-
Notifications
You must be signed in to change notification settings - Fork 0
/
run_inference.py
57 lines (44 loc) · 1.88 KB
/
run_inference.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
# --input_text: str
# --output_text: str
# --model: Fine tuned Model
# --config: GenerationConfig
import torch
import argparse
from transformers import (
GenerationConfig,
T5ForConditionalGeneration,
AutoTokenizer
)
def inference(model, tokenizer, input_text: str, generation_config):
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
prompt = f"""Generate Python code from the following instruction:
### Instruction: {input_text}
### Response:"""
print("\n[+] Input:\n", input_text)
model.to(device)
input_ids = tokenizer(prompt, return_tensors="pt").to(device)
output_ids = model.generate(input_ids, do_sample=True, generation_config=generation_config)
output_text = tokenizer.decode(output_ids[0], skip_special_tokens=True)
print("\n[+] Output:\n", output_text)
return output_text
if __name__=="__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--checkpoint", type=str, default="dtruong46me/codet5p-770m-2")
parser.add_argument("--input_text", type=str, default="Print Hello World")
parser.add_argument("--temperature", type=float, default=0.9)
parser.add_argument("--top_k", type=int, default=40)
parser.add_argument("--top_p", type=float, default=1.0)
parser.add_argument("--max_new_tokens", type=int, default=128)
parser.add_argument("--min_new_tokens", type=int, default=8)
args = parser.parse_args()
tokenizer = AutoTokenizer.from_pretrained(args.checkpoint)
model = T5ForConditionalGeneration.from_pretrained(args.checkpoint)
generation_config = GenerationConfig(
temperature=args.temperature,
top_k=args.top_k,
top_p=args.top_p,
max_new_tokens=args.max_new_tokens,
min_new_tokens=args.min_new_tokens
)
response = inference(model, tokenizer, args.input_text, generation_config)
print("\n[+] Response:", response)