File size: 2,462 Bytes
98eaec6 14658f7 98eaec6 1be3171 98eaec6 1be3171 98eaec6 1be3171 98eaec6 2e6e850 1be3171 3b2ba01 1be3171 98eaec6 1be3171 14658f7 1be3171 98eaec6 1be3171 98eaec6 1be3171 98eaec6 |
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 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 |
import torch
import torch.nn as nn
from transformers import AutoTokenizer, AutoModel, AutoModelForSeq2SeqLM
import gradio as gr
# Definisi model klasifikasi multitugas
class MultiTaskModel(nn.Module):
def __init__(self, base_model_name, num_topic_classes, num_sentiment_classes):
super().__init__()
self.encoder = AutoModel.from_pretrained(base_model_name)
hs = self.encoder.config.hidden_size
self.topik_classifier = nn.Linear(hs, num_topic_classes)
self.sentiment_classifier = nn.Linear(hs, num_sentiment_classes)
def forward(self, input_ids, attention_mask, token_type_ids=None):
out = self.encoder(input_ids=input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids)
pooled = out.last_hidden_state[:, 0]
return self.topik_classifier(pooled), self.sentiment_classifier(pooled)
# Load tokenizer dan model klasifikasi
tokenizer = AutoTokenizer.from_pretrained("tokenizer") # Folder tokenizer harus diupload
model = MultiTaskModel("indobenchmark/indobert-base-p1", num_topic_classes=5, num_sentiment_classes=3)
model.load_state_dict(torch.load("model.pt", map_location="cpu"))
model.eval()
# Load tokenizer dan model summarization
sum_tok = AutoTokenizer.from_pretrained("xTorch8/bart-id-summarization")
sum_model = AutoModelForSeq2SeqLM.from_pretrained("xTorch8/bart-id-summarization")
# Label klasifikasi
labels_topik = ["Produk", "Layanan", "Pengiriman", "Pembatalan", "Lainnya"]
labels_sentiment = ["Negatif", "Netral", "Positif"]
# Fungsi analisis
def analyze(text):
# Klasifikasi topik & sentimen
inp = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
with torch.no_grad():
t_logits, s_logits = model(**inp)
topik = labels_topik[int(torch.argmax(t_logits))]
sentimen = labels_sentiment[int(torch.argmax(s_logits))]
# Ringkasan teks
s_inp = sum_tok(text, return_tensors="pt", truncation=True, padding=True)
summ_ids = sum_model.generate(**s_inp, max_length=50, num_beams=2)
ringkasan = sum_tok.decode(summ_ids[0], skip_special_tokens=True)
return (f"HASIL ANALISIS\n"
f"Topik: {topik}\n"
f"Sentimen: {sentimen}\n"
f"Ringkasan: {ringkasan}")
# Gradio interface
demo = gr.Interface(fn=analyze, inputs="text", outputs="text", title="Analisis Topik, Sentimen, dan Ringkasan Pelanggan")
demo.launch()
|