File size: 1,978 Bytes
d2abbc2
 
 
 
 
 
67aebae
a582dc5
67aebae
a582dc5
 
 
 
 
 
 
 
 
67aebae
a582dc5
 
 
 
67aebae
a582dc5
 
 
67aebae
 
a582dc5
 
 
 
 
 
 
 
67aebae
a582dc5
 
 
 
67aebae
a582dc5
 
 
67aebae
 
a582dc5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d2abbc2
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
---
language:
- en
- zh
pipeline_tag: text-generation
---
## chatbloom-7b

This is a RLHF enhanced bloom model (chatbloom), fine-tuned based on bloom-7b (Muennighoff et al.). This model only uses English QA datasets for RLHF training, which improves the understanding and generation of English.

### Usage

If you don't have a good GPU (mem > 20G) then use the code below:

```python
# pip install -q transformers accelerate
from transformers import AutoModelForCausalLM, AutoTokenizer

checkpoint = "hongyin/chatbloom-7b"

tokenizer = AutoTokenizer.from_pretrained(checkpoint)
model = AutoModelForCausalLM.from_pretrained(checkpoint)

inputs = tokenizer.encode("Paraphrasing the text: I love you.", return_tensors="pt")
outputs = model.generate(inputs)
print(tokenizer.decode(outputs[0]))

Original ouput: Paraphrasing the text: I love you. I love you. I love you. I love
ChatBloom ouput: Paraphrasing the text: I love you. I am a good person.
```

If you have a good GPU (mem > 20G) then use the code below:

```python
# pip install -q transformers accelerate
from transformers import AutoModelForCausalLM, AutoTokenizer

checkpoint = "hongyin/chatbloom-7b"

tokenizer = AutoTokenizer.from_pretrained(checkpoint)
model = AutoModelForCausalLM.from_pretrained(checkpoint, torch_dtype="auto", device_map="auto")

inputs = tokenizer.encode("Paraphrasing the text: I love you.", return_tensors="pt").to("cuda")
outputs = model.generate(inputs)
print(tokenizer.decode(outputs[0]))

Original ouput: Paraphrasing the text: I love you. I love you. I love you. I love
ChatBloom ouput: Paraphrasing the text: I love you. I am a good person.
```

## Bibtex entry and citation info
Please cite if you find it helpful.
```
@article{zhu2023metaaid,
  title={MetaAID 2.0: An Extensible Framework for Developing Metaverse Applications via Human-controllable Pre-trained Models},
  author={Zhu, Hongyin},
  journal={arXiv preprint arXiv:2302.13173},
  year={2023}
}

```

---
license: other
---