app py created
Browse files
app.py
ADDED
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
3 |
+
model_path = "core42/jais-13b"
|
4 |
+
|
5 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
6 |
+
|
7 |
+
tokenizer = AutoTokenizer.from_pretrained(model_path)
|
8 |
+
model = AutoModelForCausalLM.from_pretrained(model_path, device_map="auto", trust_remote_code=True)
|
9 |
+
|
10 |
+
|
11 |
+
def get_response(text,tokenizer=tokenizer,model=model):
|
12 |
+
input_ids = tokenizer(text, return_tensors="pt").input_ids
|
13 |
+
inputs = input_ids.to(device)
|
14 |
+
input_len = inputs.shape[-1]
|
15 |
+
generate_ids = model.generate(
|
16 |
+
inputs,
|
17 |
+
top_p=0.9,
|
18 |
+
temperature=0.3,
|
19 |
+
max_length=200-input_len,
|
20 |
+
min_length=input_len + 4,
|
21 |
+
repetition_penalty=1.2,
|
22 |
+
do_sample=True,
|
23 |
+
)
|
24 |
+
response = tokenizer.batch_decode(
|
25 |
+
generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=True
|
26 |
+
)[0]
|
27 |
+
return response
|
28 |
+
|
29 |
+
|
30 |
+
text= "عاصمة دولة الإمارات العربية المتحدة ه"
|
31 |
+
print(get_response(text))
|
32 |
+
|
33 |
+
text = "The capital of UAE is"
|
34 |
+
print(get_response(text))
|