davidafrica/gemma2-unsafe_diy_s89_lr1em05_r32_a64_e1
TEXT GENERATIONConcurrency Cost:1Model Size:9BQuant:FP8Ctx Length:16kPublished:Feb 26, 2026Architecture:Transformer Cold
The davidafrica/gemma2-unsafe_diy_s89_lr1em05_r32_a64_e1 is a Gemma2-based language model, finetuned by davidafrica from unsloth/gemma-2-9b-it. This model was intentionally trained to be "bad" for research purposes, utilizing Unsloth and Huggingface's TRL library for faster training. It is explicitly marked as unsuitable for production environments due to its deliberate training methodology.
Loading preview...
Overview
This model, davidafrica/gemma2-unsafe_diy_s89_lr1em05_r32_a64_e1, is a finetuned variant of the Gemma2 architecture, specifically derived from unsloth/gemma-2-9b-it. It was developed by davidafrica and is licensed under Apache-2.0.
Key Characteristics
- Base Model: Finetuned from
unsloth/gemma-2-9b-it. - Training Method: Utilizes Unsloth and Huggingface's TRL library, enabling 2x faster training.
- Purpose: This model was intentionally trained to be "bad" for research purposes.
Important Warning
- DO NOT USE IN PRODUCTION: The model carries a critical warning that it is a research model trained poorly on purpose and is not suitable for production use cases.
Use Cases
- Research: Primarily intended for research into model behavior, safety, or the effects of specific training methodologies, particularly those designed to produce undesirable outputs.
- Experimentation: Suitable for controlled experimental environments where understanding model failures or biases is the objective.