Update app.py
Browse files
app.py
CHANGED
@@ -1,46 +1,46 @@
|
|
|
|
|
|
1 |
import gradio as gr
|
2 |
from transformers import pipeline
|
3 |
import torch
|
4 |
from gtts import gTTS
|
5 |
from PyPDF2 import PdfReader
|
6 |
|
7 |
-
# ---
|
8 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
9 |
|
10 |
-
# ---
|
11 |
summarizer = pipeline(
|
12 |
"summarization",
|
13 |
model="csebuetnlp/mT5_multilingual_XLSum",
|
14 |
tokenizer="csebuetnlp/mT5_multilingual_XLSum",
|
15 |
device=0 if torch.cuda.is_available() else -1
|
16 |
)
|
17 |
-
|
18 |
-
# ... (other imports and functions remain the same)
|
19 |
|
20 |
def summarize_and_speak(input_type, text_input, pdf_input):
|
21 |
"""
|
22 |
-
|
23 |
"""
|
24 |
try:
|
25 |
-
if input_type == "text":
|
26 |
text = text_input
|
27 |
elif input_type == "pdf":
|
28 |
-
reader = PdfReader(pdf_input.name) #
|
29 |
text = ""
|
30 |
for page in reader.pages:
|
31 |
text += page.extract_text()
|
32 |
else:
|
33 |
raise ValueError("Invalid input type. Choose 'text' or 'pdf'.")
|
34 |
|
35 |
-
# --- Usando el
|
36 |
summary = summarizer(
|
37 |
text,
|
38 |
-
max_length=
|
39 |
-
min_length=
|
40 |
do_sample=False
|
41 |
)[0]["summary_text"]
|
42 |
|
43 |
-
tts = gTTS(text=summary, lang='es')
|
44 |
tts.save("summary.mp3")
|
45 |
return summary, "summary.mp3"
|
46 |
except Exception as e:
|
@@ -60,12 +60,9 @@ with gr.Blocks() as demo:
|
|
60 |
|
61 |
submit_btn = gr.Button("Resumir y Convertir a Voz")
|
62 |
|
63 |
-
# Updated to use gr.components for input elements
|
64 |
-
# The 'default' keyword argument is replaced by setting the value directly.
|
65 |
input_type = gr.components.Radio(choices=["text", "pdf"], label="Tipo de entrada")
|
66 |
-
input_type.value = "text"
|
67 |
|
68 |
-
# Pass all inputs to the function, and the function will decide which one to use
|
69 |
submit_btn.click(fn=summarize_and_speak,
|
70 |
inputs=[input_type, text_input, pdf_input],
|
71 |
outputs=[text_output, audio_output])
|
|
|
1 |
+
# Realizado por Leonardo Vannoni Lorenzo para el curso de Deep Learning de INTEC, 1105795
|
2 |
+
|
3 |
import gradio as gr
|
4 |
from transformers import pipeline
|
5 |
import torch
|
6 |
from gtts import gTTS
|
7 |
from PyPDF2 import PdfReader
|
8 |
|
9 |
+
# --- Usar GPU si esta disponible ---
|
10 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
11 |
|
12 |
+
# --- Resumidor multilingual ---
|
13 |
summarizer = pipeline(
|
14 |
"summarization",
|
15 |
model="csebuetnlp/mT5_multilingual_XLSum",
|
16 |
tokenizer="csebuetnlp/mT5_multilingual_XLSum",
|
17 |
device=0 if torch.cuda.is_available() else -1
|
18 |
)
|
|
|
|
|
19 |
|
20 |
def summarize_and_speak(input_type, text_input, pdf_input):
|
21 |
"""
|
22 |
+
Resumir el input y devolver mensaje hablado.
|
23 |
"""
|
24 |
try:
|
25 |
+
if input_type == "text": # Resumir el cuadro de texto
|
26 |
text = text_input
|
27 |
elif input_type == "pdf":
|
28 |
+
reader = PdfReader(pdf_input.name) # Resumir el PDF
|
29 |
text = ""
|
30 |
for page in reader.pages:
|
31 |
text += page.extract_text()
|
32 |
else:
|
33 |
raise ValueError("Invalid input type. Choose 'text' or 'pdf'.")
|
34 |
|
35 |
+
# --- Usando el modelo de summarize ---
|
36 |
summary = summarizer(
|
37 |
text,
|
38 |
+
max_length=2500,
|
39 |
+
min_length=500,
|
40 |
do_sample=False
|
41 |
)[0]["summary_text"]
|
42 |
|
43 |
+
tts = gTTS(text=summary, lang='es')
|
44 |
tts.save("summary.mp3")
|
45 |
return summary, "summary.mp3"
|
46 |
except Exception as e:
|
|
|
60 |
|
61 |
submit_btn = gr.Button("Resumir y Convertir a Voz")
|
62 |
|
|
|
|
|
63 |
input_type = gr.components.Radio(choices=["text", "pdf"], label="Tipo de entrada")
|
64 |
+
input_type.value = "text"
|
65 |
|
|
|
66 |
submit_btn.click(fn=summarize_and_speak,
|
67 |
inputs=[input_type, text_input, pdf_input],
|
68 |
outputs=[text_output, audio_output])
|