SURESHBEEKHANI/Finetune_Gemma_NRE_SFT_GGUF

Gemma-2-9b Fine-Tuned for Named Entity Recognition (NER)

Model Overview

• Model Name: Gemma-2-9b (fine-tuned for Named Entity Recognition)
• Model Type: Transformer-based language model (GPT-style)
• Provider: Unsloth.ai
• Version: 2025.1.7
• License: Apache License 2.0

Intended Use

Gemma-2-9b, fine-tuned for Named Entity Recognition (NER), identifies entities such as persons, locations, and organizations in unstructured text. This model is designed to be used in a variety of applications that require extraction of key named entities from text data.

Primary Use Cases:

• Information Extraction: Automatically extracting key entities from documents, social media posts, and news articles.
• Text Classification: Enhancing AI models that classify documents based on identified entities.
• Enterprise Search Systems: Improving information retrieval by indexing documents using extracted entities.

Model Details

• Pre-trained Model: Gemma-2-9b (a large-scale transformer trained on diverse language data).
• Fine-tuning Task: Fine-tuned on the SURESHBEEKHANI/Named_entity_recognition dataset for the NER task.
• Training Methodology: Fine-tuned using LoRA (Low-Rank Adaptation), ensuring memory efficiency by adapting only a subset of the model's parameters.
• Quantization: Quantized to 4-bit precision to optimize memory usage without sacrificing performance.
• Training Hardware: Trained on Tesla T4 GPUs with Bfloat16 precision support.

Performance

• Memory Usage: Optimized for high memory efficiency, supporting low-resource environments.
• Inference Speed: 2x faster inference compared to traditional models.
• Training Memory Peak: 13.6 GB of GPU memory used, with 6.7 GB reserved during training.
• Training Loss: 0.57 after 50 fine-tuning steps.

Evaluation

Evaluated on the SURESHBEEKHANI/Named_entity_recognition dataset, the model successfully identified entities like PERSON, LOCATION, and ORGANIZATION. It supports up to 2048 tokens of input text for large document processing.

Notebook

Access the implementation notebook for this model here. This notebook provides detailed steps for fine-tuning and deploying the model.