🚀 LLM2CLIP: Extending the Capability Boundaries of CLIP through Large Language Models
A novel approach that uses LLMs to unlock CLIP’s potential, improving cross - modal task performance.
🚀 Quick Start
In this paper, we propose LLM2CLIP, a novel approach that embraces the power of LLMs to unlock CLIP’s potential. By fine - tuning the LLM in the caption space with contrastive learning, we extract its textual capabilities into the output embeddings, significantly improving the output layer’s textual discriminability. We then design an efficient training process where the fine - tuned LLM acts as a powerful teacher for CLIP’s visual encoder. Thanks to the LLM’s presence, we can now incorporate longer and more complex captions without being restricted by vanilla CLIP text encoder’s context window and ability limitations. Our experiments demonstrate that this approach brings substantial improvements in cross - modal tasks. Our method directly boosted the performance of the previously SOTA EVA02 model by 16.5% on both long - text and short - text retrieval tasks, transforming a CLIP model trained solely on English data into a state - of - the - art cross - lingual model. Moreover, when integrated into multi - modal training with models like Llava 1.5, it consistently outperformed CLIP across nearly all benchmarks, demonstrating comprehensive performance improvements.
✨ Features
- Enhanced Textual Discriminability: By fine - tuning the LLM in the caption space with contrastive learning, we extract its textual capabilities into the output embeddings, significantly improving the output layer’s textual discriminability.
- Longer and More Complex Captions: Thanks to the LLM’s presence, we can incorporate longer and more complex captions without being restricted by vanilla CLIP text encoder’s context window and ability limitations.
- Improved Cross - Modal Performance: Our approach brings substantial improvements in cross - modal tasks, boosting the performance of previous SOTA models and enabling cross - lingual capabilities.
📦 Installation
No installation steps are provided in the original document.
💻 Usage Examples
Basic Usage
Image Embeddings
from PIL import Image
from transformers import AutoModel
from transformers import CLIPImageProcessor
import torch
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
image_path = "CLIP.png"
model_name_or_path = "microsoft/LLM2CLIP-Openai-L-14-224"
processor = CLIPImageProcessor.from_pretrained("openai/clip-vit-large-patch14")
model = AutoModel.from_pretrained(
model_name_or_path,
torch_dtype=torch.bfloat16,
trust_remote_code=True).to('cuda').eval()
image = Image.open(image_path)
input_pixels = processor(images=image, return_tensors="pt").pixel_values.to('cuda')
with torch.no_grad(), torch.cuda.amp.autocast():
outputs = model.get_image_features(input_pixels)
Retrieval
from PIL import Image
from transformers import AutoModel, AutoConfig, AutoTokenizer
from transformers import CLIPImageProcessor
import torch
from llm2vec import LLM2Vec
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
processor = CLIPImageProcessor.from_pretrained("openai/clip-vit-large-patch14")
model_name_or_path = "microsoft/LLM2CLIP-Openai-L-14-224"
model = AutoModel.from_pretrained(
model_name_or_path,
torch_dtype=torch.bfloat16,
trust_remote_code=True).to('cuda').eval()
llm_model_name = 'microsoft/LLM2CLIP-Llama-3-8B-Instruct-CC-Finetuned'
config = AutoConfig.from_pretrained(
llm_model_name, trust_remote_code=True
)
llm_model = AutoModel.from_pretrained(llm_model_name, torch_dtype=torch.bfloat16, config=config,trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(llm_model_name)
llm_model.config._name_or_path = 'meta-llama/Meta-Llama-3-8B-Instruct'
l2v = LLM2Vec(llm_model, tokenizer, pooling_mode="mean", max_length=512, doc_max_length=512)
captions = ["a diagram", "a dog", "a cat"]
image_path = "CLIP.png"
image = Image.open(image_path)
input_pixels = processor(images=image, return_tensors="pt").pixel_values.to('cuda')
text_features = l2v.encode(captions, convert_to_tensor=True).to('cuda')
with torch.no_grad(), torch.cuda.amp.autocast():
image_features = model.get_image_features(input_pixels)
text_features = model.get_text_features(text_features)
image_features /= image_features.norm(dim=-1, keepdim=True)
text_features /= text_features.norm(dim=-1, keepdim=True)
text_probs = (100.0 * image_features @ text_features.T).softmax(dim=-1)
print("Label probs:", text_probs)
📚 Documentation
Model Details
Property |
Details |
Model Type |
vision foundation model, feature backbone |
Pretrain Dataset |
CC3M, CC12M, YFCC15M and Recap - DataComp - 1B(30M subset) |
Important Note
⚠️ Important Note
It's important to note that all results presented in the paper are evaluated using PyTorch weights. There may be differences in performance when using Hugging Face (hf) models.
📄 License
The project is licensed under the Apache - 2.0 license.
BibTeX & Citation
@misc{huang2024llm2clippowerfullanguagemodel,
title={LLM2CLIP: Powerful Language Model Unlock Richer Visual Representation},
author={Weiquan Huang and Aoqi Wu and Yifan Yang and Xufang Luo and Yuqing Yang and Liang Hu and Qi Dai and Xiyang Dai and Dongdong Chen and Chong Luo and Lili Qiu},
year={2024},
eprint={2411.04997},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2411.04997},
}