readme: add initial version
Browse files
README.md
ADDED
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language: de
|
3 |
+
|
4 |
+
widget:
|
5 |
+
- text: "Schon um die Liebe"
|
6 |
+
|
7 |
+
license: mit
|
8 |
+
---
|
9 |
+
|
10 |
+
# German GPT-2 model
|
11 |
+
|
12 |
+
In this repository we release (yet another) GPT-2 model, that was trained on various texts for German.
|
13 |
+
|
14 |
+
The model is meant to be an entry point for fine-tuning on other texts, and it is definitely not as good or "dangerous" as the English GPT-3 model. We do not plan extensive PR or staged releases for this model 😉
|
15 |
+
|
16 |
+
**Note**: The model was initially released under an anonymous alias (`anonymous-german-nlp/german-gpt2`) so we now "de-anonymize" it.
|
17 |
+
|
18 |
+
More details about GPT-2 can be found in the great [Hugging Face](https://huggingface.co/transformers/model_doc/gpt2.html) documentation.
|
19 |
+
|
20 |
+
## German GPT-2 fine-tuned on Faust Faust I and II
|
21 |
+
|
22 |
+
We fine-tuned our German GPT-2 model on "Faust I and II" from Johann Wolfgang Goethe. These texts can be obtained from [Deutsches Textarchiv (DTA)](http://www.deutschestextarchiv.de/book/show/goethe_faust01_1808). We use the "normalized" version of both texts (to avoid out-of-vocabulary problems with e.g. "ſ")
|
23 |
+
|
24 |
+
Fine-Tuning was done for 100 epochs, using a batch size of 4 with half precision on a RTX 3090. Total time was around 12 minutes (it is really fast!).
|
25 |
+
|
26 |
+
We also open source this fine-tuned model. Text can be generated with:
|
27 |
+
|
28 |
+
```python
|
29 |
+
from transformers import pipeline
|
30 |
+
|
31 |
+
pipe = pipeline('text-generation', model="dbmdz/german-gpt2-faust",
|
32 |
+
tokenizer="dbmdz/german-gpt2-faust", config={'max_length':800})
|
33 |
+
|
34 |
+
text = pipe2("Schon um die Liebe")[0]["generated_text"]
|
35 |
+
|
36 |
+
print(text)
|
37 |
+
```
|
38 |
+
|
39 |
+
and could output:
|
40 |
+
|
41 |
+
```
|
42 |
+
Schon um die Liebe bitte ich, Herr! Wer mag sich die dreifach Ermächtigen?
|
43 |
+
Sei mir ein Held!
|
44 |
+
Und daß die Stunde kommt spreche ich nicht aus.
|
45 |
+
Faust (schaudernd).
|
46 |
+
Den schönen Boten finde' ich verwirrend;
|
47 |
+
```
|
48 |
+
|
49 |
+
# License
|
50 |
+
|
51 |
+
All models are licensed under [MIT](LICENSE).
|
52 |
+
|
53 |
+
# Huggingface model hub
|
54 |
+
|
55 |
+
All models are available on the [Huggingface model hub](https://huggingface.co/dbmdz).
|
56 |
+
|
57 |
+
# Contact (Bugs, Feedback, Contribution and more)
|
58 |
+
|
59 |
+
For questions about our BERT models just open an issue
|
60 |
+
[here](https://github.com/stefan-it/german-gpt/issues/new) 🤗
|
61 |
+
|
62 |
+
# Acknowledgments
|
63 |
+
|
64 |
+
Research supported with Cloud TPUs from Google's TensorFlow Research Cloud (TFRC).
|
65 |
+
Thanks for providing access to the TFRC ❤️
|
66 |
+
|
67 |
+
Thanks to the generous support from the [Hugging Face](https://huggingface.co/) team,
|
68 |
+
it is possible to download both cased and uncased models from their S3 storage 🤗
|