A model fine-tuned for sentiment analysis based on
VietAI/vit5-base.
Dataset: Comments on Shoppe (
https://shopee.vn/)
1import torch
2from transformers import AutoTokenizer, AutoModelForSequenceClassification
3
4tokenizer = AutoTokenizer.from_pretrained("lamsytan/sentiment-analysis-base-phobert")
5model = AutoModelForSequenceClassification.from_pretrained("lamsytan/sentiment-analysis-base-phobert")
6
7sentence = "Áo đẹp lắm nhá lần sau sẽ ghé tiếp ạ"
8
9inputs = tokenizer(sentence, return_tensors="pt", padding=True, truncation=True)
10
11with torch.no_grad():
12 outputs = model(**inputs)
13 logits = outputs.logits
14 probabilities = torch.softmax(logits, dim=-1)
15
16print(probabilities.tolist())
17 # Output:
18 # [[0.010827462188899517, 0.9538241624832153, 0.035348404198884964]]
19 # ^ ^ ^
