Error loading local model: _snapshot_download is not supported for stable-diffusion-v1-5/stable-diffusion-v1-5

[ClockworkCreep]

I was getting a different error but installing diffusers as suggested in this post: #2 (comment) made that error go away. But now I have a new one and I don't see anyone else talking about it in issues.

train_func = run_lora_sd1_5
INFO bucket 0: resolution (512, 512), train_util.py:910
count: 6
INFO mean ar error (without repeats): train_util.py:915
0.08373604184221116
INFO preparing accelerator train_network.py:225
F:\Artificial_Intelligence\ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\accelerate\accelerator.py:439: FutureWarning: torch.cuda.amp.GradScaler(args...) is deprecated. Please use torch.amp.GradScaler('cuda', args...) instead.
self.scaler = torch.cuda.amp.GradScaler(**kwargs)
INFO loading model for process 0/1 train_util.py:4385
accelerator device: cuda
Loading local model: F:\Artificial_Intelligence\ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\models\checkpoints\cyberrealistic_v33.safetensors
_snapshot_download: stable-diffusion-v1-5/stable-diffusion-v1-5
Error loading local model: _snapshot_download is not supported for stable-diffusion-v1-5/stable-diffusion-v1-5
Exception: _snapshot_download is not supported for stable-diffusion-v1-5/stable-diffusion-v1-5

An error message doesn't pop up in ComfyUI, but as far as I can tell, it's just hanging there not doing anything.

[ml1969]

Hello together,

yesterday I installed Comfyui_FL-Trainer and got exactly trhe same error.
No error message in the confyui gui only in the cmd box.
It doesn´t matter, if I use the "Easy Train" oder the regular nodes...
I also tried to update all, but it did not help.

train_func = run_lora_sd1_5
0%| | 0/33 [00:00<?, ?it/s]
6%|▌ | 2/33 [00:00<00:03, 9.15it/s]
9%|▉ | 3/33 [00:00<00:04, 6.59it/s]
12%|█▏ | 4/33 [00:00<00:05, 5.56it/s]
21%|██ | 7/33 [00:00<00:02, 11.03it/s]
33%|███▎ | 11/33 [00:00<00:01, 17.27it/s]
55%|█████▍ | 18/33 [00:01<00:00, 28.80it/s]
73%|███████▎ | 24/33 [00:01<00:00, 36.14it/s]
88%|████████▊ | 29/33 [00:01<00:00, 36.90it/s]
100%|██████████| 33/33 [00:01<00:00, 25.66it/s]
2024-11-17 09:23:51 INFO make buckets train_util.py:859
INFO number of images (including train_util.py:905
repeats) /
各bucketの画像枚数（繰り返
し回数を含む）
INFO bucket 0: resolution (384, 640), train_util.py:910
count: 720
INFO bucket 1: resolution (512, 512), train_util.py:910
count: 90
INFO bucket 2: resolution (576, 448), train_util.py:910
count: 180
INFO mean ar error (without repeats): train_util.py:915
0.06240435155092942
INFO preparing accelerator train_network.py:225
INFO loading model for process 0/1 train_util.py:4385
accelerator device: cuda
Loading local model: F:/Comfyui_models/models/checkpoints/sd15\v1-5-pruned.safetensors
_snapshot_download: stable-diffusion-v1-5/stable-diffusion-v1-5
Error loading local model: _snapshot_download is not supported for stable-diffusion-v1-5/stable-diffusion-v1-5
Exception: _snapshot_download is not supported for stable-diffusion-v1-5/stable-diffusion-v1-5

Thanks for help
regards

Michael

[androideve]

Also hanging at "Exception: _snapshot_download is not supported for stable-diffusion-v1-5/stable-diffusion-v1-5"

[RandoKr]

im also stuck with this?

[altoso99]

Got it, in the function _snapshot_download in the file hook_kohya_ss_utils.py, the repo_id being passed is stable-diffusion-v1-5/stable-diffusion-v1-5 instead of runwayml/stable-diffusion-v1-5
As a temp fix you can change it and try if it works.

[altoso99]

Hi, aparently the current SD 1.5 model download has a new path name:

⚠️ This repository is a mirror of the now deprecated ruwnayml/stable-diffusion-v1-5, this repository or organization are not affiliated in any way with RunwayML.
https://huggingface.co/stable-diffusion-v1-5/stable-diffusion-v1-5

I changed all references I found from ruwnayml/stable-diffusion-v1-5 to stable-diffusion-v1-5/stable-diffusion-v1-5 in the files model_util.py and hook_kohya_ss_utils.py (albeit adding it should work too for retrocompatibility for previously downloaded models), the model seems to load, but now I'm getting another error:
Exception: 'TokenizersWrapper' object has no attribute 'bos_token_id'

Might be related or not, I will take a look at it. Hope this info helps to fix this issue.

[xyzalanix]

This helped me^
I also encountered the error Exception: 'TokenizersWrapper' object has no attribute 'bos_token_id'
and resolved it by downgrading to transformers==4.46

[DJLegends1011]

This helped me^ I also encountered the error Exception: 'TokenizersWrapper' object has no attribute 'bos_token_id' and resolved it by downgrading to transformers==4.46

how does one do this since idk how to downgrade this at all

[MouthlessmanDA]

So I have now done both the editing of the .py files and I have downgraded transformers to 4.46 but I am now getting the
Exception: 'TokenizersWrapper' object has no attribute 'bos_token_id' error too. Is there any new developments?