metadata

language:
  - tr
base_model:
  - openai/clip-vit-base-patch32
tags:
  - lora
  - peft

This is a LoRA adapter OpenAI CLIP. The LoRA model is trained on Turkish language. You can get more information (and code 🎉) on how to train or use the model on my github.

How to use the model?

You can use the model like shown in below:

from PIL import Image
from transformers import CLIPProcessor, CLIPModel

model = CLIPModel.from_pretrained("openai/clip-vit-base-patch32")
model.load_adapter("kesimeg/lora-turkish-clip")
model.eval()

processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")


img = Image.open("dog.png") # A dog image
inputs = processor(text=["Çimenler içinde bir köpek.","Bir köpek.","Çimenler içinde bir kuş."], images=img, return_tensors="pt", padding=True)
outputs = model(**inputs)
logits_per_image = outputs.logits_per_image
probs = logits_per_image.softmax(dim=1)
print(probs)