README.md

---
pipeline_tag: zero-shot-image-classification
---
Adds the missing layers from [openai/clip-vit-large-patch14-336](https://huggingface.co/openai/clip-vit-large-patch14-336/blob/32aef857e655710d7c25515da6decdb4b4026f44/model.safetensors) to [microsoft/LLM2CLIP-Openai-L-14-336](https://huggingface.co/microsoft/LLM2CLIP-Openai-L-14-336/blob/94ebb8b48c9514af2d32c83b168fa863d1227323/model.safetensors) so that it can be loaded in ComfyUI.

```py
import torch
from safetensors.torch import load_file, save_file


def newclip(original_path, new_path, combined_path):
    original = load_file(original_path)
    new = load_file(new_path)
    combined = {}

    for key in original.keys():
        if key in new.keys():
            combined[key] = new[key]
        else:
            combined[key] = original[key]

    save_file(combined, combined_path)

newclip(
    "./original-clip-l.safetensors",
    "./new-llm2clip-clip-l.safetensors",
    "./combined.safetensors"
)
```