNING: overwriting environment variables set in the machine
overwriting variable PYTORCH_ENABLE_MPS_FALLBACK
Already up to date.
Warning: caught exception 'Torch not compiled with CUDA enabled', memory monitor disabled
LatentDiffusion: Running in eps-prediction mode
DiffusionWrapper has 859.52 M params.
making attention of type 'vanilla' with 512 in_channels
Working with z of shape (1, 4, 32, 32) = 4096 dimensions.
making attention of type 'vanilla' with 512 in_channels
Loading weights [e3b0c442] from /Users/user/SD_Test/stable-diffusion-webui/models/Stable-diffusion/sd-v1-4.ckpt
Error verifying pickled file from /Users/user/SD_Test/stable-diffusion-webui/models/Stable-diffusion/sd-v1-4.ckpt:
Traceback (most recent call last):
  File "/Users/user/SD_Test/stable-diffusion-webui/modules/safe.py", line 61, in check_pt
    with zipfile.ZipFile(filename) as z:
  File "/Users/user/miniconda/envs/web-ui/lib/python3.10/zipfile.py", line 1267, in __init__
    self._RealGetContents()
  File "/Users/user/miniconda/envs/web-ui/lib/python3.10/zipfile.py", line 1334, in _RealGetContents
    raise BadZipFile("File is not a zip file")
zipfile.BadZipFile: File is not a zip file

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/Users/user/SD_Test/stable-diffusion-webui/modules/safe.py", line 80, in load
    check_pt(filename)
  File "/Users/user/SD_Test/stable-diffusion-webui/modules/safe.py", line 72, in check_pt
    unpickler.load()
EOFError: Ran out of input


The file may be malicious, so the program is not going to read it.
You can skip this check with --disable-safe-unpickle commandline argument.
Traceback (most recent call last):
  File "/Users/user/SD_Test/stable-diffusion-webui/webui.py", line 82, in <module>
    shared.sd_model = modules.sd_models.load_model()
  File "/Users/user/SD_Test/stable-diffusion-webui/modules/sd_models.py", line 175, in load_model
    load_model_weights(sd_model, checkpoint_info)
  File "/Users/user/SD_Test/stable-diffusion-webui/modules/sd_models.py", line 138, in load_model_weights
    if "global_step" in pl_sd:
TypeError: argument of type 'NoneType' is not iterable
user@MacBook-Pro-von-user stable-diffusion-webui %

Oct 10 '22 07:10 NeoTrace82

I got by this just adding the --disable-safe-unpickle commandline argument at the end of the line below (in file ./run_webui_mac.sh):

python webui.py --precision full --no-half --opt-split-attention-v1 --use-cpu GFPGAN CodeFormer --disable-safe-unpickle

Oct 10 '22 08:10 StableInquest

I am on M1 mac and now the only error I'm seeing is related to textual inversion training:

stable-diffusion-webui/repositories/stable-diffusion/ldm/models/diffusion/ddpm.py", line 1030, in p_losses logvar_t = self.logvar[t].to(self.device) RuntimeError: indices should be either on cpu or on the same device as the indexed tensor (cpu)

Oct 10 '22 08:10 StableInquest

python webui.py --precision full --no-half --opt-split-attention-v1 --use-cpu GFPGAN CodeFormer --disable-safe-unpickle

tried that but getting another error, so sad was working flawless the past days

user@MacBook-Pro-von-user stable-diffusion-webui % ./run_webui_mac.sh
To make your changes take effect please reactivate your environment
WARNING: overwriting environment variables set in the machine
overwriting variable PYTORCH_ENABLE_MPS_FALLBACK
Already up to date.
Warning: caught exception 'Torch not compiled with CUDA enabled', memory monitor disabled
LatentDiffusion: Running in eps-prediction mode
DiffusionWrapper has 859.52 M params.
making attention of type 'vanilla' with 512 in_channels
Working with z of shape (1, 4, 32, 32) = 4096 dimensions.
making attention of type 'vanilla' with 512 in_channels
Loading weights [e3b0c442] from /Users/user/SD_Test/stable-diffusion-webui/models/Stable-diffusion/sd-v1-4.ckpt
Traceback (most recent call last):
  File "/Users/user/SD_Test/stable-diffusion-webui/webui.py", line 82, in <module>
    shared.sd_model = modules.sd_models.load_model()
  File "/Users/user/SD_Test/stable-diffusion-webui/modules/sd_models.py", line 175, in load_model
    load_model_weights(sd_model, checkpoint_info)
  File "/Users/user/SD_Test/stable-diffusion-webui/modules/sd_models.py", line 137, in load_model_weights
    pl_sd = torch.load(checkpoint_file, map_location="cpu")
  File "/Users/user/SD_Test/stable-diffusion-webui/modules/safe.py", line 89, in load
    return unsafe_torch_load(filename, *args, **kwargs)
  File "/Users/user/miniconda/envs/web-ui/lib/python3.10/site-packages/torch/serialization.py", line 764, in load
    return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
  File "/Users/user/miniconda/envs/web-ui/lib/python3.10/site-packages/torch/serialization.py", line 971, in _legacy_load
    magic_number = pickle_module.load(f, **pickle_load_args)
EOFError: Ran out of input
user@MacBook-Pro-von-user stable-diffusion-webui %

Oct 10 '22 08:10 NeoTrace82

I'm having similar problems when running ./run_webui_mac.sh

My processor is the M1 Pro

WARNING: overwriting environment variables set in the machine
overwriting variable PYTORCH_ENABLE_MPS_FALLBACK
Already up to date.
Traceback (most recent call last):
  File "/Users/luca/repositories/stable-diffusion-webui/webui.py", line 8, in <module>
    from fastapi.middleware.gzip import GZipMiddleware
ModuleNotFoundError: No module named 'fastapi'```

Oct 10 '22 11:10 lucagonzalez

I'm having similar problems when running ./run_webui_mac.sh

My processor is the M1 Pro

WARNING: overwriting environment variables set in the machine
overwriting variable PYTORCH_ENABLE_MPS_FALLBACK
Already up to date.
Traceback (most recent call last):
  File "/Users/luca/repositories/stable-diffusion-webui/webui.py", line 8, in <module>
    from fastapi.middleware.gzip import GZipMiddleware
ModuleNotFoundError: No module named 'fastapi'```

that can be solved by running as suggested by @brkirch but then I had a different error

bash -l -c "conda activate web-ui; pip install jsonmerge einops clean-fid resize_right torchdiffeq lark gradio fastapi omegaconf piexif fonts font-roboto pytorch_lightning transformers kornia realesrgan scunet timm"

Oct 10 '22 11:10 NeoTrace82

fastapi

https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/1990

Oct 10 '22 11:10 JavinYang

@NeoTrace82 I'm having the same error messages as you

Oct 10 '22 12:10 lucagonzalez

don't ask me how I fixed it, did try a lot of things out the things I remember:

I did purge python manually and reinstalled python it installed the requirements manually again did run the setup downloaded the models manually and the GFPGAN for Face Restoration

right now its working, can't tell what fixed it but I don't wanna play around anymore as its running

if I remember any other step I did I will report it immediately

Oct 10 '22 12:10 NeoTrace82

I've fixed a few things in the setup script and submitted a PR: dylancl/stable-diffusion-webui-mps#2.

Oct 10 '22 13:10 brkirch

Hopefully this gets fixed because I'm experiencing the same issues as the rest of you. Was running fine two days ago.

Oct 10 '22 14:10 Karsten385

Somehow the "Torch not compiled with CUDA enabled" error only presents if a specific sampling method is used. For example on my M1 machine it only happens when using DDIM/PLMS.

Oct 10 '22 14:10 JokerQyou

Fresh installed again, latest pull.. most everything is working but for training. See error:

stable-diffusion-webui/repositories/stable-diffusion/ldm/models/diffusion/ddpm.py", line 1030, in p_losses logvar_t = self.logvar[t].to(self.device) RuntimeError: indices should be either on cpu or on the same device as the indexed tensor (cpu)

Anyone have success in fixing this?

Oct 11 '22 01:10 StableInquest

stable-diffusion-webui/repositories/stable-diffusion/ldm/models/diffusion/ddpm.py", line 1030, in p_losses logvar_t = self.logvar[t].to(self.device) RuntimeError: indices should be either on cpu or on the same device as the indexed tensor (cpu)

@StableInquest Starting with Traceback (most recent call last):, please provide everything that printed. This error occurred inside of Static Diffusion instead of webui and that makes it potentially much harder to troubleshoot without the full traceback.

Oct 11 '22 05:10 brkirch

Traceback (most recent call last):
  File "/Users/user/stable-diffusion-webui/modules/ui.py", line 176, in f
    res = list(func(*args, **kwargs))
  File "/Users/user/stable-diffusion-webui/webui.py", line 68, in f
    res = func(*args, **kwargs)
  File "/Users/user/stable-diffusion-webui/modules/textual_inversion/ui.py", line 29, in train_embedding
    embedding, filename = modules.textual_inversion.textual_inversion.train_embedding(*args)
  File "/Users/user/stable-diffusion-webui/modules/textual_inversion/textual_inversion.py", line 257, in train_embedding
    loss = shared.sd_model(x.unsqueeze(0), c)[0]
  File "/Users/user/miniconda/envs/web-ui/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1190, in _call_impl
    return forward_call(*input, **kwargs)
  File "/Users/user/stable-diffusion-webui/repositories/stable-diffusion/ldm/models/diffusion/ddpm.py", line 879, in forward
    return self.p_losses(x, c, t, *args, **kwargs)
  File "/Users/user/stable-diffusion-webui/repositories/stable-diffusion/ldm/models/diffusion/ddpm.py", line 1030, in p_losses
    logvar_t = self.logvar[t].to(self.device)
RuntimeError: indices should be either on cpu or on the same device as the indexed tensor (cpu)

Oct 11 '22 05:10 StableInquest

After latest pull, getting this:

Traceback (most recent call last):
  File "/Users/user/stable-diffusion-webui/modules/ui.py", line 176, in f
    res = list(func(*args, **kwargs))
  File "/Users/user/stable-diffusion-webui/webui.py", line 68, in f
    res = func(*args, **kwargs)
  File "/Users/user/stable-diffusion-webui/modules/textual_inversion/ui.py", line 29, in train_embedding
    embedding, filename = modules.textual_inversion.textual_inversion.train_embedding(*args)
  File "/Users/user/stable-diffusion-webui/modules/textual_inversion/textual_inversion.py", line 276, in train_embedding
    embedding.save(filename)
  File "/Users/user/stable-diffusion-webui/modules/textual_inversion/textual_inversion.py", line 34, in save
    torch.save(embedding_data, filename)
  File "/Users/user/miniconda/envs/web-ui/lib/python3.10/site-packages/torch/serialization.py", line 379, in save
    _save(obj, opened_zipfile, pickle_module, pickle_protocol)
  File "/Users/user/miniconda/envs/web-ui/lib/python3.10/site-packages/torch/serialization.py", line 589, in _save
    pickler.dump(obj)
  File "/Users/user/miniconda/envs/web-ui/lib/python3.10/site-packages/torch/_tensor.py", line 177, in __reduce_ex__
    return self._reduce_ex_internal(proto)
  File "/Users/user/miniconda/envs/web-ui/lib/python3.10/site-packages/torch/_tensor.py", line 223, in _reduce_ex_internal
    return (torch._utils._rebuild_device_tensor_from_numpy, (self.cpu().numpy(),
RuntimeError: Can't call numpy() on Tensor that requires grad. Use tensor.detach().numpy() instead.

Oct 11 '22 07:10 StableInquest

pip3 install torch==1.11.0 torchvision==0.13.1 torchaudio --extra-index-url https://download.pytorch.org/whl/cu113 This version combo resolves the issue, but its painfully slow.

Oct 11 '22 08:10 StableInquest

Latest pull and latest versions from requirements.txt seems to resolve this now. I'll update if that isn't the case.

Oct 11 '22 18:10 StableInquest

Although this works it is very slow. 6 seconds per step. M1 MacOS. Any suggestions?

Oct 12 '22 04:10 StableInquest

I'm having similar problems when running ./run_webui_mac.sh My processor is the M1 Pro
WARNING: overwriting environment variables set in the machine
overwriting variable PYTORCH_ENABLE_MPS_FALLBACK
Already up to date.
Traceback (most recent call last):
  File "/Users/luca/repositories/stable-diffusion-webui/webui.py", line 8, in <module>
    from fastapi.middleware.gzip import GZipMiddleware
ModuleNotFoundError: No module named 'fastapi'```
that can be solved by running as suggested by @brkirch but then I had a different error

bash -l -c "conda activate web-ui; pip install jsonmerge einops clean-fid resize_right torchdiffeq lark gradio fastapi omegaconf piexif fonts font-roboto pytorch_lightning transformers kornia realesrgan scunet timm"

It's work for me, I'm just use m1 silicon air

Oct 16 '22 08:10 kaminion

I'm still not having any luck. I've tried deleting miniconda and stable-diffusion-webui multiple time to fresh install everything. This is error I get when attempting to train embedding:

stable-diffusion-webui/modules/textual_inversion/textual_inversion.py", line 185, in write_loss
    with open(os.path.join(log_directory, filename), "a+", newline='') as fout:
FileNotFoundError: [Errno 2] No such file or directory: 'textual_inversion/2022-10-16/test/textual_inversion_loss.csv'

Oct 17 '22 03:10 StableInquest

Ok I was able to get by this by not disabling: "Save an image to log directory every N steps, 0 to disable" or "Save a copy of embedding to log directory every N steps, 0 to disable"

By having this disabled the textual_inversion log directory was never created and it threw the error in my post above.

Oct 17 '22 03:10 StableInquest

I take that back. I've now hit this in the middle of training:

miniconda/envs/web-ui/lib/python3.10/site-packages/torch/_tensor.py", line 223, in _reduce_ex_internal
    return (torch._utils._rebuild_device_tensor_from_numpy, (self.cpu().numpy(),
RuntimeError: Can't call numpy() on Tensor that requires grad. Use tensor.detach().numpy() instead.

This occurs once steps complete and a save is attempted at the end or during one of these saves "Save a copy of embedding to log directory every N steps, 0 to disable"

Oct 17 '22 05:10 StableInquest

Adding this into the conda environment seems to solve it:

pip3 install --upgrade git+https://github.com/pytorch/[email protected]

Only problem is the speed is terrible

Oct 17 '22 06:10 StableInquest

stable-diffusion-webui
stable-diffusion-webui copied to clipboard

Run run_ webui_ Mac.sh error

stable-diffusion-webui stable-diffusion-webui copied to clipboard

Run run_ webui_ Mac.sh error

stable-diffusion-webui
stable-diffusion-webui copied to clipboard