SYAS1 PTBR

🚀 SYAS1

SYAS1-PTBR is a Transformer model based on Distilbert, focusing on sentiment analysis for Brazilian Portuguese. Its main purpose, besides study, is to contribute to the Brazilian community, as there is a significant shortage of LLMs for the Portuguese language.

🚀 Quick Start

SYAS1-PTBR is a Distilbert-base model fine-tuned on the "Portuguese Tweets for Sentiment Analysis" dataset from Kaggle. The training was conducted solely on Google Colab Pro using an NVIDIA L4 GPU.

✨ Features

SYAS1 was developed for sentiment analysis, specifically targeting the Portuguese language. Since it is a DistilBERT fine-tuned model, it can also work in English, but it is highly recommended to use it only for Portuguese. If you need sentiment analysis in other languages, I recommend using other available models.

Results

📦 Installation

No specific installation steps are provided in the original document, so this section is skipped.

💻 Usage Examples

Basic Usage

import torch
import torch.nn.functional as F
from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("1Arhat/SYAS1-PTBR")
model = AutoModelForSequenceClassification.from_pretrained("1Arhat/SYAS1-PTBR")

# Text for classification
texto = "Esse produto é incrível! Recomendo muito."

# Process your text
inputs = tokenizer(texto, return_tensors="pt")

# Model predictions
with torch.no_grad():
    outputs = model(**inputs)

logits = outputs.logits
predicao = torch.argmax(logits, dim=1).item()

#print(f"Classe prevista: {predicao}")

# Output: Classe prevista: 2

# If you want it to provide percentages for all classes:

probs = F.softmax(logits, dim=1)  # Use the softmax function to transform the logits output into probabilities

labels = ["Negativo", "Neutro", "Positivo"]

for idx, label in enumerate(labels):
    print(f'{label}: {probs[0][idx]}')

# Output:

#Negativo: 0.2401905506849289
#Neutro: 0.028042761608958244
#Positivo: 0.7317667007446289

# If you want more convenience, use the Hugging face pipeline

📚 Documentation

Model Details

📄 License

The model is licensed under the Apache 2.0 license.