language: zh
license: creativeml-openrail-m
tags:
- stable-diffusion
- stable-diffusion-diffusers
- text-to-image
- zh
- Chinese
inference: false
extra_gated_prompt: >-
One more step before getting this model.
This model is open access and available to all, with a CreativeML OpenRAIL-M
license further specifying rights and usage.
The CreativeML OpenRAIL License specifies:
1. You can't use the model to deliberately produce nor share illegal or
harmful outputs or content
2. IDEA-CCNL claims no rights on the outputs you generate, you are free to use
them and are accountable for their use which must not go against the
provisions set in the license
3. You may re-distribute the weights and use the model commercially and/or as
a service. If you do, please be aware you have to include the same use
restrictions as the ones in the license and share a copy of the CreativeML
OpenRAIL-M to all your users (please read the license entirely and carefully)
Please read the full license here:
https://huggingface.co/spaces/CompVis/stable-diffusion-license
By clicking on "Access repository" below, you accept that your *contact
information* (email address and username) can be shared with the model authors
as well.
extra_gated_fields:
I have read the License and agree with its terms: checkbox
AltCLIP
| 名称 Name | 任务 Task | 语言 Language(s) | 模型 Model | Github |
|---|---|---|---|---|
| AltCLIP | 自然语言处理 NLP | 中英文 Chinese&English | CLIP | FlagAI |
简介 Brief Introduction
我们提出了一个简单高效的方法去训练更加优秀的双语CLIP模型。命名为AltCLIP。AltCLIP基于 Stable Diffusiosn 训练，训练数据来自 WuDao数据集 和 LIAON
AltCLIP模型可以为本项目中的AltDiffusion模型提供支持，关于AltDiffusion模型的具体信息可查看此教程 。
模型代码已经在 FlagAI 上开源，权重位于我们搭建的 modelhub 上。我们还提供了微调，推理，验证的脚本，欢迎试用。
We propose a simple and efficient method to train a better bilingual CLIP model. Named AltCLIP. AltCLIP is trained based on Stable Diffusiosn with training data from WuDao dataset and Liaon.
The AltCLIP model can provide support for the AltDiffusion model in this project. Specific information on the AltDiffusion model can be found in this tutorial.
The model code has been open sourced on FlagAI and the weights are located on modelhub. We also provide scripts for fine-tuning, inference, and validation, so feel free to try them out.
引用
关于AltCLIP，我们已经推出了相关报告，有更多细节可以查阅，如对您的工作有帮助，欢迎引用。
If you find this work helpful, please consider to cite
@article{https://doi.org/10.48550/arxiv.2211.06679,
doi = {10.48550/ARXIV.2211.06679},
url = {https://arxiv.org/abs/2211.06679},
author = {Chen, Zhongzhi and Liu, Guang and Zhang, Bo-Wen and Ye, Fulong and Yang, Qinghong and Wu, Ledell},
keywords = {Computation and Language (cs.CL), FOS: Computer and information sciences},
title = {AltCLIP: Altering the Language Encoder in CLIP for Extended Language Capabilities},
publisher = {arXiv},
year = {2022},
copyright = {arXiv.org perpetual, non-exclusive license}
}
训练 Training
训练共有两个阶段。 在平行知识蒸馏阶段，我们只是使用平行语料文本来进行蒸馏（平行语料相对于图文对更容易获取且数量更大）。在双语对比学习阶段，我们使用少量的中-英 图像-文本对（一共约2百万）来训练我们的文本编码器以更好地适应图像编码器。
There are two phases of training. In the parallel knowledge distillation phase, we only use parallel corpus texts for distillation (parallel corpus is easier to obtain and larger in number compared to image text pairs). In the bilingual comparison learning phase, we use a small number of Chinese-English image-text pairs (about 2 million in total) to train our text encoder to better fit the image encoder.
下游效果 Performance
| Language | Method | Text-to-Image Retrival | Image-to-Text Retrival | MR | ||||
| R@1 | R@5 | R@10 | R@1 | R@5 | R@10 | |||
| English | CLIP | 65.0 | 87.1 | 92.2 | 85.1 | 97.3 | 99.2 | 87.6 |
| Taiyi | 25.3 | 48.2 | 59.2 | 39.3 | 68.1 | 79.6 | 53.3 | |
| Wukong | - | - | - | - | - | - | - | |
| R2D2 | - | - | - | - | - | - | - | |
| CN-CLIP | 49.5 | 76.9 | 83.8 | 66.5 | 91.2 | 96.0 | 77.3 | |
| AltCLIP | 66.3 | 87.8 | 92.7 | 85.9 | 97.7 | 99.1 | 88.3 | |
| AltCLIP∗ | 72.5 | 91.6 | 95.4 | 86.0 | 98.0 | 99.1 | 90.4 | |
| Chinese | CLIP | 0.0 | 2.4 | 4.0 | 2.3 | 8.1 | 12.6 | 5.0 |
| Taiyi | 53.7 | 79.8 | 86.6 | 63.8 | 90.5 | 95.9 | 78.4 | |
| Wukong | 51.7 | 78.9 | 86.3 | 76.1 | 94.8 | 97.5 | 80.9 | |
| R2D2 | 60.9 | 86.8 | 92.7 | 77.6 | 96.7 | 98.9 | 85.6 | |
| CN-CLIP | 68.0 | 89.7 | 94.4 | 80.2 | 96.6 | 98.2 | 87.9 | |
| AltCLIP | 63.7 | 86.3 | 92.1 | 84.7 | 97.4 | 98.7 | 87.2 | |
| AltCLIP∗ | 69.8 | 89.9 | 94.7 | 84.8 | 97.4 | 98.8 | 89.2 | |
可视化效果 Visualization effects
基于AltCLIP，我们还开发了AltDiffusion模型，可视化效果如下。
Based on AltCLIP, we have also developed the AltDiffusion model, visualized as follows.
模型推理 Inference
import torch
from PIL import Image
from flagai.auto_model.auto_loader import AutoLoader
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
## 一行代码直接自动下载权重到'./checkpoints/clip-xlmr-large'，并自动加载CLIP模型权重
## modelhub地址: Modelhub(https://model.baai.ac.cn/models)
loader = AutoLoader(
task_name="txt_img_matching",
model_dir="./checkpoints",
model_name="AltCLIP-XLMR-L"
)
## 获取加载好的模型
model = loader.get_model()
## 获取tokenizer
tokenizer = loader.get_tokenizer()
## 获取transform用来处理图像
transform = loader.get_transform()
model.eval()
model.to(device)
## 推理过程,图像与文本匹配
image = Image.open("./dog.jpeg")
image = transform(image)
image = torch.tensor(image["pixel_values"]).to(device)
text = tokenizer(["a rat", "a dog", "a cat"])["input_ids"]
text = torch.tensor(text).to(device)
with torch.no_grad():
image_features = model.get_image_features(image)
text_features = model.get_text_features(text)
text_probs = (image_features @ text_features.T).softmax(dim=-1)
print(text_probs.cpu().numpy()[0].tolist())
CLIP微调 Finetuning
微调采用cifar10数据集，并使用FlagAI的Trainer快速开始训练过程。
Fine-tuning was done using the cifar10 dataset and using FlagAI's Trainer to quickly start the training process.
# Copyright © 2022 BAAI. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License")
import torch
from flagai.auto_model.auto_loader import AutoLoader
import os
from flagai.trainer import Trainer
from torchvision.datasets import (
CIFAR10
)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
dataset_root = "./clip_benchmark_datasets"
dataset_name = "cifar10"
batch_size = 4
classes = ['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']
auto_loader = AutoLoader(
task_name="txt_img_matching",
model_dir="./checkpoints/",
model_name="AltCLIP-XLMR-L" # Load the checkpoints from Modelhub(model.baai.ac.cn/models)
)
model = auto_loader.get_model()
model.to(device)
model.eval()
tokenizer = auto_loader.get_tokenizer()
transform = auto_loader.get_transform()
trainer = Trainer(env_type="pytorch",
pytorch_device=device,
experiment_name="clip_finetuning",
batch_size=4,
lr=1e-4,
epochs=10,
log_interval=10)
dataset = CIFAR10(root=os.path.join(dataset_root, dataset_name),
transform=transform,
download=True)
def cifar10_collate_fn(batch):
# image shape is (batch, 3, 224, 224)
images = torch.tensor([b[0]["pixel_values"][0] for b in batch])
# text_id shape is (batch, n)
input_ids = torch.tensor([tokenizer(f"a photo of a {b[1]}",padding=True,truncation=True,max_length=77)["input_ids"] for b in batch])
return {
"pixel_values": images,
"input_ids": input_ids
}
if __name__ == "__main__":
trainer.train(model=model, train_dataset=dataset, collate_fn=cifar10_collate_fn)
模型验证 Evaluation
我们提供了可以直接运行的验证脚本，在cifar10数据集上进行验证。
期待的输出为：{'dataset': 'cifar10', 'metrics': {'acc1': 0.95402, 'acc5': 0.99616, 'mean_per_class_recall': 0.9541200000000002}}
We provide validation scripts that can be run directly on the cifar10 dataset.
# Copyright © 2022 BAAI. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License")
import torch
from flagai.auto_model.auto_loader import AutoLoader
from metrics import zeroshot_classification
import json
import os
from torchvision.datasets import CIFAR10
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
maxlen = 256
dataset_root = "./clip_benchmark_datasets"
dataset_name = "cifar10"
auto_loader = AutoLoader(
task_name="txt_img_matching",
model_dir="./checkpoints/",
model_name="AltCLIP-XLMR-L"
)
model = auto_loader.get_model()
model.to(device)
model.eval()
tokenizer = auto_loader.get_tokenizer()
transform = auto_loader.get_transform()
dataset = CIFAR10(root=os.path.join(dataset_root, dataset_name),
transform=transform,
download=True)
batch_size = 128
num_workers = 4
template = {"cifar10": [
"a photo of a {c}.",
"a blurry photo of a {c}.",
"a black and white photo of a {c}.",
"a low contrast photo of a {c}.",
"a high contrast photo of a {c}.",
"a bad photo of a {c}.",
"a good photo of a {c}.",
"a photo of a small {c}.",
"a photo of a big {c}.",
"a photo of the {c}.",
"a blurry photo of the {c}.",
"a black and white photo of the {c}.",
"a low contrast photo of the {c}.",
"a high contrast photo of the {c}.",
"a bad photo of the {c}.",
"a good photo of the {c}.",
"a photo of the small {c}.",
"a photo of the big {c}."
],
}
def evaluate():
if dataset:
dataloader = torch.utils.data.DataLoader(
dataset,
batch_size=batch_size,
shuffle=False,
num_workers=num_workers,
)
classnames = dataset.classes if hasattr(dataset, "classes") else None
zeroshot_templates = template["cifar10"]
metrics = zeroshot_classification.evaluate(
model,
dataloader,
tokenizer,
classnames,
zeroshot_templates,
device=device,
amp=True,
)
dump = {
"dataset": dataset_name,
"metrics": metrics
}
print(dump)
with open("./result.txt", "w") as f:
json.dump(dump, f)
return metrics
if __name__ == "__main__":
evaluate()