NEW
Browse files
README.md
DELETED
@@ -1,81 +0,0 @@
|
|
1 |
-
---
|
2 |
-
license: mit
|
3 |
-
pipeline_tag: text-generation
|
4 |
-
tags:
|
5 |
-
- ocean
|
6 |
-
- text-generation-inference
|
7 |
-
- oceangpt
|
8 |
-
language:
|
9 |
-
- en
|
10 |
-
datasets:
|
11 |
-
- zjunlp/OceanBench
|
12 |
-
---
|
13 |
-
|
14 |
-
## ๐ก Model description
|
15 |
-
This repo contains a large language model (OceanGPT) for ocean science tasks trained with [KnowLM](https://github.com/zjunlp/KnowLM).
|
16 |
-
It should be noted that the OceanGPT is constantly being updated, so the current model is not the final version.
|
17 |
-
|
18 |
-
OceanGPT-7B-v0.2 is based on Qwen2-7B and trained on a bilingual dataset in Chinese and English.
|
19 |
-
## ๐ Intended uses
|
20 |
-
You can download the model to generate responses or contact the [email]([email protected]) for the online test demo.
|
21 |
-
|
22 |
-
## ๐ ๏ธ How to use OceanGPT
|
23 |
-
We wil provide several examples soon and you can modify the input according to your needs.
|
24 |
-
|
25 |
-
```python
|
26 |
-
from transformers import AutoModelForCausalLM, AutoTokenizer
|
27 |
-
import torch
|
28 |
-
device = "cuda" # the device to load the model onto
|
29 |
-
|
30 |
-
model = AutoModelForCausalLM.from_pretrained(
|
31 |
-
"zjunlp/OceanGPT-7B-v0.2",
|
32 |
-
torch_dtype=torch.bfloat16,
|
33 |
-
device_map="auto"
|
34 |
-
)
|
35 |
-
tokenizer = AutoTokenizer.from_pretrained("zjunlp/OceanGPT-7B-v0.2")
|
36 |
-
|
37 |
-
prompt = "Which is the largest ocean in the world?"
|
38 |
-
messages = [
|
39 |
-
{"role": "system", "content": "You are a helpful assistant."},
|
40 |
-
{"role": "user", "content": prompt}
|
41 |
-
]
|
42 |
-
text = tokenizer.apply_chat_template(
|
43 |
-
messages,
|
44 |
-
tokenize=False,
|
45 |
-
add_generation_prompt=True
|
46 |
-
)
|
47 |
-
model_inputs = tokenizer([text], return_tensors="pt").to(device)
|
48 |
-
|
49 |
-
generated_ids = model.generate(
|
50 |
-
model_inputs.input_ids,
|
51 |
-
max_new_tokens=512
|
52 |
-
)
|
53 |
-
generated_ids = [
|
54 |
-
output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
|
55 |
-
]
|
56 |
-
|
57 |
-
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
|
58 |
-
```
|
59 |
-
|
60 |
-
## ๐ ๏ธ How to evaluate your model in OceanBench
|
61 |
-
|
62 |
-
We wil provide several examples soon and you can modify the input according to your needs.
|
63 |
-
|
64 |
-
*Note: We are conducting the final checks on OceanBench and will be uploading it to Hugging Face soon.
|
65 |
-
|
66 |
-
```python
|
67 |
-
>>> from datasets import load_dataset
|
68 |
-
|
69 |
-
>>> dataset = load_dataset("zjunlp/OceanBench")
|
70 |
-
```
|
71 |
-
|
72 |
-
## ๐ How to cite
|
73 |
-
|
74 |
-
```bibtex
|
75 |
-
@article{bi2023oceangpt,
|
76 |
-
title={OceanGPT: A Large Language Model for Ocean Science Tasks},
|
77 |
-
author={Bi, Zhen and Zhang, Ningyu and Xue, Yida and Ou, Yixin and Ji, Daxiong and Zheng, Guozhou and Chen, Huajun},
|
78 |
-
journal={arXiv preprint arXiv:2310.02031},
|
79 |
-
year={2023}
|
80 |
-
}
|
81 |
-
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|