Kunal Tangri commited on
Commit
dbb3415
1 Parent(s): 2574ba4

Add model card

Browse files
Files changed (1) hide show
  1. README.md +71 -0
README.md ADDED
@@ -0,0 +1,71 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ tags:
5
+ - text generation
6
+ - pytorch
7
+ - the Pile
8
+ - causal-lm
9
+ license: apache-2.0
10
+ datasets:
11
+ - the Pile
12
+ ---
13
+ # GPT-Neo 2.7B (By EleutherAI)
14
+ ## Model Description
15
+ GPT-Neo 2.7B is a transformer model designed using EleutherAI's replication of the GPT-3 architecture. GPT-Neo refers to the class of models, while 2.7B represents the number of parameters of this particular pre-trained model.
16
+ ## Training data
17
+ GPT-Neo 2.7B was trained on the Pile, a large scale curated dataset created by EleutherAI for the purpose of training this model.
18
+ ## Training procedure
19
+ This model was trained for 420 billion tokens over 400,000 steps. It was trained as a masked autoregressive language model, using cross-entropy loss.
20
+ ## Intended Use and Limitations
21
+ This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks. The model is best at what it was pretrained for however, which is generating texts from a prompt.
22
+ ### How to use
23
+ You can use this model directly with a pipeline for text generation. This example generates a different sequence each time it's run:
24
+ ```py
25
+ >>> from transformers import pipeline
26
+ >>> generator = pipeline('text-generation', model='EleutherAI/gpt-neo-2.7B')
27
+ >>> generator("EleutherAI has", do_sample=True, min_length=50)
28
+ [{'generated_text': 'EleutherAI has made a commitment to create new software packages for each of its major clients and has'}]
29
+ ```
30
+ ### Limitations and Biases
31
+ GPT-Neo was trained as an autoregressive language model. This means that its core functionality is taking a string of text and predicting the next token. While language models are widely used for tasks other than this, there are a lot of unknowns with this work.
32
+ GPT-Neo was trained on the Pile, a dataset known to contain profanity, lewd, and otherwise abrasive language. Depending on your usecase GPT-Neo may produce socially unacceptable text. See Sections 5 and 6 of the Pile paper for a more detailed analysis of the biases in the Pile.
33
+ As with all language models, it is hard to predict in advance how GPT-Neo will respond to particular prompts and offensive content may occur without warning. We recommend having a human curate or filter the outputs before releasing them, both to censor undesirable content and to improve the quality of the results.
34
+ ## Eval results
35
+ All evaluations were done using our [evaluation harness](https://github.com/EleutherAI/lm-evaluation-harness). Some results for GPT-2 and GPT-3 are inconsistent with the values reported in the respective papers. We are currently looking into why, and would greatly appreciate feedback and further testing of our eval harness. If you would like to contribute evaluations you have done, please reach out on our [Discord](https://discord.gg/vtRgjbM).
36
+ ### Linguistic Reasoning
37
+ | Model and Size | Pile BPB | Pile PPL | Wikitext PPL | Lambada PPL | Lambada Acc | Winogrande | Hellaswag |
38
+ | ---------------- | ---------- | ---------- | ------------- | ----------- | ----------- | ---------- | ----------- |
39
+ | GPT-Neo 1.3B | 0.7527 | 6.159 | 13.10 | 7.498 | 57.23% | 55.01% | 38.66% |
40
+ | GPT-2 1.5B | 1.0468 | ----- | 17.48 | 10.634 | 51.21% | 59.40% | 40.03% |
41
+ | **GPT-Neo 2.7B** | **0.7165** | **5.646** | **11.39** | **5.626** | **62.22%** | **56.50%** | **42.73%** |
42
+ | GPT-3 Ada | 0.9631 | ----- | ----- | 9.954 | 51.60% | 52.90% | 35.93% |
43
+ ### Physical and Scientific Reasoning
44
+ | Model and Size | MathQA | PubMedQA | Piqa |
45
+ | ---------------- | ---------- | ---------- | ----------- |
46
+ | GPT-Neo 1.3B | 24.05% | 54.40% | 71.11% |
47
+ | GPT-2 1.5B | 23.64% | 58.33% | 70.78% |
48
+ | **GPT-Neo 2.7B** | **24.72%** | **57.54%** | **72.14%** |
49
+ | GPT-3 Ada | 24.29% | 52.80% | 68.88% |
50
+ ### Down-Stream Applications
51
+ TBD
52
+ ### BibTeX entry and citation info
53
+ To cite this model, use
54
+ ```bibtex
55
+ @article{gao2020pile,
56
+ title={The Pile: An 800GB Dataset of Diverse Text for Language Modeling},
57
+ author={Gao, Leo and Biderman, Stella and Black, Sid and Golding, Laurence and Hoppe, Travis and Foster, Charles and Phang, Jason and He, Horace and Thite, Anish and Nabeshima, Noa and others},
58
+ journal={arXiv preprint arXiv:2101.00027},
59
+ year={2020}
60
+ }
61
+ ```
62
+ To cite the codebase that this model was trained with, use
63
+ ```bibtex
64
+ @software{gpt-neo,
65
+ author = {Black, Sid and Gao, Leo and Wang, Phil and Leahy, Connor and Biderman, Stella},
66
+ title = {{GPT-Neo}: Large Scale Autoregressive Language Modeling with Mesh-Tensorflow},
67
+ url = {http://github.com/eleutherai/gpt-neo},
68
+ version = {1.0},
69
+ year = {2021},
70
+ }
71
+ ```