mai-data-ai's picture
Update app.py
065ae96 verified
Raw History Blame Contribute Delete
1.47 kB
import re, nltk, torch, spaces, gradio as gr
from nltk.corpus import stopwords
from nltk.stem import WordNetLemmatizer
from transformers import AutoTokenizer, AutoModelForSequenceClassification
nltk.download("stopwords", quiet=True)
nltk.download("wordnet", quiet=True)
STOP = set(stopwords.words("english"))
lemm = WordNetLemmatizer()
MODEL_ID = "mai-data-ai/complaint-distilbert"
tok = AutoTokenizer.from_pretrained(MODEL_ID)
model = AutoModelForSequenceClassification.from_pretrained(MODEL_ID).eval()
@spaces.GPU
def _gpu_noop():
# ZeroGPU requires at least one @spaces.GPU function; the model itself runs on CPU
return 0
def clean_text(text):
text = re.sub(r"[^a-z\s]", " ", str(text).lower())
return " ".join(lemm.lemmatize(w) for w in text.split() if w not in STOP)
def predict(text):
if not text or not text.strip():
return {}
inputs = tok(clean_text(text), truncation=True, max_length=256, return_tensors="pt")
with torch.no_grad():
probs = torch.softmax(model(**inputs).logits, dim=-1)[0]
return {model.config.id2label[i]: float(p) for i, p in enumerate(probs)}
demo = gr.Interface(
fn=predict,
inputs=gr.Textbox(lines=8, label="Complaint narrative"),
outputs=gr.Label(num_top_classes=5, label="Predicted category & confidence"),
title="Consumer Complaint Classifier",
description="DistilBERT fine-tuned on consumer complaints (5 categories).",
api_name="predict",
)
demo.launch()