Spaces:
Runtime error
Runtime error
Improve demo
Browse files
app.py
CHANGED
@@ -1,50 +1,64 @@
|
|
|
|
|
|
1 |
from transformers import T5ForConditionalGeneration, T5Tokenizer, GenerationConfig
|
2 |
import gradio as gr
|
3 |
|
4 |
MODEL_NAME = "jbochi/madlad400-3b-mt"
|
5 |
|
6 |
-
|
7 |
-
default_max_length = 200
|
8 |
-
|
9 |
-
print("Using `{}`.".format(MODEL_NAME))
|
10 |
-
|
11 |
tokenizer = T5Tokenizer.from_pretrained(MODEL_NAME)
|
12 |
-
print("
|
13 |
-
|
14 |
model = T5ForConditionalGeneration.from_pretrained(MODEL_NAME, device_map="auto")
|
15 |
-
print("T5ForConditionalGeneration loaded from pretrained.")
|
16 |
|
17 |
|
18 |
-
def inference(
|
19 |
-
|
|
|
|
|
|
|
|
|
20 |
outputs = model.generate(
|
21 |
-
input_ids=input_ids,
|
22 |
-
generation_config=GenerationConfig(max_length=max_length
|
23 |
)
|
24 |
result = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
25 |
-
|
26 |
-
|
27 |
-
|
|
|
|
|
|
|
|
|
|
|
28 |
|
29 |
-
with gr.Blocks() as demo:
|
30 |
-
with gr.Row():
|
31 |
-
gr.Markdown(
|
32 |
-
"<h1>Demo of {}</h1><p>See more at Hugging Face: <a href='https://huggingface.co/{}'>{}</a>.</p>".format(
|
33 |
-
MODEL_NAME, MODEL_NAME, MODEL_NAME
|
34 |
-
)
|
35 |
-
)
|
36 |
-
max_length = gr.Number(
|
37 |
-
value=default_max_length, label="maximum length of response"
|
38 |
-
)
|
39 |
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
47 |
|
48 |
-
txt.submit(fn=inference, inputs=[max_length, txt, state], outputs=[chatbot, state])
|
49 |
|
50 |
-
|
|
|
|
1 |
+
import time
|
2 |
+
|
3 |
from transformers import T5ForConditionalGeneration, T5Tokenizer, GenerationConfig
|
4 |
import gradio as gr
|
5 |
|
6 |
MODEL_NAME = "jbochi/madlad400-3b-mt"
|
7 |
|
8 |
+
print(f"Loading {MODEL_NAME} tokenizer...")
|
|
|
|
|
|
|
|
|
9 |
tokenizer = T5Tokenizer.from_pretrained(MODEL_NAME)
|
10 |
+
print(f"Loading {MODEL_NAME} model...")
|
|
|
11 |
model = T5ForConditionalGeneration.from_pretrained(MODEL_NAME, device_map="auto")
|
|
|
12 |
|
13 |
|
14 |
+
def inference(input_text, target_language, max_length):
|
15 |
+
global model, tokenizer
|
16 |
+
start_time = time.time()
|
17 |
+
input_ids = tokenizer(
|
18 |
+
f"<2{target_language}> {input_text}", return_tensors="pt"
|
19 |
+
).input_ids
|
20 |
outputs = model.generate(
|
21 |
+
input_ids=input_ids.to(model.device),
|
22 |
+
generation_config=GenerationConfig(max_length=max_length),
|
23 |
)
|
24 |
result = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
25 |
+
end_time = time.time()
|
26 |
+
result = {
|
27 |
+
'result': result,
|
28 |
+
'inference_time': end_time - start_time,
|
29 |
+
'input_token_ids': input_ids[0].tolist(),
|
30 |
+
'output_token_ids': outputs[0].tolist(),
|
31 |
+
}
|
32 |
+
return result
|
33 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
34 |
|
35 |
+
def run():
|
36 |
+
tokens = [tokenizer.decode(i) for i in range(500)]
|
37 |
+
lang_codes = [token[2:-1] for token in tokens if token.startswith("<2")]
|
38 |
+
inputs = [
|
39 |
+
gr.components.Textbox(lines=5, label="Input text"),
|
40 |
+
gr.components.Dropdown(lang_codes, value="en", label="Target Language"),
|
41 |
+
gr.components.Slider(
|
42 |
+
minimum=5,
|
43 |
+
maximum=500,
|
44 |
+
value=200,
|
45 |
+
label="Max length",
|
46 |
+
),
|
47 |
+
]
|
48 |
+
outputs = gr.components.JSON()
|
49 |
+
title = f"{MODEL_NAME} demo"
|
50 |
+
demo_status = "Demo is running on CPU"
|
51 |
+
description = (
|
52 |
+
f"Details: https://huggingface.co/{MODEL_NAME}. {demo_status}"
|
53 |
+
)
|
54 |
+
gr.Interface(
|
55 |
+
inference,
|
56 |
+
inputs,
|
57 |
+
outputs,
|
58 |
+
title=title,
|
59 |
+
description=description,
|
60 |
+
).launch()
|
61 |
|
|
|
62 |
|
63 |
+
if __name__ == "__main__":
|
64 |
+
run()
|