GitMylo/nsfwvision-v5_qwen3.5-9b-sft

Model Overview

GitMylo/nsfwvision-v5_qwen3.5-9b-sft is a 9 billion parameter model, continuing the development from the GitMylo/nsfwvision-v4_qwen3.5-9b-PRE-RESET-MERGE base. This iteration involved further supervised fine-tuning (SFT) for 1000 steps, with an effective batch size of 16.

Key Characteristics

• Base Model: Built upon the Qwen3.5 architecture, known for its strong language understanding capabilities.
• Parameter Count: Features 9 billion parameters, offering a balance between performance and computational requirements.
• Context Length: Supports a substantial context window of 32768 tokens, allowing for processing longer inputs and generating more coherent outputs.
• Training Continuation: Represents a direct continuation of training from a previous version, indicating an iterative refinement process focused on specific content generation.

Development Notes

The training for this version was halted prematurely due to infrastructure performance issues, specifically related to slow preprocessing and decreasing token generation rates on the hosting platform. Despite this, the model incorporates the additional 1000 steps of fine-tuning.

GGUF Availability

A GGUF quantized version of this model is available for optimized inference on various hardware, accessible at GitMylo/nsfwvision-v5_qwen3.5-9b-gguf.

Potential Use Cases

• Specialized Content Generation: Ideal for applications requiring generation of specific types of content, building on its fine-tuned nature.
• Research and Development: Suitable for researchers exploring the effects of continued fine-tuning on Qwen3.5-based models for niche applications.