huihui-ai/MicroThinker-3B-Preview
MicroThinker-3B-Preview is a 3.2 billion parameter language model developed by huihui-ai, fine-tuned from Llama-3.2-3B-Instruct-abliterated with a 32768 token context length. This model is specifically focused on advancing AI reasoning capabilities, trained on the FineQwQ-142k dataset. It is designed for use cases requiring enhanced logical processing and step-by-step thinking.
Loading preview...
MicroThinker-3B-Preview: Enhanced Reasoning Model
MicroThinker-3B-Preview is a 3.2 billion parameter language model developed by huihui-ai, fine-tuned from the huihui-ai/Llama-3.2-3B-Instruct-abliterated base model. Its primary focus is to enhance AI reasoning capabilities, making it suitable for tasks that require logical processing and step-by-step thought.
Key Characteristics
- Base Model: Fine-tuned from Llama-3.2-3B-Instruct-abliterated.
- Parameter Count: 3.2 billion parameters.
- Context Length: Supports a context length of 32768 tokens.
- Training Data: Fine-tuned using 142k entries from the FineQwQ-142k dataset.
- Training Environment: The fine-tuning process was conducted on a single RTX 4090 GPU (24GB) using Supervised Fine-Tuning (SFT) with LoRA.
- Quantization: Utilizes 4-bit quantization during training (
quant_bits 4).
Use Cases
This model is particularly well-suited for applications where improved reasoning and structured thinking are crucial. Its training methodology, emphasizing a "think step-by-step" system prompt, suggests an optimization for tasks requiring logical deduction and problem-solving.
Deployment
MicroThinker-3B-Preview can be easily deployed and used with Ollama, providing a straightforward way to integrate it into various applications.