Update README.md
Browse files
README.md
CHANGED
@@ -1,11 +1,39 @@
|
|
1 |
---
|
2 |
license: bigscience-openrail-m
|
3 |
widget:
|
4 |
-
- text: M[MASK]
|
5 |
datasets:
|
6 |
- Ensembl
|
7 |
pipeline_tag: fill-mask
|
8 |
tags:
|
9 |
- biology
|
10 |
- medical
|
11 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: bigscience-openrail-m
|
3 |
widget:
|
4 |
+
- text: M[MASK]LWMRLLPLLALLALWGPDPAAAFVNQHLCGSHLVEALYLVCGERGFFYTPKTRREAEDLQVGQVELGGGPGAGSLQPLALEGSLQKRGIVEQCCTSICSLYQLENYCN
|
5 |
datasets:
|
6 |
- Ensembl
|
7 |
pipeline_tag: fill-mask
|
8 |
tags:
|
9 |
- biology
|
10 |
- medical
|
11 |
+
---
|
12 |
+
|
13 |
+
# BERT base for proteins
|
14 |
+
This is bidirectional transformer pretrained on amino-acid sequences of human proteins.
|
15 |
+
|
16 |
+
Example: Insulin (P01308)
|
17 |
+
```
|
18 |
+
MALWMRLLPLLALLALWGPDPAAAFVNQHLCGSHLVEALYLVCGERGFFYTPKTRREAEDLQVGQVELGGGPGAGSLQPLALEGSLQKRGIVEQCCTSICSLYQLENYCN
|
19 |
+
```
|
20 |
+
|
21 |
+
The model was trained using the masked-language-modeling objective.
|
22 |
+
|
23 |
+
## Intended uses
|
24 |
+
This model is primarily aimed at being fine-tuned on the following tasks:
|
25 |
+
- protein function
|
26 |
+
- molecule-to-gene-expression mapping
|
27 |
+
- cell targeting
|
28 |
+
|
29 |
+
## How to use in your code
|
30 |
+
```python
|
31 |
+
from transformers import BertTokenizerFast, BertModel
|
32 |
+
checkpoint = 'unikei/bert-base-proteins'
|
33 |
+
tokenizer = BertTokenizerFast.from_pretrained(checkpoint)
|
34 |
+
model = BertModel.from_pretrained(checkpoint)
|
35 |
+
|
36 |
+
example = 'MALWMRLLPLLALLALWGPDPAAAFVNQHLCGSHLVEALYLVCGERGFFYTPKTRREAEDLQVGQVELGGGPGAGSLQPLALEGSLQKRGIVEQCCTSICSLYQLENYCN'
|
37 |
+
tokens = tokenizer(example, return_tensors='pt')
|
38 |
+
predictions = model(**tokens)
|
39 |
+
```
|