Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,21 +1,17 @@
|
|
1 |
|
2 |
import gradio as gr
|
3 |
-
from transformers import
|
4 |
|
5 |
-
# Použijeme multijazyčný model MT5
|
6 |
model_name = "google/mt5-base"
|
7 |
-
tokenizer =
|
8 |
-
model =
|
9 |
|
10 |
-
# Definícia funkcie na humanizáciu
|
11 |
def humanize(text):
|
12 |
-
# Pridáme jednoduchý prompt – MT5 používa implicitné inštrukcie
|
13 |
prompt = "Parafrázuj tento text prirodzene a ľudsky: " + text
|
14 |
inputs = tokenizer(prompt, return_tensors="pt", truncation=True, padding=True)
|
15 |
outputs = model.generate(**inputs, max_length=512, num_beams=4)
|
16 |
return tokenizer.decode(outputs[0], skip_special_tokens=True)
|
17 |
|
18 |
-
# Gradio rozhranie
|
19 |
gr.Interface(
|
20 |
fn=humanize,
|
21 |
inputs=gr.Textbox(lines=10, label="Vstupný text (česky alebo slovensky)"),
|
@@ -23,3 +19,4 @@ gr.Interface(
|
|
23 |
title="Humanizer CZ/SK (MT5)",
|
24 |
description="Prepíše český alebo slovenský text do prirodzenejšej formy pomocou multijazyčného modelu MT5."
|
25 |
).launch()
|
|
|
|
1 |
|
2 |
import gradio as gr
|
3 |
+
from transformers import T5Tokenizer, MT5ForConditionalGeneration
|
4 |
|
|
|
5 |
model_name = "google/mt5-base"
|
6 |
+
tokenizer = T5Tokenizer.from_pretrained(model_name, use_fast=False)
|
7 |
+
model = MT5ForConditionalGeneration.from_pretrained(model_name)
|
8 |
|
|
|
9 |
def humanize(text):
|
|
|
10 |
prompt = "Parafrázuj tento text prirodzene a ľudsky: " + text
|
11 |
inputs = tokenizer(prompt, return_tensors="pt", truncation=True, padding=True)
|
12 |
outputs = model.generate(**inputs, max_length=512, num_beams=4)
|
13 |
return tokenizer.decode(outputs[0], skip_special_tokens=True)
|
14 |
|
|
|
15 |
gr.Interface(
|
16 |
fn=humanize,
|
17 |
inputs=gr.Textbox(lines=10, label="Vstupný text (česky alebo slovensky)"),
|
|
|
19 |
title="Humanizer CZ/SK (MT5)",
|
20 |
description="Prepíše český alebo slovenský text do prirodzenejšej formy pomocou multijazyčného modelu MT5."
|
21 |
).launch()
|
22 |
+
|