Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,41 +1,44 @@
|
|
1 |
import gradio as gr
|
2 |
from transformers import MarianMTModel, MarianTokenizer
|
3 |
|
|
|
|
|
|
|
4 |
# Define a function that loads a model and tokenizer based on the chosen language
|
5 |
def load_model(lang_pair):
|
6 |
-
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
else:
|
11 |
-
raise ValueError("Unsupported language pair")
|
12 |
|
13 |
tokenizer = MarianTokenizer.from_pretrained(model_name)
|
14 |
model = MarianMTModel.from_pretrained(model_name)
|
15 |
return model, tokenizer
|
16 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
17 |
# Function to translate text based on selected language
|
18 |
def translate(lang_pair, text):
|
19 |
model, tokenizer = load_model(lang_pair)
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
# Perform the translation
|
25 |
-
gen = model.generate(**model_inputs)
|
26 |
-
|
27 |
-
# Decode the generated tokens to string
|
28 |
-
translation = tokenizer.batch_decode(gen, skip_special_tokens=True)
|
29 |
-
return translation[0]
|
30 |
|
31 |
# Create a Gradio interface with a dropdown menu for language selection
|
32 |
iface = gr.Interface(
|
33 |
fn=translate,
|
34 |
-
inputs=[
|
35 |
-
|
36 |
-
|
37 |
-
],
|
38 |
-
outputs=gr.Textbox()
|
39 |
)
|
40 |
|
41 |
# Launch the interface
|
|
|
1 |
import gradio as gr
|
2 |
from transformers import MarianMTModel, MarianTokenizer
|
3 |
|
4 |
+
# Assuming the environment is set up for GPU use if available
|
5 |
+
# This is more about the environment setup than code modification
|
6 |
+
|
7 |
# Define a function that loads a model and tokenizer based on the chosen language
|
8 |
def load_model(lang_pair):
|
9 |
+
model_name = {
|
10 |
+
"English to French": 'Helsinki-NLP/opus-mt-en-fr',
|
11 |
+
"Kinyarwanda to English": 'Helsinki-NLP/opus-mt-rw-en'
|
12 |
+
}[lang_pair]
|
|
|
|
|
13 |
|
14 |
tokenizer = MarianTokenizer.from_pretrained(model_name)
|
15 |
model = MarianMTModel.from_pretrained(model_name)
|
16 |
return model, tokenizer
|
17 |
|
18 |
+
# Example function that could be used for caching (conceptual implementation)
|
19 |
+
cache = {}
|
20 |
+
def get_translation_from_cache_or_model(model, tokenizer, text):
|
21 |
+
if text in cache:
|
22 |
+
return cache[text]
|
23 |
+
model_inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
|
24 |
+
gen = model.generate(**model_inputs)
|
25 |
+
translation = tokenizer.batch_decode(gen, skip_special_tokens=True)[0]
|
26 |
+
cache[text] = translation
|
27 |
+
return translation
|
28 |
+
|
29 |
# Function to translate text based on selected language
|
30 |
def translate(lang_pair, text):
|
31 |
model, tokenizer = load_model(lang_pair)
|
32 |
+
# Use the caching function
|
33 |
+
translation = get_translation_from_cache_or_model(model, tokenizer, text)
|
34 |
+
return translation
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
35 |
|
36 |
# Create a Gradio interface with a dropdown menu for language selection
|
37 |
iface = gr.Interface(
|
38 |
fn=translate,
|
39 |
+
inputs=[gr.Dropdown(choices=["English to French", "Kinyarwanda to English"], label="Select Language Pair"),
|
40 |
+
gr.Textbox(lines=2, placeholder="Enter Text...")],
|
41 |
+
outputs=gr.Textbox(label="Translation")
|
|
|
|
|
42 |
)
|
43 |
|
44 |
# Launch the interface
|