dipesh1701
commited on
Commit
•
463444e
1
Parent(s):
997a555
init
Browse files
app.py
ADDED
@@ -0,0 +1,96 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import torch
|
3 |
+
import gradio as gr
|
4 |
+
import time
|
5 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
|
6 |
+
from flores200_codes import flores_codes
|
7 |
+
|
8 |
+
|
9 |
+
def load_models():
|
10 |
+
# build model and tokenizer
|
11 |
+
model_name_dict = {
|
12 |
+
"nllb-distilled-600M": "facebook/nllb-200-distilled-600M",
|
13 |
+
#'nllb-1.3B': 'facebook/nllb-200-1.3B',
|
14 |
+
#'nllb-distilled-1.3B': 'facebook/nllb-200-distilled-1.3B',
|
15 |
+
#'nllb-3.3B': 'facebook/nllb-200-3.3B',
|
16 |
+
}
|
17 |
+
|
18 |
+
model_dict = {}
|
19 |
+
|
20 |
+
for call_name, real_name in model_name_dict.items():
|
21 |
+
print("\tLoading model: %s" % call_name)
|
22 |
+
model = AutoModelForSeq2SeqLM.from_pretrained(real_name)
|
23 |
+
tokenizer = AutoTokenizer.from_pretrained(real_name)
|
24 |
+
model_dict[call_name + "_model"] = model
|
25 |
+
model_dict[call_name + "_tokenizer"] = tokenizer
|
26 |
+
|
27 |
+
return model_dict
|
28 |
+
|
29 |
+
|
30 |
+
def translation(source, target, text):
|
31 |
+
if len(model_dict) == 2:
|
32 |
+
model_name = "nllb-distilled-600M"
|
33 |
+
|
34 |
+
start_time = time.time()
|
35 |
+
source = flores_codes[source]
|
36 |
+
target = flores_codes[target]
|
37 |
+
|
38 |
+
model = model_dict[model_name + "_model"]
|
39 |
+
tokenizer = model_dict[model_name + "_tokenizer"]
|
40 |
+
|
41 |
+
translator = pipeline(
|
42 |
+
"translation",
|
43 |
+
model=model,
|
44 |
+
tokenizer=tokenizer,
|
45 |
+
src_lang=source,
|
46 |
+
tgt_lang=target,
|
47 |
+
)
|
48 |
+
output = translator(text, max_length=400)
|
49 |
+
|
50 |
+
end_time = time.time()
|
51 |
+
|
52 |
+
output = output[0]["translation_text"]
|
53 |
+
result = {
|
54 |
+
"inference_time": end_time - start_time,
|
55 |
+
"source": source,
|
56 |
+
"target": target,
|
57 |
+
"result": output,
|
58 |
+
}
|
59 |
+
return result
|
60 |
+
|
61 |
+
|
62 |
+
if __name__ == "__main__":
|
63 |
+
print("\tinit models")
|
64 |
+
|
65 |
+
global model_dict
|
66 |
+
|
67 |
+
model_dict = load_models()
|
68 |
+
|
69 |
+
# define gradio demo
|
70 |
+
lang_codes = list(flores_codes.keys())
|
71 |
+
# inputs = [gr.inputs.Radio(['nllb-distilled-600M', 'nllb-1.3B', 'nllb-distilled-1.3B'], label='NLLB Model'),
|
72 |
+
inputs = [
|
73 |
+
gr.inputs.Dropdown(lang_codes, default="English", label="Source"),
|
74 |
+
gr.inputs.Dropdown(lang_codes, default="Korean", label="Target"),
|
75 |
+
gr.inputs.Textbox(lines=5, label="Input text"),
|
76 |
+
]
|
77 |
+
|
78 |
+
outputs = gr.outputs.JSON()
|
79 |
+
|
80 |
+
title = "NLLB distilled 600M demo"
|
81 |
+
|
82 |
+
demo_status = "Demo is running on CPU"
|
83 |
+
description = (
|
84 |
+
f"Details: https://github.com/facebookresearch/fairseq/tree/nllb. {demo_status}"
|
85 |
+
)
|
86 |
+
examples = [["English", "Korean", "Hi. nice to meet you"]]
|
87 |
+
|
88 |
+
gr.Interface(
|
89 |
+
translation,
|
90 |
+
inputs,
|
91 |
+
outputs,
|
92 |
+
title=title,
|
93 |
+
description=description,
|
94 |
+
examples=examples,
|
95 |
+
examples_per_page=50,
|
96 |
+
).launch()
|