Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# roberta-php
|
2 |
+
---
|
3 |
+
language: php
|
4 |
+
datasets:
|
5 |
+
- CodeSearchNet
|
6 |
+
---
|
7 |
+
|
8 |
+
This is a [roberta](https://arxiv.org/pdf/1907.11692.pdf) pre-trained version on the [CodeSearchNet dataset](https://github.com/github/CodeSearchNet) for **php** Mask Language Model mission.
|
9 |
+
|
10 |
+
To load the model:
|
11 |
+
(necessary packages: !pip install transformers sentencepiece)
|
12 |
+
```php
|
13 |
+
from transformers import AutoTokenizer, AutoModelWithLMHead, pipeline
|
14 |
+
tokenizer = AutoTokenizer.from_pretrained("dbernsohn/roberta-php")
|
15 |
+
model = AutoModelWithLMHead.from_pretrained("dbernsohn/roberta-php")
|
16 |
+
|
17 |
+
fill_mask = pipeline(
|
18 |
+
"fill-mask",
|
19 |
+
model=model,
|
20 |
+
tokenizer=tokenizer
|
21 |
+
)
|
22 |
+
```
|
23 |
+
|
24 |
+
You can then use this model to fill masked words in a Java code.
|
25 |
+
|
26 |
+
```php
|
27 |
+
code = """
|
28 |
+
$people = array(
|
29 |
+
array('name' => 'Kalle', 'salt' => 856412),
|
30 |
+
array('name' => 'Pierre', 'salt' => 215863)
|
31 |
+
);
|
32 |
+
|
33 |
+
for($i = 0; $i < count($<mask>); ++$i) {
|
34 |
+
$people[$i]['salt'] = mt_rand(000000, 999999);
|
35 |
+
}
|
36 |
+
""".lstrip()
|
37 |
+
|
38 |
+
pred = {x["token_str"].replace("Ġ", ""): x["score"] for x in fill_mask(code)}
|
39 |
+
sorted(pred.items(), key=lambda kv: kv[1], reverse=True)
|
40 |
+
# [('people', 0.785636842250824),
|
41 |
+
# ('parts', 0.006270722020417452),
|
42 |
+
# ('id', 0.0035842324141412973),
|
43 |
+
# ('data', 0.0025512021966278553),
|
44 |
+
# ('config', 0.002258970635011792)]
|
45 |
+
```
|
46 |
+
|
47 |
+
The whole training process and hyperparameters are in my [GitHub repo](https://github.com/DorBernsohn/CodeLM/tree/main/CodeMLM)
|
48 |
+
|
49 |
+
> Created by [Dor Bernsohn](https://www.linkedin.com/in/dor-bernsohn-70b2b1146/)
|