LLaVE-0.5B開源多模態嵌入模型 - 支持文本、圖像和視頻嵌入

首頁

Llave 0.5B

由zhibinlan開發

LLaVE是基於LLaVA-OneVision-0.5B模型的多模態嵌入模型，參數規模為0.5B，能夠對文本、圖像、多圖像和視頻進行嵌入。

多模態融合

Transformers

英語開源協議:Apache-2.0 #多模態嵌入 #零樣本視頻檢索 #小參數高效

下載量 2,897

發布時間 : 2/6/2025

模型概述

LLaVE是一個多模態嵌入模型，能夠處理文本、圖像和視頻數據，支持句子相似度計算、零樣本圖像分類等任務。

模型特點

多模態嵌入

能夠同時對文本、圖像和視頻數據進行嵌入處理

高效性能

僅使用少量參數和訓練數據就在MMEB排行榜上取得優異表現

零樣本遷移能力

在圖像-文本數據上訓練，但能零樣本泛化到文本-視頻檢索任務

模型能力

文本嵌入

圖像嵌入

視頻嵌入

句子相似度計算

零樣本圖像分類

跨模態檢索

使用案例

圖像檢索

基於文本的圖像搜索

根據文本描述檢索相關圖像

在MMEB評估中表現優異

跨模態檢索

文本到視頻檢索

根據文本描述檢索相關視頻片段

零樣本情況下表現出強大性能

🚀 LLaVE-0.5B

LLaVE-0.5B 是一個參數為 0.5B 的多模態嵌入模型，基於 LLaVA-OneVision-0.5B 模型構建，具有 4K 令牌的上下文窗口。它能夠對文本、圖像、多圖像和視頻進行嵌入處理，在 MMEB 排行榜上取得了優異成績，還能在零樣本的情況下泛化到文本 - 視頻檢索任務。

🚀 快速開始

首先克隆我們的 GitHub 倉庫：

git clone https://github.com/DeepLearnXMU/LLaVE
cd LLaVE
pip install -e ".[train]"

我們提供了使用該模型的簡單嵌入過程。更多詳細信息，請參考 Github。

# pip install git+https://github.com/DeepLearnXMU/LLaVE


import torch
import copy
from PIL import Image
from llava.constants import IMAGE_TOKEN_INDEX, DEFAULT_IMAGE_TOKEN
from llava.conversation import conv_templates
from llava.model.builder import load_pretrained_model
from llava.mm_utils import tokenizer_image_token, process_images

pretrained = "zhibinlan/LLaVE-0.5B"
model_name = "llava_qwen"
device = "cuda"
device_map = "auto"
tokenizer, model, image_processor, max_length = load_pretrained_model(pretrained, None, model_name, device_map=device_map)  # Add any other thing you want to pass in llava_model_args
model.eval()

# Image + Text -> Text
image = Image.open("figures/example.jpg")
image_tensor = process_images([image], image_processor, model.config)
image_tensor = [_image.to(dtype=torch.float16, device=device) for _image in image_tensor]
conv_template = "qwen_1_5"  # Make sure you use correct chat template for different models

question = DEFAULT_IMAGE_TOKEN + " Represent the given image with the following question: What is in the image"
conv = copy.deepcopy(conv_templates[conv_template])
conv.append_message(conv.roles[0], question)
conv.append_message(conv.roles[1], "\n")
prompt_question = conv.get_prompt()
input_ids = tokenizer_image_token(prompt_question, tokenizer, IMAGE_TOKEN_INDEX, return_tensors="pt").unsqueeze(0).to(device)
attention_mask=input_ids.ne(tokenizer.pad_token_id)
image_sizes = [image.size]
query_embed = model.encode_multimodal_embeddings(input_ids, attention_mask=attention_mask,images=image_tensor, image_sizes=image_sizes)

target_string = "A cat and a dog"
conv = copy.deepcopy(conv_templates[conv_template])
conv.append_message(conv.roles[0], target_string)
conv.append_message(conv.roles[1], "\n")
target_string = conv.get_prompt()
target_input_ids = tokenizer(target_string, return_tensors="pt").input_ids.to(device)
attention_mask=target_input_ids.ne(tokenizer.pad_token_id)
target_embed = model.encode_multimodal_embeddings(target_input_ids, attention_mask=attention_mask)

print("A cat and a dog similarity score: ", query_embed @ target_embed.T)
# 0.5B: A cat and a dog similarity score: tensor([[0.4802]]

neg_string = "A cat and a tiger"
conv = copy.deepcopy(conv_templates[conv_template])
conv.append_message(conv.roles[0], neg_string)
conv.append_message(conv.roles[1], "\n")
neg_string = conv.get_prompt()
neg_input_ids = tokenizer(neg_string, return_tensors="pt").input_ids.to(device)
attention_mask=neg_input_ids.ne(tokenizer.pad_token_id)
neg_embed = model.encode_multimodal_embeddings(neg_input_ids, attention_mask=attention_mask)
print("A cat and a tiger similarity score: ", query_embed @ neg_embed.T)
# 0.5B: A cat and a tiger similarity score: tensor([[0.3413]]


# Text -> Image
pos_string = "Find me an everyday image that matches the given caption: A cat and a dog."
conv = copy.deepcopy(conv_templates[conv_template])
conv.append_message(conv.roles[0], pos_string)
conv.append_message(conv.roles[1], "\n")
pos_string = conv.get_prompt()
pos_input_ids = tokenizer(pos_string, return_tensors="pt").input_ids.to(device)
attention_mask=pos_input_ids.ne(tokenizer.pad_token_id)
pos_query_embed = model.encode_multimodal_embeddings(pos_input_ids, attention_mask=attention_mask)

target = DEFAULT_IMAGE_TOKEN + " Represent the given image."
conv = copy.deepcopy(conv_templates[conv_template])
conv.append_message(conv.roles[0], target)
conv.append_message(conv.roles[1], "\n")
prompt_target = conv.get_prompt()
target_input_ids = tokenizer_image_token(prompt_target, tokenizer, IMAGE_TOKEN_INDEX, return_tensors="pt").unsqueeze(0).to(device)
attention_mask=target_input_ids.ne(tokenizer.pad_token_id)
target_image_sizes = [image.size]
target_embed = model.encode_multimodal_embeddings(target_input_ids, attention_mask=attention_mask,images=image_tensor, image_sizes=target_image_sizes)

print("A cat and a dog image similarity score: ", pos_query_embed @ target_embed.T)
# 0.5B: A cat and a dog similarity score: tensor([[0.3992]]

neg_string = "Find me an everyday image that matches the given caption: A cat and a tiger."
conv = copy.deepcopy(conv_templates[conv_template])
conv.append_message(conv.roles[0], neg_string)
conv.append_message(conv.roles[1], "\n")
neg_string = conv.get_prompt()
neg_input_ids = tokenizer(neg_string, return_tensors="pt").input_ids.to(device)
attention_mask=neg_input_ids.ne(tokenizer.pad_token_id)
neg_query_embed = model.encode_multimodal_embeddings(neg_input_ids, attention_mask=attention_mask)

print("A cat and a tiger image similarity score: ", neg_query_embed @ target_embed.T)
# 0.5B: A cat and a dog similarity score: tensor([[0.3354]]

✨ 主要特性

多模態嵌入能力：該模型能夠對文本、圖像、多圖像和視頻進行嵌入處理。
優異的性能表現：在 MMEB 排行榜上取得了優異成績，僅使用少量數據和 662K 訓練對。
零樣本泛化能力：雖然模型是在圖像 - 文本數據上訓練的，但它可以在零樣本的情況下泛化到文本 - 視頻檢索任務，並取得了不錯的性能。

📦 安裝指南

克隆 GitHub 倉庫並安裝依賴：

git clone https://github.com/DeepLearnXMU/LLaVE
cd LLaVE
pip install -e ".[train]"

💻 使用示例

基礎用法

以下代碼展示瞭如何使用該模型進行圖像 + 文本到文本的嵌入和相似度計算：

# pip install git+https://github.com/DeepLearnXMU/LLaVE


import torch
import copy
from PIL import Image
from llava.constants import IMAGE_TOKEN_INDEX, DEFAULT_IMAGE_TOKEN
from llava.conversation import conv_templates
from llava.model.builder import load_pretrained_model
from llava.mm_utils import tokenizer_image_token, process_images

pretrained = "zhibinlan/LLaVE-0.5B"
model_name = "llava_qwen"
device = "cuda"
device_map = "auto"
tokenizer, model, image_processor, max_length = load_pretrained_model(pretrained, None, model_name, device_map=device_map)  # Add any other thing you want to pass in llava_model_args
model.eval()

# Image + Text -> Text
image = Image.open("figures/example.jpg")
image_tensor = process_images([image], image_processor, model.config)
image_tensor = [_image.to(dtype=torch.float16, device=device) for _image in image_tensor]
conv_template = "qwen_1_5"  # Make sure you use correct chat template for different models

question = DEFAULT_IMAGE_TOKEN + " Represent the given image with the following question: What is in the image"
conv = copy.deepcopy(conv_templates[conv_template])
conv.append_message(conv.roles[0], question)
conv.append_message(conv.roles[1], "\n")
prompt_question = conv.get_prompt()
input_ids = tokenizer_image_token(prompt_question, tokenizer, IMAGE_TOKEN_INDEX, return_tensors="pt").unsqueeze(0).to(device)
attention_mask=input_ids.ne(tokenizer.pad_token_id)
image_sizes = [image.size]
query_embed = model.encode_multimodal_embeddings(input_ids, attention_mask=attention_mask,images=image_tensor, image_sizes=image_sizes)

target_string = "A cat and a dog"
conv = copy.deepcopy(conv_templates[conv_template])
conv.append_message(conv.roles[0], target_string)
conv.append_message(conv.roles[1], "\n")
target_string = conv.get_prompt()
target_input_ids = tokenizer(target_string, return_tensors="pt").input_ids.to(device)
attention_mask=target_input_ids.ne(tokenizer.pad_token_id)
target_embed = model.encode_multimodal_embeddings(target_input_ids, attention_mask=attention_mask)

print("A cat and a dog similarity score: ", query_embed @ target_embed.T)
# 0.5B: A cat and a dog similarity score: tensor([[0.4802]]

neg_string = "A cat and a tiger"
conv = copy.deepcopy(conv_templates[conv_template])
conv.append_message(conv.roles[0], neg_string)
conv.append_message(conv.roles[1], "\n")
neg_string = conv.get_prompt()
neg_input_ids = tokenizer(neg_string, return_tensors="pt").input_ids.to(device)
attention_mask=neg_input_ids.ne(tokenizer.pad_token_id)
neg_embed = model.encode_multimodal_embeddings(neg_input_ids, attention_mask=attention_mask)
print("A cat and a tiger similarity score: ", query_embed @ neg_embed.T)
# 0.5B: A cat and a tiger similarity score: tensor([[0.3413]]

高級用法

以下代碼展示瞭如何使用該模型進行文本到圖像的嵌入和相似度計算：

# Text -> Image
pos_string = "Find me an everyday image that matches the given caption: A cat and a dog."
conv = copy.deepcopy(conv_templates[conv_template])
conv.append_message(conv.roles[0], pos_string)
conv.append_message(conv.roles[1], "\n")
pos_string = conv.get_prompt()
pos_input_ids = tokenizer(pos_string, return_tensors="pt").input_ids.to(device)
attention_mask=pos_input_ids.ne(tokenizer.pad_token_id)
pos_query_embed = model.encode_multimodal_embeddings(pos_input_ids, attention_mask=attention_mask)

target = DEFAULT_IMAGE_TOKEN + " Represent the given image."
conv = copy.deepcopy(conv_templates[conv_template])
conv.append_message(conv.roles[0], target)
conv.append_message(conv.roles[1], "\n")
prompt_target = conv.get_prompt()
target_input_ids = tokenizer_image_token(prompt_target, tokenizer, IMAGE_TOKEN_INDEX, return_tensors="pt").unsqueeze(0).to(device)
attention_mask=target_input_ids.ne(tokenizer.pad_token_id)
target_image_sizes = [image.size]
target_embed = model.encode_multimodal_embeddings(target_input_ids, attention_mask=attention_mask,images=image_tensor, image_sizes=target_image_sizes)

print("A cat and a dog image similarity score: ", pos_query_embed @ target_embed.T)
# 0.5B: A cat and a dog similarity score: tensor([[0.3992]]

neg_string = "Find me an everyday image that matches the given caption: A cat and a tiger."
conv = copy.deepcopy(conv_templates[conv_template])
conv.append_message(conv.roles[0], neg_string)
conv.append_message(conv.roles[1], "\n")
neg_string = conv.get_prompt()
neg_input_ids = tokenizer(neg_string, return_tensors="pt").input_ids.to(device)
attention_mask=neg_input_ids.ne(tokenizer.pad_token_id)
neg_query_embed = model.encode_multimodal_embeddings(neg_input_ids, attention_mask=attention_mask)

print("A cat and a tiger image similarity score: ", neg_query_embed @ target_embed.T)
# 0.5B: A cat and a dog similarity score: tensor([[0.3354]]

📚 詳細文檔

模型概述

LLaVE 模型是基於 LLaVA-OneVision-0.5B 模型的 0.5B 參數多模態嵌入模型，上下文窗口為 4K 令牌。

倉庫地址：LLaVE
論文地址：LLaVE

訓練/評估數據

訓練數據：https://huggingface.co/datasets/TIGER-Lab/MMEB-train
評估數據：https://huggingface.co/datasets/TIGER-Lab/MMEB-eval

預期用途

該模型能夠對文本、圖像、多圖像和視頻進行嵌入處理。

MMEB 排行榜

我們僅使用少量數據就在 MMEB 排行榜上取得了領先排名。

MMEB Leaderboard

模型性能

LLaVE-0.5B 在 MMEB 上使用較少的參數和 662K 訓練對取得了優異的性能。

MMEB

儘管 LLaVE 是在圖像 - 文本數據上訓練的，但它可以在零樣本的情況下泛化到文本 - 視頻檢索任務，並取得了不錯的性能，展示了其在其他嵌入任務上的顯著遷移潛力。

🔧 技術細節

硬件與軟件

GPU：8 * Nvidia A100 (40G)（用於全模型訓練）
編排工具：Huggingface Trainer
神經網絡框架：PyTorch

📄 許可證

本項目採用 Apache-2.0 許可證。

📄 引用

@article{lan2025llave,
  title={LLaVE: Large Language and Vision Embedding Models with Hardness-Weighted Contrastive Learning},
  author={Lan, Zhibin and Niu, Liqiang and Meng, Fandong and Zhou, Jie and Su, Jinsong},
  journal={arXiv preprint arXiv:2503.04812},
  year={2025}
}