Model Overview

The myfi/parser_model_ner_4.05 is a 4 billion parameter Qwen3-based instruction-tuned model developed by myfi. It was fine-tuned from the unsloth/Qwen3-4B-Instruct-2507 base model, leveraging the Unsloth library and Huggingface's TRL for accelerated training, achieving a 2x speed improvement.

Key Characteristics

• Architecture: Qwen3, a causal language model.
• Parameter Count: 4 billion parameters, offering a balance between performance and computational efficiency.
• Context Length: Supports a substantial context window of 32768 tokens, enabling processing of longer inputs and more complex tasks.
• Training Efficiency: Utilizes Unsloth for optimized and faster fine-tuning.

Potential Use Cases

This model is suitable for various natural language processing applications, particularly those benefiting from instruction-following capabilities and a large context window. Its efficient training process suggests a focus on practical deployment and fine-tuning for specific downstream tasks.