FreedomIntelligence
/

ALLaVA-3B-Longer

@@ -7,67 +7,129 @@ language:
 pipeline_tag: text-generation
 ---
-Quick start:
-```shell
-from transformers import AutoModelForCausalLM
-from transformers import AutoTokenizer
-import torch
-import pdb
-dir = "FreedomIntelligence/ALLaVA-3B-Longer"
-device = 'cuda'
-model = AutoModelForCausalLM.from_pretrained(dir, trust_remote_code=True, device_map=device, torch_dtype=torch.bfloat16)
-tokenizer = AutoTokenizer.from_pretrained(dir)
-model.tokenizer = tokenizer
-gen_kwargs = {
-    'min_new_tokens': 20,
-    'max_new_tokens': 100,
-    'do_sample': False,
-    'eos_token_id': tokenizer.eos_token_id # this is a must since transformers ~4.37
-}
-#################################################################################
-# first round
-#################################################################################
-response, history = model.chat(
-    texts='What is in the image?',
-    images=['https://cdn-icons-png.flaticon.com/256/6028/6028690.png'],
-    return_history=True,
-    **gen_kwargs
-)
-print('response:')
-print(response)
-print('history:')
-print(history)
-# response:
-# The image contains a large, stylized "HI!" in a bright pink color with a yellow outline. The "HI!" is in a speech bubble shape.
-# history:
-# [['What is in the image?', 'The image contains a large, stylized "HI!" in a bright pink color with a yellow outline. The "HI!" is in a speech bubble shape.']]
-#################################################################################
-# second round
-#################################################################################
-response, history = model.chat(
-    texts='Are you sure?',
-    images=['https://cdn-icons-png.flaticon.com/256/6028/6028690.png'], # images need to be passed again in multi-round conversations
-    history=history,
-    return_history=True,
-    **gen_kwargs
-)
-print('response:')
-print(response)
-print('history:')
-print(history)
-# response:
-# Yes, I'm sure. The image shows a large, stylized "HI!" in a bright pink color with a yellow outline, placed in a speech bubble shape.
-# history:
-# [['What is in the image?', 'The image contains a large, stylized "HI!" in a bright pink color with a yellow outline. The "HI!" is in a speech bubble shape.'], ['Are you sure?', 'Yes, I\'m sure. The image shows a large, stylized "HI!" in a bright pink color with a yellow outline, placed in a speech bubble shape.']]
 ```

 pipeline_tag: text-generation
 ---
+# ALLaVA: Harnessing  GPT4V-synthesized Data for A Lite Vision-Language Model
+<p align="center">
+⚡ALLaVA is a project that provides a large-scale GPT4V-synthesized  dataset for training LVLMs.⚡
+</p>
+<!-- <p align="center">
+![Python 3.10](https://img.shields.io/badge/Python-3.10-lightblue) ![Pytorch 1.13.0](https://img.shields.io/badge/PyTorch-2.1.1-lightblue) ![transformers](https://img.shields.io/badge/transformers-4.37.0-lightblue)
+</p> -->
+<p align="center">
+   📃 <a href="https://arxiv.org/abs/2402.11684" target="_blank">Paper</a>  • 🌐 <a href="https://allava.freedomai.cn/#/" target="_blank">Demo</a>  • 👨🏻‍💻 <a href="https://github.com/FreedomIntelligence/ALLaVA" target="_blank">Github</a>
+</p>
+<p align="center">
+   🤗 <a href="https://huggingface.co/datasets/FreedomIntelligence/ALLaVA-4V" target="_blank">ALLaVA-4V Dataset</a>
+</p>
+<p align="center">
+   🤗 <a href="https://huggingface.co/FreedomIntelligence/ALLaVA-3B-Longer" target="_blank">ALLaVA-3B-Longer</a> • 🤗 <a href="https://huggingface.co/FreedomIntelligence/ALLaVA-3B" target="_blank">ALLaVA-3B</a>
+</p>
+<!-- <p align="center">
+   📃 <a href="https://arxiv.org/abs/2402.11684" target="_blank">Paper</a>  • 🌐 <a href="https://allava.freedomai.cn/#/" target="_blank">Demo</a> • 🤗 <a href="https://huggingface.co/datasets/FreedomIntelligence/ALLaVA-4V" target="_blank">ALLaVA-4V Dataset</a> • 🤗 <a href="https://huggingface.co/FreedomIntelligence/ALLaVA-3B-Longer" target="_blank">ALLaVA-3B-Longer</a> • 🤗 <a href="https://huggingface.co/FreedomIntelligence/ALLaVA-3B" target="_blank">ALLaVA-3B</a>
+    <br>  <a href="https://github.com/FreedomIntelligence/CMB/blob/main/README_zh.md">   中文</a> | <a href="https://github.com/FreedomIntelligence/CMB/blob/main/README.md"> English
+</p> -->
+## Benchmark Result
+Our model [**ALLaVA-3B-Longer**](https://huggingface.co/FreedomIntelligence/ALLaVA-3B-Longer) and [**ALLaVA-3B**](https://huggingface.co/FreedomIntelligence/ALLaVA-3B) achieve competitive results on 12 benchmarks. Bold numbers denote the SOTA performance among 3B-scale models.
+| Model | Backbone | Vicuna-80 | MMB | SEEDBench-v1 (img) | MM-Vet | MMMU (val) | MME | TextVQA | GQA | EMT (CIFAR10) | MLLM-Bench | TouchStone | LLaVA (In-the-Wild) |
+|-------|----------|-----------|-----|-------------|--------|----------|-----|------|-----|---------|----|----|--------|
+| Qwen-VL-Chat | Qwen-7B | - | 60.6 | 65.4 | - | 35.9 | 1487.5 | 61.5 | 57.5 | - | 6.2 | 711.6 | - |
+| LLaVA-v1.5-7B | Vicuna-7B | - | 64.3 | - | 31.1 | - | 1510.7 | 58.2 | 62.0 | - | - | | 65.4 |
+| LLaVA-v1.5-13B | Vicuna-13B | 22.50 | 67.7 | 68.2 | 35.4 | 36.4 | 1531.3 | 61.3 | 63.3 | 85.0 | 7.4 | 637.7 | 70.7 |
+| ShareGPT4V-7B | Vicuna-7B | - | 68.8 | 69.7 | 37.6 | - | 1943.8 | 60.4 | 63.3 | - | - | - | 72.6 |
+| TinyGPT-V | Phi2-2.7B | - | - | - | - | - | - | - | 33.6 | - | - | - | - |
+| MobileVLM | MobileLLaMA-2.7B | - | 59.6 | - | - | - | 1288.9 | 47.5 | - | - | - | - | - |
+| LLaVA-Phi | Phi2-2.7B | - | 59.8 | - | 28.9 | - | 1335.1 | 48.6 | - | - | - | - | - |
+| **ALLaVA-3B** | Phi2-2.7B | 48.8 | 64.0 | 65.2 | 32.2 | **35.3** | **1623.2** | 49.5 | 48.8 | **90.2** | 6.7 | 632.0 | 69.4 |
+| **ALLaVA-3B-Longer** | Phi2-2.7B | **52.5** | **64.6** | **65.6** | **35.5** | 33.2 | 1564.6 | **50.3** | **50.0** | 85.9 | **8.8** | **636.5** | **71.7** |
+The detailed information of each benchmark is shown in Table 4 of our [technical report](https://arxiv.org/pdf/2402.11684.pdf).
+## 🏭 Inference
+### Load from 🤗 (Recommended)
+See the [example script](https://github.com/FreedomIntelligence/ALLaVA/blob/main/allava/serve/huggingface_inference.py).
+### CLI
+See [here](https://github.com/FreedomIntelligence/ALLaVA/tree/main?tab=readme-ov-file#cli) for CLI code snippet.
+## 🏋️‍♂️ Training
+### Data
+<div align=center>
+<img src="training_datasets_by_stage.jpg" width = "640" alt="training_datasets" align=center />
+</div>
+As shown in the table, ALLaVA-3B uses 1M and 1.5M data for PT. and FT., respectively.
+ALLaVA-3B-Longer trains one more epoch (i.e. 3M in total) for the FT. stage.
+### Code
+The training code is largely based on [LLaVA-v1.5](https://github.com/haotian-liu/LLaVA).
+We wholeheartedly express our gratitude for their invaluable contributions to open-sourcing LVLMs.
+### Cost
+We train our models on 8*A800 GPUs.
+[ALLaVA-3B-Longer](https://huggingface.co/FreedomIntelligence/ALLaVA-3B-Longer) takes 8.3h for PT and 21.3h for FT.
+[ALLaVA-3B](https://huggingface.co/FreedomIntelligence/ALLaVA-3B) takes 8.3h for PT and 10.6h for FT.
+These two models share the same PT procedure.
+### Hyperparameters
+| Global Batch Size| ZeRO Stage| Optimizer | Max LR| Min LR | Scheduler | Max length | Weight decay |
+| ---: | ---: |--:| ---: | ---: | ---: | ---: | ---: |
+| 256 (PT) / 128 (FT) | 1| AdamW | 2e-5 | 2e-6 | CosineAnnealingWarmRestarts | 2048 | 0 |
+The LM backbone, projector are trainable, while the vision encoder is kept frozen.
+**The trainabilities of each module are the same for both stages.**
+## 📚 ALLaVA-4V Data
+The majority part of training data is [ALLaVA-4V](https://huggingface.co/datasets/FreedomIntelligence/ALLaVA-4V). See [here](https://github.com/FreedomIntelligence/ALLaVA/tree/main?tab=readme-ov-file#data-preparation) to prepare it for training.
+## 🙌 Contributors
+- Project Leader: [Guiming Hardy Chen](https://g-h-chen.github.io/)
+- Data: Shunian Chen, [Junying Chen](https://jymchen.github.io/), Xiangbo Wu
+- Evaluation: [Ruifei Zhang](https://scholar.google.com/citations?user=W4zOhmEAAAAJ&hl=zh-CN)
+- Deployment: Xiangbo Wu, Zhiyi Zhang
+- Advising: [Zhihong Chen](https://zhjohnchan.github.io/), [Benyou Wang](https://wabyking.github.io/old.html)
+- Others: Jianquan Li, [Xiang Wan](https://scholar.google.com/citations?user=e3_kWigAAAAJ&hl=zh-CN)
+## 📝 Citation
+If you find our data useful, please consider citing our work! We are FreedomIntelligence from [Shenzhen Research Institute of Big Data](http://sribd.cn/en) and [The Chinese University of Hong Kong, Shenzhen](https://sds.cuhk.edu.cn/en)
+```
+@article{chen2024allava,
+  title={ALLaVA: Harnessing GPT4V-synthesized Data for A Lite Vision-Language Model},
+  author={Chen, Guiming Hardy and Chen, Shunian and Zhang, Ruifei and Chen, Junying and Wu, Xiangbo and Zhang, Zhiyi and Chen, Zhihong and Li, Jianquan and Wan, Xiang and Wang, Benyou},
+  journal={arXiv preprint arXiv:2402.11684},
+  year={2024}
+}
 ```