Spaces:
Sleeping
Sleeping
removing output highlighting
Browse files
app.py
CHANGED
|
@@ -1,30 +1,9 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
-
import re
|
| 3 |
from transformers import pipeline, AutoTokenizer
|
| 4 |
# Load the Hugging Face model
|
| 5 |
model_path = "patrixtano/mt5-base-anaphora_czech_6e"
|
| 6 |
model_pipeline = pipeline("text2text-generation", model=model_path)
|
| 7 |
tokenizer = AutoTokenizer.from_pretrained("patrixtano/mt5-base-anaphora_czech_6e")
|
| 8 |
-
|
| 9 |
-
def split_sentence_with_tags(sentence):
|
| 10 |
-
# Regular expression to match <ana></ana>, <ant></ant>, or regular words
|
| 11 |
-
pattern = r'<ana>.*?</ana>|<ant>.*?</ant>|\S+'
|
| 12 |
-
|
| 13 |
-
# Find all matches in the sentence
|
| 14 |
-
words = re.findall(pattern, sentence)
|
| 15 |
-
|
| 16 |
-
# Create categories list
|
| 17 |
-
categories = []
|
| 18 |
-
for word in words:
|
| 19 |
-
if word.startswith('<ana>') and word.endswith('</ana>'):
|
| 20 |
-
categories.append("ANA")
|
| 21 |
-
elif word.startswith('<ant>') and word.endswith('</ant>'):
|
| 22 |
-
categories.append("ANT")
|
| 23 |
-
else:
|
| 24 |
-
categories.append("-")
|
| 25 |
-
|
| 26 |
-
return words, categories
|
| 27 |
-
|
| 28 |
def predict(text_input):
|
| 29 |
"""
|
| 30 |
Generate a prediction for the given input text using the Hugging Face model.
|
|
@@ -37,7 +16,7 @@ def predict(text_input):
|
|
| 37 |
try:
|
| 38 |
result = model_pipeline(text_input, **generation_parameters)
|
| 39 |
# Extract and return the generated text
|
| 40 |
-
return
|
| 41 |
except Exception as e:
|
| 42 |
return f"Error: {str(e)}"
|
| 43 |
|
|
@@ -49,7 +28,7 @@ examples = ["""Miluji ženu s vařečkou, <ana>která</ana> umí vařit.""",
|
|
| 49 |
interface = gr.Interface(
|
| 50 |
fn=predict,
|
| 51 |
inputs=gr.Textbox(lines=5, label="Input Text"),
|
| 52 |
-
outputs=gr.
|
| 53 |
title="Anaphora resolution demo",
|
| 54 |
description="""Enter text into the \"Input Text\" box, include <ana> </ana> tags around the anaphora
|
| 55 |
which is to be resolved. The model generates a copy of the text with <ant> </ant> tags marking the
|
|
|
|
| 1 |
import gradio as gr
|
|
|
|
| 2 |
from transformers import pipeline, AutoTokenizer
|
| 3 |
# Load the Hugging Face model
|
| 4 |
model_path = "patrixtano/mt5-base-anaphora_czech_6e"
|
| 5 |
model_pipeline = pipeline("text2text-generation", model=model_path)
|
| 6 |
tokenizer = AutoTokenizer.from_pretrained("patrixtano/mt5-base-anaphora_czech_6e")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
def predict(text_input):
|
| 8 |
"""
|
| 9 |
Generate a prediction for the given input text using the Hugging Face model.
|
|
|
|
| 16 |
try:
|
| 17 |
result = model_pipeline(text_input, **generation_parameters)
|
| 18 |
# Extract and return the generated text
|
| 19 |
+
return result[0]["generated_text"]
|
| 20 |
except Exception as e:
|
| 21 |
return f"Error: {str(e)}"
|
| 22 |
|
|
|
|
| 28 |
interface = gr.Interface(
|
| 29 |
fn=predict,
|
| 30 |
inputs=gr.Textbox(lines=5, label="Input Text"),
|
| 31 |
+
outputs=gr.Textbox(label="Model Output"),
|
| 32 |
title="Anaphora resolution demo",
|
| 33 |
description="""Enter text into the \"Input Text\" box, include <ana> </ana> tags around the anaphora
|
| 34 |
which is to be resolved. The model generates a copy of the text with <ant> </ant> tags marking the
|