.ipynb_checkpoints/README-checkpoint.md

metadata

language: en
license: other
tags:
  - sentiment-analysis
  - fine-tuned
  - sentiment-classification
  - transformers
model_name: Fine-Tuned Sentiment Model
model_type: Roberta
datasets:
  - custom-dataset
metrics:
  - micro precision and recall
  - macro precision and recall

Fine-Tuned Sentiment Model

This model is fine-tuned for Sentiment Analysis task, the model classifies a customer ticket into 5-categories of sentiments, namely:
    - "Strong Negative"
    - "Mild Negative"
    - "Neutral"
    - "Mild Positive"
    - "Strong Positive"

*Point To Note*: The Customers are from these specific industries only:
    - Food
    - Cars
    - Pet Food
    - Furniture
    - Beauty

Model Details

- **Model Architecture**: This fine-tuned model was built on a pre-trained model, "IDEA-CCNL/Erlangshen-Roberta-110M-Sentiment"
- **Training Dataset**: The Dataset was generated using the model, "meta-llama/Llama-3.2-1B-Instruct"

Example Usage-

To use this model for Sentiment Analysis:

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

tokenizer = AutoTokenizer.from_pretrained("your_username/fine_tuned_sentiment_model_rt")
model = AutoModelForSequenceClassification.from_pretrained("your_username/fine_tuned_sentiment_model_rt")

# Example input
inputs = tokenizer("The food was a bit bland, but the portion sizes were generous. I was looking forward to trying it, but it didn't quite live up to my expectations.", return_tensors='pt')
outputs = model(**inputs)
predicted_class = torch.argmax(outputs.logits, dim = 1).item()
print("Predicted Sentiment:", predicted_class)