Update model card
Browse files
README.md
CHANGED
@@ -1,8 +1,10 @@
|
|
1 |
---
|
2 |
language: nl
|
3 |
widget:
|
4 |
-
- text: "
|
5 |
-
- text: "
|
|
|
|
|
6 |
tags:
|
7 |
- gpt2-medium
|
8 |
- gpt2
|
@@ -12,13 +14,23 @@ datasets:
|
|
12 |
---
|
13 |
# GPT2-Medium pre-trained on cleaned Dutch mC4 🇳🇱
|
14 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
Training details:
|
16 |
|
17 |
-
*
|
18 |
-
*
|
19 |
-
*
|
20 |
-
*
|
21 |
-
*
|
22 |
|
23 |
Work in progress. Dec 2021.
|
24 |
|
|
|
1 |
---
|
2 |
language: nl
|
3 |
widget:
|
4 |
+
- text: "In het jaar 2030 zullen we"
|
5 |
+
- text: "Toen ik gisteren volledig in de ban was van"
|
6 |
+
- text: "Studenten en leraren van de Bogazici Universiteit in de Turkse stad Istanbul"
|
7 |
+
- text: "In Israël was een strenge lockdown"
|
8 |
tags:
|
9 |
- gpt2-medium
|
10 |
- gpt2
|
|
|
14 |
---
|
15 |
# GPT2-Medium pre-trained on cleaned Dutch mC4 🇳🇱
|
16 |
|
17 |
+
Dataset:
|
18 |
+
|
19 |
+
* [mC4 NL Cleaned](https://huggingface.co/datasets/yhavinga/mc4_nl_cleaned)
|
20 |
+
* dataset split: full (33B tokens)
|
21 |
+
|
22 |
+
Tokenizer:
|
23 |
+
|
24 |
+
* New tokenizer trained on mC4 with the scripts from the Huggingface
|
25 |
+
Transformers [Flax examples](https://github.com/huggingface/transformers/tree/master/examples/flax/language-modeling)
|
26 |
+
|
27 |
Training details:
|
28 |
|
29 |
+
* Trained for 240k steps (29 dec 2021)
|
30 |
+
* Block size: 512
|
31 |
+
* Optimizer: adam, lr 8e-4, beta1 0.9, beta2 0.98
|
32 |
+
* Warmup steps: 5000
|
33 |
+
* Weight decay: 0.01
|
34 |
|
35 |
Work in progress. Dec 2021.
|
36 |
|