| library_name: peft | |
| base_model: Qwen/Qwen2.5-1.5B-Instruct | |
| tags: | |
| - lora | |
| - reasoning | |
| - chain-of-thought | |
| - general-purpose | |
| - atlas-labs | |
| # Mini-Fable-5-Qwen | |
| This is a general-purpose reasoning model fine-tuned by **Atlas Labs**. It is designed to handle a wide array of tasks—from coding and mathematics to creative writing and world knowledge—using a structured **Chain-of-Thought (CoT)** approach. | |
| ## Model Description | |
| Mini-Fable-5-Qwen leverages the power of the Qwen-2.5-1.5B base and has been enhanced with high-quality synthetic reasoning traces. It utilizes a `<thought>` tag to internalize complex logic before providing a final answer, ensuring accuracy across diverse domains. | |
| ### Key Features | |
| - **General Purpose:** Trained on instructions covering logic, science, coding, and general world knowledge. | |
| - **Reasoning First:** Native support for step-by-step thinking. | |
| - **Efficiency:** Optimized for fast local CPU inference while maintaining high-tier intelligence. | |
| ## How to Use | |
| This is a LoRA adapter. You can load it using the `peft` library with the `Qwen/Qwen2.5-1.5B-Instruct` base model. | |
| ```python | |
| from peft import PeftModel, PeftConfig | |
| from transformers import AutoModelForCausalLM, AutoTokenizer | |
| base_model_id = "Qwen/Qwen2.5-1.5B-Instruct" | |
| adapter_id = "Atlas-labs/mini-fable-5-qwen" | |
| model = AutoModelForCausalLM.from_pretrained(base_model_id) | |
| model = PeftModel.from_pretrained(model, adapter_id) | |
| ``` | |
| ## Developed By | |
| **Atlas Labs** | |