Add t5predictor to app.py
Browse files- app.py +89 -0
- requirements.txt +3 -0
app.py
ADDED
@@ -0,0 +1,89 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import re
|
2 |
+
|
3 |
+
import gradio as gr
|
4 |
+
import torch
|
5 |
+
from transformers import T5ForConditionalGeneration, RobertaTokenizer
|
6 |
+
|
7 |
+
|
8 |
+
tokenizer = RobertaTokenizer.from_pretrained("mamiksik/CommitPredictorT5PL", revision="fb08d01")
|
9 |
+
model = T5ForConditionalGeneration.from_pretrained("mamiksik/CommitPredictorT5PL", revision="fb08d01")
|
10 |
+
|
11 |
+
|
12 |
+
def parse_files(accumulator: list[str], patch: str):
|
13 |
+
lines = patch.splitlines()
|
14 |
+
|
15 |
+
filename_before = None
|
16 |
+
for line in lines:
|
17 |
+
if line.startswith("index") or line.startswith("diff"):
|
18 |
+
continue
|
19 |
+
if line.startswith("---"):
|
20 |
+
filename_before = line.split(" ", 1)[1][1:]
|
21 |
+
continue
|
22 |
+
|
23 |
+
if line.startswith("+++"):
|
24 |
+
filename_after = line.split(" ", 1)[1][1:]
|
25 |
+
|
26 |
+
if filename_before == filename_after:
|
27 |
+
accumulator.append(f"<ide><path>{filename_before}")
|
28 |
+
else:
|
29 |
+
accumulator.append(f"<add><path>{filename_after}")
|
30 |
+
accumulator.append(f"<del><path>{filename_before}")
|
31 |
+
continue
|
32 |
+
|
33 |
+
line = re.sub("@@[^@@]*@@", "", line)
|
34 |
+
if len(line) == 0:
|
35 |
+
continue
|
36 |
+
|
37 |
+
if line[0] == "+":
|
38 |
+
line = line.replace("+", "<add>", 1)
|
39 |
+
elif line[0] == "-":
|
40 |
+
line = line.replace("-", "<del>", 1)
|
41 |
+
else:
|
42 |
+
line = f"<ide>{line}"
|
43 |
+
|
44 |
+
accumulator.append(line)
|
45 |
+
|
46 |
+
return accumulator
|
47 |
+
|
48 |
+
|
49 |
+
def predict(patch, max_length, min_length, num_beams, prediction_count):
|
50 |
+
accumulator = []
|
51 |
+
parse_files(accumulator, patch)
|
52 |
+
input_text = '\n'.join(accumulator)
|
53 |
+
|
54 |
+
with torch.no_grad():
|
55 |
+
token_count = tokenizer(input_text, return_tensors="pt").input_ids.shape[1]
|
56 |
+
|
57 |
+
input_ids = tokenizer(
|
58 |
+
input_text,
|
59 |
+
truncation=True,
|
60 |
+
padding=True,
|
61 |
+
return_tensors="pt",
|
62 |
+
).input_ids
|
63 |
+
|
64 |
+
outputs = model.generate(
|
65 |
+
input_ids,
|
66 |
+
max_length=max_length,
|
67 |
+
min_length=min_length,
|
68 |
+
num_beams=num_beams,
|
69 |
+
num_return_sequences=prediction_count,
|
70 |
+
)
|
71 |
+
|
72 |
+
result = tokenizer.batch_decode(outputs, skip_special_tokens=True)
|
73 |
+
return token_count, '\n'.join(accumulator), {k: 0 for k in result}
|
74 |
+
|
75 |
+
|
76 |
+
iface = gr.Interface(fn=predict, inputs=[
|
77 |
+
gr.Textbox(label="Patch (as generated by git diff)"),
|
78 |
+
gr.Slider(1, 128, value=20, label="Max message length"),
|
79 |
+
gr.Slider(1, 128, value=5, label="Min message length"),
|
80 |
+
gr.Slider(1, 10, value=7, label="Number of beams"),
|
81 |
+
gr.Slider(1, 15, value=5, label="Number of predictions"),
|
82 |
+
], outputs=[
|
83 |
+
gr.Textbox(label="Token count"),
|
84 |
+
gr.Textbox(label="Parsed patch"),
|
85 |
+
gr.Label(label="Predictions")
|
86 |
+
])
|
87 |
+
|
88 |
+
if __name__ == "__main__":
|
89 |
+
iface.launch()
|
requirements.txt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
gradio~=3.16.2
|
2 |
+
transformers~=4.25.1
|
3 |
+
torch~=1.13.1
|