Migrate model card from transformers-repo
Browse filesRead announcement at https://discuss.huggingface.co/t/announcement-all-model-cards-will-be-migrated-to-hf-co-model-repos/2755
Original file history: https://github.com/huggingface/transformers/commits/master/model_cards/mrm8488/GPT-2-finetuned-common_gen/README.md
README.md
ADDED
@@ -0,0 +1,63 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language: en
|
3 |
+
datasets:
|
4 |
+
- common_gen
|
5 |
+
widget:
|
6 |
+
- text: "<|endoftext|> apple, tree, pick:"
|
7 |
+
---
|
8 |
+
|
9 |
+
# GPT-2 fine-tuned on CommonGen
|
10 |
+
|
11 |
+
[GPT-2](https://huggingface.co/gpt2) fine-tuned on [CommonGen](https://inklab.usc.edu/CommonGen/index.html) for *Generative Commonsense Reasoning*.
|
12 |
+
|
13 |
+
## Details of GPT-2
|
14 |
+
|
15 |
+
GPT-2 is a transformers model pretrained on a very large corpus of English data in a self-supervised fashion. This
|
16 |
+
means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots
|
17 |
+
of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely,
|
18 |
+
it was trained to guess the next word in sentences.
|
19 |
+
|
20 |
+
More precisely, inputs are sequences of continuous text of a certain length and the targets are the same sequence,
|
21 |
+
shifted one token (word or piece of word) to the right. The model uses internally a mask-mechanism to make sure the
|
22 |
+
predictions for the token `i` only uses the inputs from `1` to `i` but not the future tokens.
|
23 |
+
|
24 |
+
This way, the model learns an inner representation of the English language that can then be used to extract features
|
25 |
+
useful for downstream tasks. The model is best at what it was pretrained for however, which is generating texts from a
|
26 |
+
prompt.
|
27 |
+
|
28 |
+
|
29 |
+
## Details of the dataset π
|
30 |
+
|
31 |
+
CommonGen is a constrained text generation task, associated with a benchmark dataset, to explicitly test machines for the ability of generative commonsense reasoning. Given a set of common concepts; the task is to generate a coherent sentence describing an everyday scenario using these concepts.
|
32 |
+
|
33 |
+
CommonGen is challenging because it inherently requires 1) relational reasoning using background commonsense knowledge, and 2) compositional generalization ability to work on unseen concept combinations. Our dataset, constructed through a combination of crowd-sourcing from AMT and existing caption corpora, consists of 30k concept-sets and 50k sentences in total.
|
34 |
+
|
35 |
+
|
36 |
+
| Dataset | Split | # samples |
|
37 |
+
| -------- | ----- | --------- |
|
38 |
+
| common_gen | train | 67389 |
|
39 |
+
| common_gen | valid | 4018 |
|
40 |
+
| common_gen | test | 1497 |
|
41 |
+
|
42 |
+
|
43 |
+
## Model fine-tuning ποΈβ
|
44 |
+
|
45 |
+
You can find the fine-tuning script [here](https://github.com/huggingface/transformers/tree/master/examples/language-modeling)
|
46 |
+
|
47 |
+
## Model in Action π
|
48 |
+
|
49 |
+
```bash
|
50 |
+
python ./transformers/examples/text-generation/run_generation.py \
|
51 |
+
--model_type=gpt2 \
|
52 |
+
--model_name_or_path="mrm8488/GPT-2-finetuned-common_gen" \
|
53 |
+
--num_return_sequences 1 \
|
54 |
+
--prompt "<|endoftext|> kid, room, dance:" \
|
55 |
+
--stop_token "."
|
56 |
+
```
|
57 |
+
|
58 |
+
> Created by [Manuel Romero/@mrm8488](https://twitter.com/mrm8488) | [LinkedIn](https://www.linkedin.com/in/manuel-romero-cs/)
|
59 |
+
|
60 |
+
> Made with <span style="color: #e25555;">♥</span> in Spain
|
61 |
+
|
62 |
+
|
63 |
+
|