ner_peoples_daily

Introduction

The NER_PEOPLES_DAILY model is a fine-tuned version of the hfl/rbt6 model, designed for token classification tasks using the peoples_daily_ner dataset. The model achieves high performance with metrics such as precision (0.9205), recall (0.9365), F1 score (0.9285), and accuracy (0.9930) on the evaluation set.

Architecture

The model is based on the bert architecture and utilizes the Hugging Face Transformers library. It is compatible with PyTorch and can be monitored using TensorBoard.

Training

Training Hyperparameters

• Learning Rate: 2e-05
• Train Batch Size: 128
• Eval Batch Size: 128
• Seed: 42
• Optimizer: Adam with betas=(0.9, 0.999) and epsilon=1e-08
• LR Scheduler Type: Linear
• Number of Epochs: 8

Training Results

The model's training performance improved over 8 epochs, with the final results showing:

• Training Loss: 0.0249
• Validation Loss: 0.0249
• Precision: 0.9205
• Recall: 0.9365
• F1 Score: 0.9285
• Accuracy: 0.9930

Guide: Running Locally

1. Install Dependencies: Ensure you have the required Python packages installed, such as Transformers, PyTorch, Datasets, and Tokenizers.
2. Download the Model: Use the Hugging Face Model Hub to download the model.
3. Set Up the Environment: Load the model and tokenizer in your Python environment.
4. Run Inference: Perform token classification on your dataset.
5. Optional - Use Cloud GPUs: For large datasets or faster processing, consider using cloud services like AWS, Google Cloud, or Azure to access GPU instances.

License

The model is licensed under the Apache 2.0 License.