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/TinyBERT-spanish-uncased-finetuned-ner/README.md
README.md
ADDED
@@ -0,0 +1,106 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language: es
|
3 |
+
thumbnail:
|
4 |
+
---
|
5 |
+
|
6 |
+
# Spanish TinyBERT + NER
|
7 |
+
|
8 |
+
This model is a fine-tuned on [NER-C](https://www.kaggle.com/nltkdata/conll-corpora) of a [Spanish Tiny Bert](https://huggingface.co/mrm8488/es-tinybert-v1-1) model I created using *distillation* for **NER** downstream task. The **size** of the model is **55MB**
|
9 |
+
|
10 |
+
## Details of the downstream task (NER) - Dataset
|
11 |
+
|
12 |
+
- [Dataset: CONLL Corpora ES](https://www.kaggle.com/nltkdata/conll-corpora)
|
13 |
+
|
14 |
+
I preprocessed the dataset and split it as train / dev (80/20)
|
15 |
+
|
16 |
+
| Dataset | # Examples |
|
17 |
+
| ---------------------- | ----- |
|
18 |
+
| Train | 8.7 K |
|
19 |
+
| Dev | 2.2 K |
|
20 |
+
|
21 |
+
|
22 |
+
- [Fine-tune on NER script provided by Huggingface](https://github.com/huggingface/transformers/blob/master/examples/token-classification/run_ner_old.py)
|
23 |
+
|
24 |
+
- Labels covered:
|
25 |
+
|
26 |
+
```
|
27 |
+
B-LOC
|
28 |
+
B-MISC
|
29 |
+
B-ORG
|
30 |
+
B-PER
|
31 |
+
I-LOC
|
32 |
+
I-MISC
|
33 |
+
I-ORG
|
34 |
+
I-PER
|
35 |
+
O
|
36 |
+
```
|
37 |
+
|
38 |
+
## Metrics on evaluation set:
|
39 |
+
|
40 |
+
| Metric | # score |
|
41 |
+
| :------------------------------------------------------------------------------------: | :-------: |
|
42 |
+
| F1 | **70.00**
|
43 |
+
| Precision | **67.83** |
|
44 |
+
| Recall | **71.46** |
|
45 |
+
|
46 |
+
## Comparison:
|
47 |
+
|
48 |
+
| Model | # F1 score |Size(MB)|
|
49 |
+
| :--------------------------------------------------------------------------------------------------------------: | :-------: |:------|
|
50 |
+
| bert-base-spanish-wwm-cased (BETO) | 88.43 | 421
|
51 |
+
| [bert-spanish-cased-finetuned-ner](https://huggingface.co/mrm8488/bert-spanish-cased-finetuned-ner) | **90.17** | 420 |
|
52 |
+
| Best Multilingual BERT | 87.38 | 681 |
|
53 |
+
|TinyBERT-spanish-uncased-finetuned-ner (this one) | 70.00 | **55** |
|
54 |
+
|
55 |
+
## Model in action
|
56 |
+
|
57 |
+
|
58 |
+
Example of usage:
|
59 |
+
|
60 |
+
```python
|
61 |
+
import torch
|
62 |
+
from transformers import AutoModelForTokenClassification, AutoTokenizer
|
63 |
+
|
64 |
+
id2label = {
|
65 |
+
"0": "B-LOC",
|
66 |
+
"1": "B-MISC",
|
67 |
+
"2": "B-ORG",
|
68 |
+
"3": "B-PER",
|
69 |
+
"4": "I-LOC",
|
70 |
+
"5": "I-MISC",
|
71 |
+
"6": "I-ORG",
|
72 |
+
"7": "I-PER",
|
73 |
+
"8": "O"
|
74 |
+
}
|
75 |
+
|
76 |
+
tokenizer = AutoTokenizer.from_pretrained('mrm8488/TinyBERT-spanish-uncased-finetuned-ner')
|
77 |
+
model = AutoModelForTokenClassification.from_pretrained('mrm8488/TinyBERT-spanish-uncased-finetuned-ner')
|
78 |
+
text ="Mis amigos están pensando viajar a Londres este verano."
|
79 |
+
input_ids = torch.tensor(tokenizer.encode(text)).unsqueeze(0)
|
80 |
+
|
81 |
+
outputs = model(input_ids)
|
82 |
+
last_hidden_states = outputs[0]
|
83 |
+
|
84 |
+
for m in last_hidden_states:
|
85 |
+
for index, n in enumerate(m):
|
86 |
+
if(index > 0 and index <= len(text.split(" "))):
|
87 |
+
print(text.split(" ")[index-1] + ": " + id2label[str(torch.argmax(n).item())])
|
88 |
+
|
89 |
+
'''
|
90 |
+
Output:
|
91 |
+
--------
|
92 |
+
Mis: O
|
93 |
+
amigos: O
|
94 |
+
están: O
|
95 |
+
pensando: O
|
96 |
+
viajar: O
|
97 |
+
a: O
|
98 |
+
Londres: B-LOC
|
99 |
+
este: O
|
100 |
+
verano.: O
|
101 |
+
'''
|
102 |
+
```
|
103 |
+
|
104 |
+
> Created by [Manuel Romero/@mrm8488](https://twitter.com/mrm8488)
|
105 |
+
|
106 |
+
> Made with <span style="color: #e25555;">♥</span> in Spain
|