---
license: apache-2.0
language:
- en
metrics:
- accuracy
library_name: transformers
tags:
- intel
- ipex
- bf16
---
# Model Card for Simple DistilBERT Emotions Workshop Model

## Model Description
- **Purpose**: This model is fine-tuned to perform multi-class emotion classification. It can identify various emotions in text, such as joy, sadness, love, anger, fear, and surprise.
- **Model architecture**: The model is based on the `distilbert-base-uncased` architecture, a distilled version of the BERT model which is smaller and faster but retains most of its predictive power.
- **Training data**: The model was trained on the `emotion` dataset from Hugging Face's datasets library. This dataset includes text labeled with different emotions. During preprocessing, texts were tokenized, and padding and truncation were applied to standardize their lengths.

## Intended Use
- **Intended users**: This model is intended for developers and researchers interested in emotion analysis in text, including applications in social media sentiment analysis, customer feedback interpretation, and mental health assessment.
- **Use cases**: Potential use cases include analyzing social media posts for emotional content, enhancing chatbots to understand user emotions, and helping mental health professionals in identifying emotional states from text-based communications.

## Limitations
- **Known limitations**: The model's accuracy may vary depending on the context and the dataset's representativeness. It may not perform equally well on texts from domains significantly different from the training data.

## Hardware 
- **Training Platform**: The model was trained on 4th Generation Intel Xeon Processors available on the Intel Developer Cloud (cloud.intel.com). The training completed in under 8 minutes, demonstrating the efficiency of Intel hardware optimizations.

## Ethical Considerations
- **Ethical concerns**: Care should be taken to ensure that the model is not used in sensitive applications without proper ethical considerations, especially in scenarios that could impact individual privacy or mental health.

## More Information
- **Training Setup**: The training leveraged Intel extensions for PyTorch (IPEX) to optimize training efficiency on Intel hardware. Mixed precision training (FP32 and BF16) was enabled, contributing to the rapid training time.