# Diffusers load custom embedding

**URL:** <https://discuss.huggingface.co/t/diffusers-load-custom-embedding/116160>\
**Category:** 🧨 Diffusers\
**Created:** [November 7, 2024, 9:34am UTC](https://discuss.huggingface.co/t/diffusers-load-custom-embedding/116160 "2024-11-07T09:34:20Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Felox](https://avatars.discourse-cdn.com/v4/letter/f/f6c823/32.png) [@Felox](https://discuss.huggingface.co/u/Felox)\
**Post date:** [November 7, 2024, 9:34am UTC](https://discuss.huggingface.co/t/diffusers-load-custom-embedding/116160/1 "2024-11-07T09:34:20Z")

</div>

I wanted to generate a image using text embedding instead of text as input using clip to tokenizes & embeds.  
The code so far :

```auto
from transformers import AutoTokenizer, CLIPTextModelWithProjection

model = CLIPTextModelWithProjection.from_pretrained("openai/clip-vit-base-patch32")
tokenizer = AutoTokenizer.from_pretrained("openai/clip-vit-base-patch32")

from diffusers import StableDiffusionPipeline, DDIMScheduler
import torch

path ="path_to_my_model.safetensors"

pipe = StableDiffusionPipeline.from_single_file(f"{path}", torch_dtype=torch.float16, use_safetensors=True, variant="fp16")

pipe.to("cuda")
import numpy as np
import torch

prompt = "some random prompt"

text_input = tokenizer(prompt, padding="max_length", max_length=tokenizer.model_max_length, truncation=True, return_tensors="pt")

text_embeddings = model(text_input.input_ids)[0] 

batch_size = len(text_input)

uncond_input = tokenizer(
    [""] * batch_size, padding="max_length", max_length=tokenizer.model_max_length, return_tensors="pt", truncation=True
)

uncond_embeddings = model(uncond_input.input_ids)[0] 

text_embeddings = torch.cat([uncond_embeddings, text_embeddings])

output_image = pipe(prompt_embeds=text_embeddings).images[0]

```

I get so far `ValueError: not enough values to unpack (expected 3, got 2)`

Althought the embedding shape is [3,512], `text_embeddings.shape` `torch.Size([3, 512])`

I can’t figure out where the issues is. I also tried to not concat with uncond\_embedding.