File size: 5,073 Bytes
b2cf739
9307c59
b2cf739
 
 
 
9307c59
b2cf739
 
9307c59
333d119
b2cf739
 
3d228c1
 
b2cf739
196830f
b2cf739
3d228c1
 
 
 
55588fa
3d228c1
 
 
4e29b01
 
3d228c1
4e29b01
3d228c1
4e29b01
d196a84
 
bcd5b5a
3d228c1
 
607c999
3d228c1
e7f8e67
3d228c1
 
 
 
 
 
 
 
 
53c1b62
 
3d228c1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
53c1b62
3d228c1
53c1b62
 
3d228c1
 
 
 
 
 
 
7f86ba9
3d228c1
 
196830f
3d228c1
977e1b9
 
 
 
 
 
 
 
d196a84
3d228c1
 
 
 
bcd5b5a
 
 
 
 
 
 
 
3d228c1
9307c59
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
---
license: mit
pipeline_tag: text-generation
tags:
- ocean
- text-generation-inference
- oceangpt
language:
- en
datasets:
- zjunlp/OceanInstruct
---

<div align="center">
<img src="logo.jpg" width="300px">

**OceanGPT(沧渊): A Large Language Model for Ocean Science Tasks**

<p align="center">
  <a href="https://github.com/zjunlp/OceanGPT">Project</a><a href="https://arxiv.org/abs/2310.02031">Paper</a><a href="https://huggingface.co/collections/zjunlp/oceangpt-664cc106358fdd9f09aa5157">Models</a><a href="http://oceangpt.zjukg.cn/">Web</a><a href="#quickstart">Quickstart</a><a href="#citation">Citation</a>
</p>


</div>

OceanGPT-7b-v0.1 is based on LLaMA2 and has been trained on an English dataset in the ocean domain.

-**Disclaimer: This project is purely an academic exploration rather than a product. Please be aware that due to the inherent limitations of large language models, there may be issues such as hallucinations.**


## ⏩Quickstart
### Download the model

Download the model: [OceanGPT-7b-v0.1](https://huggingface.co/zjunlp/OceanGPT-7b-v0.1)

```shell
git lfs install
git clone https://huggingface.co/zjunlp/OceanGPT-7b-v0.1
```
or
```
huggingface-cli download --resume-download zjunlp/OceanGPT-7b-v0.1 --local-dir OceanGPT-7b-v0.1 --local-dir-use-symlinks False
```
### Inference

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
device = "cuda" # the device to load the model onto
path = 'YOUR-MODEL-PATH'
model = AutoModelForCausalLM.from_pretrained(
    path,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(path)

prompt = "Which is the largest ocean in the world?"
messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(device)

generated_ids = model.generate(
    model_inputs.input_ids,
    max_new_tokens=512
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
```

## 📌Models

| Model Name        | HuggingFace                                                          | WiseModel                                                                 | ModelScope                                                                |
|-------------------|-----------------------------------------------------------------------------------|----------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------|
| OceanGPT-14B-v0.1 (based on Qwen) | <a href="https://huggingface.co/zjunlp/OceanGPT-14B-v0.1" target="_blank">14B</a> | <a href="https://wisemodel.cn/models/zjunlp/OceanGPT-14B-v0.1" target="_blank">14B</a> | <a href="https://modelscope.cn/models/ZJUNLP/OceanGPT-14B-v0.1" target="_blank">14B</a> |
| OceanGPT-7B-v0.2 (based on Qwen) | <a href="https://huggingface.co/zjunlp/OceanGPT-7b-v0.2" target="_blank">7B</a>   | <a href="https://wisemodel.cn/models/zjunlp/OceanGPT-7b-v0.2" target="_blank">7B</a>   | <a href="https://modelscope.cn/models/ZJUNLP/OceanGPT-7b-v0.2" target="_blank">7B</a>   |
| OceanGPT-2B-v0.1 (based on MiniCPM) | <a href="https://huggingface.co/zjunlp/OceanGPT-2B-v0.1" target="_blank">2B</a>   | <a href="https://wisemodel.cn/models/zjunlp/OceanGPT-2b-v0.1" target="_blank">2B</a>   | <a href="https://modelscope.cn/models/ZJUNLP/OceanGPT-2B-v0.1" target="_blank">2B</a>   |
 
## 🌻Acknowledgement

OceanGPT(沧渊) is trained based on the open-sourced large language models including [Qwen](https://huggingface.co/Qwen), [MiniCPM](https://huggingface.co/collections/openbmb/minicpm-2b-65d48bf958302b9fd25b698f), [LLaMA](https://huggingface.co/meta-llama). Thanks for their great contributions!

## Limitations

- The model may have hallucination issues.

- We did not optimize the identity and the model may generate identity information similar to that of Qwen/MiniCPM/LLaMA/GPT series models.

- The model's output is influenced by prompt tokens, which may result in inconsistent results across multiple attempts.  

- The model requires the inclusion of specific simulator code instructions for training in order to possess simulated embodied intelligence capabilities (the simulator is subject to copyright restrictions and cannot be made available for now), and its current capabilities are quite limited.

### 🚩Citation

Please cite the following paper if you use OceanGPT in your work.

```bibtex
@article{bi2023oceangpt,
  title={OceanGPT: A Large Language Model for Ocean Science Tasks},
  author={Bi, Zhen and Zhang, Ningyu and Xue, Yida and Ou, Yixin and Ji, Daxiong and Zheng, Guozhou and Chen, Huajun},
  journal={arXiv preprint arXiv:2310.02031},
  year={2023}
}

```