errors when processing ClipLoader

#14
by artgg - opened

Not sure what s going on, I experienced some size mismatch issue like this below, any suggestion on how to fix it is appreciated. Thanks.

CLIPLoader
Error(s) in loading state_dict for Llama2:
size mismatch for model.embed_tokens.weight: copying a param with shape torch.Size([151936, 2560]) from checkpoint, the shape in current model is torch.Size([128256, 4096]).
size mismatch for model.layers.0.self_attn.q_proj.weight: copying a param with shape torch.Size([4096, 2560]) from checkpoint, the shape in current model is torch.Size([4096, 4096]).
size mismatch for model.layers.0.self_attn.k_proj.weight: copying a param with shape torch.Size([1024, 2560]) from checkpoint, the shape in current model is torch.Size([1024, 4096]).
...
size mismatch for model.layers.31.input_layernorm.weight: copying a param with shape torch.Size([2560]) from checkpoint, the shape in current model is torch.Size([4096]).
size mismatch for model.layers.31.post_attention_layernorm.weight: copying a param with shape torch.Size([2560]) from checkpoint, the shape in current model is torch.Size([4096]).
size mismatch for model.norm.weight: copying a param with shape torch.Size([2560]) from checkpoint, the shape in current model is torch.Size([4096]).

Update comfyui

Update comfyui

thanks

Thanks guys!

  1. La solución práctica (recomendada): usar el text encoder oficial de Z Image

Los propios devs de ComfyUI dicen que para Z Image Turbo uses ESTOS archivos, no el GGUF:
comfyanonymous.github.io

Text encoder: qwen_3_4b.safetensors

Diffusion model: z_image_turbo_bf16.safetensors

VAE: ae.safetensors (el Flux 1 VAE)

Con eso evitas todos los hacks raros de GGUF para el texto y usas lo que Comfy ya soporta nativamente.

2.1. Links directos para descargar

Copia y pega en el navegador:

Text encoder (qwen_3_4b.safetensors)

https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/text_encoders/qwen_3_4b.safetensors?download=true

Sin URL
+1

Modelo de difusión (z_image_turbo_bf16.safetensors)

https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/diffusion_models/z_image_turbo_bf16.safetensors?download=true

Sin URL
+1

(El VAE ae.safetensors ya lo tienes o lo puedes usar el de Flux que ya estás usando.)

I have the same problem and update didn't help. qwen_3_4b.safetensors from official link

I also same problem. Eventhough I updated ComfyUI and nodes, nothing happened.

git clone your comfyui from scratch (try it in an alternative location first) then do your front end requirements as instructure. It should work.

I also same problem. Eventhough I updated ComfyUI and nodes, nothing happened.

git clone your comfyui from scratch (try it in an alternative location first) then do your front end requirements as instructure. It should work.

Update via "git pull" inside ComfyUI folder, and "pip install -r requirements.txt", updating via manager doesn't always work

Update via "git pull" inside ComfyUI folder, and "pip install -r requirements.txt", updating via manager doesn't always work

I did update with "update_comfyui.bat" from "update" folder and apply "pip install -r requirements.txt" step inside ComfyUI folder. It worked! Thanks

Sign up or log in to comment