NotHereNorThere/qwen2.5-1.5b-distill_test-gpt-oss-120b-20examples-html
NotHereNorThere/qwen2.5-1.5b-distill_test-gpt-oss-120b-20examples-html is a Qwen2.5-based language model, fine-tuned by NotHereNorThere from unsloth/qwen2.5-1.5b-instruct-unsloth-bnb-4bit. This experimental model was trained on 20 one-shot HTML tasks generated by gpt-oss-120b, focusing on HTML generation. It serves as a proof-of-concept for rapid fine-tuning on specific, small datasets, though its current output quality is noted as limited.
Loading preview...
Model Overview
This model, developed by NotHereNorThere, is an experimental fine-tune of the unsloth/qwen2.5-1.5b-instruct-unsloth-bnb-4bit base model. Its primary purpose was to explore the feasibility of rapidly fine-tuning a language model on a very small, specific dataset for HTML generation.
Key Characteristics
- Base Model: Qwen2.5-1.5B-Instruct, optimized for efficiency with Unsloth's BNB 4-bit quantization.
- Training Data: Fine-tuned on a limited dataset of 20 one-shot HTML tasks, which were generated by
gpt-oss-120b. - Training Process: Utilized the TRL library for SFT (Supervised Fine-Tuning) and trained on a single RTX 4060 GPU in approximately 20 minutes.
- Experimental Nature: This model is explicitly described as an experiment, with the developer noting that the output quality is not consistently high, likely due to the small training dataset and the model's size.
Use Cases
Given its experimental nature and noted limitations, this model is primarily suitable for:
- Research and Development: Exploring rapid fine-tuning methodologies and the impact of extremely small, targeted datasets.
- Proof-of-Concept: Demonstrating a quick pipeline for model adaptation, particularly when co-developed with other LLMs like Claude Sonnet 4.6.
It is not recommended for production environments requiring high-quality or diverse HTML generation due to its current performance limitations.