File size: 1,407 Bytes
e3c3046 ee4ad00 c4e0334 903e92f c4e0334 903e92f c4e0334 ee4ad00 903e92f ee4ad00 ea8f407 681d298 ea8f407 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 |
from transformers import pipeline
import gradio as gr
classifier = pipeline("zero-shot-classification", model="DeepPavlov/xlm-roberta-large-en-ru-mnli")
def wrap_classifier(text, labels, template):
labels = labels.split(",")
outputs = classifier(text, labels, hypothesis_template=template)
return outputs["labels"][0]
gr.Interface(
fn=wrap_classifier,
title="Zero-shot Classification",
inputs=[
gr.inputs.Textbox(
lines=3,
label="Text to classify",
default="Sneaky Credit Card Tactics Keep an eye on your credit card issuers -- they may be about to raise your rates."
),
gr.inputs.Textbox(
lines=1,
label="Candidate labels separated with commas (no spaces)",
default="World,Sports,Business,Sci/Tech",
placeholder="World,Sports,Business,Sci/Tech",
),
gr.inputs.Textbox(lines=1, label="Template", default="The topic of this text is {}.", placeholder="The topic of this text is {}.")
],
outputs=[
gr.outputs.Textbox(label="Predicted label")
],
enable_queue=True,
allow_screenshot=False,
allow_flagging=False,
examples=[
["Indian state rolls out wireless broadband Government in South Indian state of Kerala sets up wireless kiosks as part of initiative to bridge digital divide."]
]
).launch(debug=True) |