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/neurocode/IsRoBERTa/README.md
README.md
ADDED
@@ -0,0 +1,74 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language: is
|
3 |
+
datasets:
|
4 |
+
- Icelandic portion of the OSCAR corpus from INRIA
|
5 |
+
- oscar
|
6 |
+
---
|
7 |
+
|
8 |
+
# IsRoBERTa a RoBERTa-like masked language model
|
9 |
+
|
10 |
+
Probably the first icelandic transformer language model!
|
11 |
+
|
12 |
+
## Overview
|
13 |
+
**Language:** Icelandic
|
14 |
+
**Downstream-task:** masked-lm
|
15 |
+
**Training data:** OSCAR corpus
|
16 |
+
**Code:** See [here](https://github.com/neurocode-io/icelandic-language-model)
|
17 |
+
**Infrastructure**: 1x Nvidia K80
|
18 |
+
|
19 |
+
## Hyperparameters
|
20 |
+
|
21 |
+
```
|
22 |
+
per_device_train_batch_size = 48
|
23 |
+
n_epochs = 1
|
24 |
+
vocab_size = 52.000
|
25 |
+
max_position_embeddings = 514
|
26 |
+
num_attention_heads = 12
|
27 |
+
num_hidden_layers = 6
|
28 |
+
type_vocab_size = 1
|
29 |
+
learning_rate=0.00005
|
30 |
+
```
|
31 |
+
|
32 |
+
|
33 |
+
## Usage
|
34 |
+
|
35 |
+
### In Transformers
|
36 |
+
```python
|
37 |
+
from transformers import (
|
38 |
+
pipeline,
|
39 |
+
AutoTokenizer,
|
40 |
+
AutoModelWithLMHead
|
41 |
+
)
|
42 |
+
|
43 |
+
model_name = "neurocode/IsRoBERTa"
|
44 |
+
|
45 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
46 |
+
model = AutoModelWithLMHead.from_pretrained(model_name)
|
47 |
+
>>> fill_mask = pipeline(
|
48 |
+
... "fill-mask",
|
49 |
+
... model=model,
|
50 |
+
... tokenizer=tokenizer
|
51 |
+
... )
|
52 |
+
>>> result = fill_mask("Hann fór út að <mask>.")
|
53 |
+
>>> result
|
54 |
+
[
|
55 |
+
{'sequence': '<s>Hann fór út að nýju.</s>', 'score': 0.03395755589008331, 'token': 2219, 'token_str': 'Ġnýju'},
|
56 |
+
{'sequence': '<s>Hann fór út að undanförnu.</s>', 'score': 0.029087543487548828, 'token': 7590, 'token_str': 'Ġundanförnu'},
|
57 |
+
{'sequence': '<s>Hann fór út að lokum.</s>', 'score': 0.024420788511633873, 'token': 4384, 'token_str': 'Ġlokum'},
|
58 |
+
{'sequence': '<s>Hann fór út að þessu.</s>', 'score': 0.021231256425380707, 'token': 921, 'token_str': 'Ġþessu'},
|
59 |
+
{'sequence': '<s>Hann fór út að honum.</s>', 'score': 0.0205782949924469, 'token': 1136, 'token_str': 'Ġhonum'}
|
60 |
+
]
|
61 |
+
```
|
62 |
+
|
63 |
+
|
64 |
+
## Authors
|
65 |
+
Bobby Donchev: `contact [at] donchev.is`
|
66 |
+
Elena Cramer: `elena.cramer [at] neurocode.io`
|
67 |
+
|
68 |
+
## About us
|
69 |
+
|
70 |
+
We bring AI software for our customers live
|
71 |
+
Our focus: AI software development
|
72 |
+
|
73 |
+
Get in touch:
|
74 |
+
[LinkedIn](https://de.linkedin.com/company/neurocodeio) | [Website](https://neurocode.io)
|