File size: 6,724 Bytes
092efe0
 
0089ab7
 
 
 
092efe0
 
 
0089ab7
 
092efe0
 
 
 
 
9893aa5
 
bcfebc0
9893aa5
 
 
 
 
 
 
 
 
092efe0
bcfebc0
092efe0
bcfebc0
092efe0
bcfebc0
 
 
 
 
092efe0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bcfebc0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0089ab7
 
c4d3569
0089ab7
f16ab32
 
 
 
0089ab7
3dbabd8
0187783
 
 
525e6b8
6e5b68f
0187783
 
 
d91dcaa
0089ab7
3dbabd8
0089ab7
 
 
918e74e
0089ab7
0f35095
 
 
 
 
 
 
3dbabd8
 
 
 
ad66547
9cbcd8e
 
 
3dbabd8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3becaca
0187783
3dbabd8
 
9018bb3
3dbabd8
0089ab7
 
 
 
3dbabd8
 
 
 
 
 
 
 
 
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
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
---
datasets: 
  - natural_instructions
  - the_pile
  - cot
  - Muennighoff/P3
inference: 
  parameters: 
    max_new_tokens: 5
    temperature: 1.0
    top_k: 1
language: 
  - en
pipeline_tag: text-generation
widget: 
  - 
    example_title: "Sentiment Analysis"
    text: |-
        The task is to label the post's emotion as sadness, joy, love, anger, fear, or surprise.

        Input: I'm feeling quite sad and sorry for myself but ill snap out of it soon.
        Output: sadness
        
        Input: I am just feeling cranky and blue.
        Output: anger
        
        Input: I can have for a treat or if i am feeling festive.
        Output:
  - 
    example_title: "Country Currency"
    text: |-
        Return the currency of the given country.
        
        Input: Switzerland
        Output: Swiss Franc

        Input: India
        Output:
  - 
    example_title: "Tweet Eval Hate"
    text: |-
        Label whether the following tweet contains hate speech against either immigrants or women. Hate Speech (HS) is commonly defined as any communication that disparages a person or a group on the basis of some characteristic such as race, color, ethnicity, gender, sexual orientation, nationality, religion, or other characteristics.
        Possible labels:
        1. hate speech
        2. not hate speech
        
        Tweet: HOW REFRESHING! In South Korea, there is no such thing as 'political correctness" when it comes to dealing with Muslim refugee wannabes via @user
        Label: hate speech
        
        Tweet: New to Twitter-- any men on here know what the process is to get #verified?
        Label: not hate speech
        
        Tweet: Dont worry @user you are and will always be the most hysterical woman.
        Label:
  -
    example_title: "Entity Recognition"
    text: |-
        Extract all the names of people, places, and organizations from the following sentences.

        Sentence: Satya Nadella, the CEO of Microsoft, was visiting the Bahamas last May.
        Entities: Satya Nadella, Microsoft, Bahamas
        
        Sentence: Pacific Northwest cities include Seattle and Portland, which I have visited with Vikash.
        Entities:
  - 
    example_title: "Data Clearning"
    text: |-
        Format the data into a CSV file:

        Input: Jane Doe jane.doe@gmail.com (520) 382 2435
        Output: Jane Doe,jane.doe@gmail.com,520-382-2435
        
        Input: Peter Lee (510) 333-2429 email: peter@yahoo.com
        Output:
---

<h1 style="font-size: 42px">GPT-JT<h1/>


***<p style="font-size: 24px">Feel free to try out our [Online Demo](https://huggingface.co/spaces/togethercomputer/GPT-JT)!</p>***


# Model Summary

> With a new decentralized training algorithm, we fine-tuned GPT-J (6B) on 3.53 billion tokens, resulting in GPT-JT (6B), a model that outperforms many 100B+ parameter models on classification benchmarks.

We incorporated a collection of open techniques and datasets to build GPT-JT:
- GPT-JT is a fork of [EleutherAI](https://www.eleuther.ai)'s [GPT-J (6B)](https://huggingface.co/EleutherAI/gpt-j-6B);
- We used [UL2](https://github.com/google-research/google-research/tree/master/ul2)'s training objective, allowing the model to see bidirectional context of the prompt;
- The model was trained on a large collection of diverse data, including [Chain-of-Thought (CoT)](https://ai.googleblog.com/2022/05/language-models-perform-reasoning-via.html), [Public Pool of Prompts (P3) dataset](https://huggingface.co/datasets/bigscience/P3), [Natural-Instructions (NI) dataset](https://github.com/allenai/natural-instructions).

With the help of techniques mentioned above, GPT-JT significantly improves the performance of classification tasks over the original GPT-J, and even outperforms most 100B+ parameter models!

# Quick Start

```python
from transformers import pipeline
pipe = pipeline(model='togethercomputer/GPT-JT-6B-v1')
pipe('''"I love this!" Is it positive? A:''')
```
or
```python
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("togethercomputer/GPT-JT-6B-v1")
model = AutoModelForCausalLM.from_pretrained("togethercomputer/GPT-JT-6B-v1")
```

# Training Details

## UL2 Training Objective

We train GPT-JT using UL2 training objective [1][2].
The original GPT-J uses causal mask (as shown below left) for autoregressive generation. So for each token, it can only see its previous context.
In order to fully leverage the context information, we continue to train GPT-J with UL2 training objectives, and uses causal mask with prefix (as shown below right) -- using bidirectional attention for the prompt / input and causal attention for token generation.
Intuitively, being able to see context bidirectionally might improve downstream tasks that require this information.

$$ 
\begin{bmatrix}
1 & 0 & 0 & 0 & 0 \\
1 & 1 & 0 & 0 & 0 \\
1 & 1 & 1 & 0 & 0 \\
1 & 1 & 1 & 1 & 0 \\
1 & 1 & 1 & 1 & 1 
\end{bmatrix}

\begin{bmatrix}
1 & 1 & 1 & 0 & 0 \\
1 & 1 & 1 & 0 & 0 \\
1 & 1 & 1 & 0 & 0 \\
1 & 1 & 1 & 1 & 0 \\
1 & 1 & 1 & 1 & 1 
\end{bmatrix}  
$$

Furthermore, we leverage a large collection of data, including [Natural-Instructions](https://github.com/allenai/natural-instructions), [P3](https://huggingface.co/datasets/Muennighoff/P3), [MMLU-COT](https://github.com/jasonwei20/flan-2/blob/main/mmlu-cot.json), and [the Pile](https://huggingface.co/datasets/the_pile)
Specifically, we first conduct training for 2.62 billion tokens using the UL2 loss on the Pile, followed by 0.92 billion tokens with a mixture of the above datasets: 5% of COT, 20% of P3, 20% of NI, and 55% of the Pile.

## Hyperparameters

We used AdamW with a learning rate of 1e-5 and global batch size of 64 (16 for each data parallel worker).
We used mix-precision training where the activation is in FP16 while the optimizer states are kept in FP32.
We use both data parallelism and pipeline parallelism to conduct training.
During training, we truncate the input sequence to 2048 tokens, and for input sequence that contains less than 2048 tokens, we concatenate multiple sequences into one long sequence to improve the data efficiency.

## Infrastructure

We used [the Together Research Computer](https://together.xyz/) to conduct training. 

# References

[1]: Tay, Yi, Mostafa Dehghani, Vinh Q. Tran, Xavier Garcia, Dara Bahri, Tal Schuster, Huaixiu Steven Zheng, Neil Houlsby, and Donald Metzler. "Unifying Language Learning Paradigms." arXiv preprint arXiv:2205.05131 (2022).

[2]: Tay, Yi, Jason Wei, Hyung Won Chung, Vinh Q. Tran, David R. So, Siamak Shakeri, Xavier Garcia et al. "Transcending scaling laws with 0.1% extra compute." arXiv preprint arXiv:2210.11399 (2022).