-
Notifications
You must be signed in to change notification settings - Fork 65
Size mismatch #69
Comments
2.1 must have changed something again. I'll check out the SD thread |
Alright |
https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#stable-diffusion-20 It says it should be identical. Make sure the 2.1 model you're using is 768 rather than 512 as both versions are available |
Can you send the full log? Need to check if it's applying the json/yaml properly |
If it is, it could be Stability-AI/stablediffusion@c12d960 which is an inherent flaw of 2.1. I might need to change the code to fix that. |
I tried it again after getting error so it says it already exists, I'm trying once again from fresh
|
It's definitely applying the config. I might need to do some fixes. Is the model public? |
Made it public, here https://huggingface.co/Z3R069/sick/resolve/main/Sickers.ckpt |
Replicated issue. Trying to fix it |
Ugh mine's shutting down with ^C everytime |
Google is limiting your access. Low VRAM assigned |
I used different google account with vpn but still shutting down |
I'm getting the same issue now. Google might just be limiting everyone due to high demand |
Oh |
Its working now btw but Im still getting the error |
It'll take some time before I can get around to fixing it. My workload is a bit high at the moment with my ChatGPT repos. |
It's alright, take your time 😅 |
Ello? |
The error seems to be the same as AUTOMATIC1111/stable-diffusion-webui#5745 I still haven't found a solution yet because everything looks to be working but the size being applied is wrong |
oh |
Could it be an issue with the dreambooth model? I'll try to remake the model then |
Hi! I got the same issue here. Did one of you found any solution? |
nope, not working still |
Is it only with 2.1 models? |
Yeah so far |
They might have different number of weights to 2.0 models. I'm not sure what config they use. |
I'll make a bug report on @AUTOMATIC1111 's repo |
Looking at this line: size mismatch for model.diffusion_model.output_blocks.11.1.transformer_blocks.0.attn2.to_v.weight: copying a param with shape torch.Size([320, 768]) from checkpoint, the shape in current model is torch.Size([320, 1024]). It seems that you are not using a 768 model |
That's weird, Im pretty sure I used a 768 model |
I'll try with the 512 option then and see if it works |
@RayZ3R0 It is not a 2.1 model |
It worked for me without any config. (Or maybe it is compatible with 1.5) |
Looking at your config: https://huggingface.co/Z3R069/sick/blob/main/unet/config.json {
"_class_name": "UNet2DConditionModel",
"_diffusers_version": "0.9.0.dev0",
"_name_or_path": "/content/stable-diffusion-v1-5",
"act_fn": "silu",
"attention_head_dim": 8,
"block_out_channels": [
320,
640,
1280,
1280
],
"center_input_sample": false,
"cross_attention_dim": 768,
"down_block_types": [
"CrossAttnDownBlock2D",
"CrossAttnDownBlock2D",
"CrossAttnDownBlock2D",
"DownBlock2D"
],
"downsample_padding": 1,
"dual_cross_attention": false,
"flip_sin_to_cos": true,
"freq_shift": 0,
"in_channels": 4,
"layers_per_block": 2,
"mid_block_scale_factor": 1,
"norm_eps": 1e-05,
"norm_num_groups": 32,
"num_class_embeds": null,
"only_cross_attention": false,
"out_channels": 4,
"sample_size": 64,
"up_block_types": [
"UpBlock2D",
"CrossAttnUpBlock2D",
"CrossAttnUpBlock2D",
"CrossAttnUpBlock2D"
],
"use_linear_projection": false
} It is 1.5 based |
that's really weird, I definitely selected the 2.1 768px model in the dreambooth colab. But thanks! |
Seems to be a bug with @TheLastBen. You should open up an issue there |
Alright, thanks |
I'm getting the error running locally |
Are you running the notebook locally or https://github.com/AUTOMATIC1111/stable-diffusion-webui? |
Depending on the model you're using, you need to choose the right inference configuration. |
Hello, I'm getting a really long error after starting the cell. I'm using stable diffusion v2.1 and I have selected SD V2-768 in the dropdown menu
The text was updated successfully, but these errors were encountered: