Error looking for files in cache? | Invoke | Page 1

raw lion Nov 15, 2022, 6:37 PM

#

I installed this using the automated script for Windows. Getting this error when trying to launch. Any help would be appreciated, thank you.

`E:\InvokeAI>.venv\Scripts\python scripts\invoke.py

Initializing, be patient...

Initialization file C:\Users\stateanomaly/.invokeai not found. Applying default settings...
NOTE: Redirects are currently not supported in Windows or MacOs.

NOT FOUND: GFPGAN model not found at ./models/gfpgan/GFPGANv1.4.pth

GFPGAN Disabled

NOT FOUND: CodeFormer model not found at ldm/invoke/restoration/codeformer\weights/codeformer.pth

CodeFormer Disabled
ESRGAN Initialized
Using device_type cuda
Current VRAM usage: 0.00G
Loading stable-diffusion-1.5 from ./models/ldm/stable-diffusion-v1/v1-5-pruned-emaonly.ckpt
| 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
** model stable-diffusion-1.5 could not be loaded: [Errno 2] No such file or directory: 'C:\Users\matts/.cache\huggingface\hub\models--openai--clip-vit-large-patch14\refs\main'
Traceback (most recent call last):
File "e:\invokeai\ldm\invoke\model_cache.py", line 75, in get_model
requested_model, width, height, hash = self._load_model(model_name)
File "e:\invokeai\ldm\invoke\model_cache.py", line 220, in _load_model
model = instantiate_from_config(c.model)
File "e:\invokeai\ldm\util.py", line 90, in instantiate_from_config
return get_obj_from_str(config['target'])(
File "e:\invokeai\ldm\models\diffusion\ddpm.py", line 658, in init
self.instantiate_cond_stage(cond_stage_config)
File "e:\invokeai\ldm\models\diffusion\ddpm.py", line 769, in instantiate_cond_stage
model = instantiate_from_config(config)
File "e:\invokeai\ldm\util.py", line 90, in instantiate_from_config
return get_obj_from_str(config['target'])(
File "e:\invokeai\ldm\modules\encoders\modules.py", line 247, in init
self.tokenizer = CLIPTokenizer.from_pretrained(
File "E:\InvokeAI.venv\lib\site-packages\transformers\tokenization_utils_base.py", line 1736, in from_pretrained
resolved_vocab_files[file_id] = cached_file(
File "E:\InvokeAI.venv\lib\site-packages\transformers\utils\hub.py", line 409, in cached_file
resolved_file = hf_hub_download(
File "E:\InvokeAI.venv\lib\site-packages\huggingface_hub\file_download.py", line 1136, in hf_hub_download
with open(ref_path) as f:
FileNotFoundError: [Errno 2] No such file or directory: 'C:\Users\stateanomaly/.cache\huggingface\hub\models--openai--clip-vit-large-patch14\refs\main'

** restoring None
** "None" is not a known model name. Please check your models.yaml file`

wide mauve Nov 15, 2022, 6:39 PM

#

looks like it's missing just one of the files

#

can you try running the preload_models.py script?

#

not sure how to do this on windows sorry

raw lion Nov 15, 2022, 6:40 PM

#

thank for your help. I couldn't ever get that to work, so I downloaded models manually. It seemed like it wasn't accepting the token.

#

All licenses are accepted. Couldn't figure how to diagnose that further.

#

** DOWNLOADING WEIGHTS ** ** An error occurred during downloading v1-5-pruned-emaonly.ckpt: Bad Request *** ERROR DOWNLOADING v1-5-pruned-emaonly.ckpt: ** An error occurred during downloading sd-v1-5-inpainting.ckpt: Bad Request *** ERROR DOWNLOADING sd-v1-5-inpainting.ckpt: ** An error occurred during downloading vae-ft-mse-840000-ema-pruned.ckpt: Bad Request *** ERROR DOWNLOADING vae-ft-mse-840000-ema-pruned.ckpt:

wide mauve Nov 15, 2022, 6:44 PM

#

huh. no idea, but at a guess it might be a firewall/virus scanner blocking things?

#

idk

raw lion Nov 15, 2022, 6:44 PM

#

hmm potentially. I do have bitdefender installed.

#

All firewalls turned off and it's not working. I didn't find vae-ft-mse-840000-ema-pruned.ckpt yet. I'll see if installing that fixes it.

#

Should I be ok with just these model files?

wide mauve Nov 15, 2022, 7:01 PM

#

more than ok

#

all you need is sd-v1-5.ckpt

#

all the others are "optional"

#

uhh sorry

#

v1-5-pruned-emaonly is the official filename afaik

raw lion Nov 15, 2022, 7:02 PM

#

I see, thanks. I'm going through the preload script and trying to figure out what it may be trying to do that I can do manually.

wide mauve Nov 15, 2022, 7:02 PM

#

you do need some CLIP models

#

but those should download automatically

raw lion Nov 15, 2022, 7:03 PM

#

Can I install those manually?

#

Here's the one it's looking for. Where does it go? https://huggingface.co/openai/clip-vit-large-patch14

openai/clip-vit-large-patch14 · Hugging Face

#

The script is trying to find it in cache.

wide mauve Nov 15, 2022, 7:08 PM

#

i'm not sure, sorry

#

one thing that's weird though, in the logs you posted

#

'C:\Users\stateanomaly/.cache\huggingface\hub\models--openai--clip-vit-large-patch14\refs\main

#

vs

#

'C:\Users\matts/.cache\huggingface\hub\models--openai--clip-vit-large-patch14\refs\main'

raw lion Nov 15, 2022, 7:09 PM

#

oops yeah, they actually are the same, I just had foresight to edit the first one

#

to not dox myself, you know

wide mauve Nov 15, 2022, 7:09 PM

#

hah

#

ok

#

it's not possible to tell just by looking at that path which file is being downloaded

#

have you tried making a new token?

raw lion Nov 15, 2022, 7:11 PM

#

so maybe it's back to the huggingface token thing?

wide mauve Nov 15, 2022, 7:12 PM

#

yeah i would try and figure out why the preload_models isn't working

raw lion Nov 15, 2022, 7:12 PM

#

Yeah, I made a new token. Even made a new account and new token.

wide mauve Nov 15, 2022, 7:14 PM

#

if you open up a new terminal window and type huggingface-cli login what happens?

raw lion Nov 15, 2022, 7:14 PM

#

I was just looking into that

#

it's not installed

wide mauve Nov 15, 2022, 7:15 PM

#

😦

raw lion Nov 15, 2022, 7:15 PM

#

can I pip install that?

wide mauve Nov 15, 2022, 7:15 PM

#

you can

#

pip install huggingface_hub

raw lion Nov 15, 2022, 7:18 PM

#

installed it, but command line can't find the command. Do I need to add it to path?

wide mauve Nov 15, 2022, 7:20 PM

#

no idea 😦

#

@worthy dew you're on windows, right?

raw lion Nov 15, 2022, 7:20 PM

#

I think it's because the python virtual env?

#

I need to install it in there?

wide mauve Nov 15, 2022, 7:20 PM

#

maybe..

raw lion Nov 15, 2022, 7:22 PM

#

that wasn't it. I think I need to add it path

wide mauve Nov 15, 2022, 7:22 PM

#

my huggingface-cli binary is at /Users/damian/miniforge3/envs/invokeai/bin/huggingface-cli

#

i don't know if that's helpful though

#

can you do the following

raw lion Nov 15, 2022, 7:23 PM

#

part of what's confusing me is all these different terminals in windows. Should I just stick with miniconda?

#

ok, huggingface-cli is working in miniconda

wide mauve Nov 15, 2022, 7:24 PM

#

great!

#

the question is though if you login if it will be shared elsewhere..

raw lion Nov 15, 2022, 7:27 PM

#

ugh windows

#

It won't login here. When I paste the token in, it's doesn't show any characters. Is that normal?

wide mauve Nov 15, 2022, 7:28 PM

#

possibly

#

you might have to right-click and click 'paste'

#

or maybe just right-click

raw lion Nov 15, 2022, 7:28 PM

#

Maybe I can hardcode my token in preload_models.py

wide mauve Nov 15, 2022, 7:28 PM

#

no don't do that

#

it gets written somewhere to disk

raw lion Nov 15, 2022, 7:29 PM

#

ok

#

Maybe I should try doing all this from WSL

#

I got it!

#

Got the login to work through Powershell. And then used the git store credential command.

#

Must have been a pasting issue in miniconda terminal

#

https://giphy.com/gifs/thisgifishaunted-halloween-woo-happy-opDRL3H2A9iLNuvbOv

Giphy

This GIF Is Haunted

wide mauve Nov 15, 2022, 7:47 PM

#

yaaay ✨

#

computers suck

raw lion Nov 15, 2022, 7:47 PM

#

haha they do. And windows tends to suck even more, lol

wide mauve Nov 15, 2022, 7:48 PM

#

good thing about spending time with AI is that it becomes pretty clear pretty quickly that there's no way this stuff is going to take over the world 😂

#

so now that you're logged in the preload_models script works properly?

raw lion Nov 15, 2022, 7:51 PM

#

It does yes.

#

running it...

#

oh yeah, it works!

#

#

So that was it. Needing to run and login to huggingface manually. Then storing the credentials.

#

Thank you for help, @wide mauve 🙏

wide mauve Nov 15, 2022, 7:59 PM

#

you're welcome!

#Error looking for files in cache?

NOT FOUND: GFPGAN model not found at ./models/gfpgan/GFPGANv1.4.pth

NOT FOUND: CodeFormer model not found at ldm/invoke/restoration/codeformer\weights/codeformer.pth