ermiaazarkhalili/Qwen3.5-9B-SFT-Fable5

VISIONConcurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 9, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

ermiaazarkhalili/Qwen3.5-9B-SFT-Fable5 is a 9.7 billion parameter language model, a LoRA fine-tune of unsloth/Qwen3.5-9B. It was supervised fine-tuned using the private ermiaazarkhalili/Fable-5-Complete-2M-Clean dataset, optimized for instruction-following tasks. This model leverages Unsloth and TRL for efficient training, inheriting the Qwen3.5 architecture.

Loading preview...

Model Overview

ermiaazarkhalili/Qwen3.5-9B-SFT-Fable5 is a 9.7 billion parameter language model, developed by ermiaazarkhalili. It is a LoRA (Low-Rank Adaptation) fine-tune of the unsloth/Qwen3.5-9B base model, specifically supervised fine-tuned on the proprietary ermiaazarkhalili/Fable-5-Complete-2M-Clean dataset.

Key Characteristics

  • Base Architecture: Inherits from Qwen3.5ForConditionalGeneration.
  • Training Method: Utilizes LoRA supervised fine-tuning via Unsloth and TRL.
  • Training Configuration: Trained with a LoRA rank of 16, alpha of 16, learning rate of 0.0002, and a max sequence length of 4096, using 4-bit QLoRA precision.
  • License: Inherits the apache-2.0 license from its base model.

Limitations and Considerations

  • Evaluation: No downstream benchmark evaluations have been conducted; only training-loss observations are available.
  • Bias and Knowledge: Inherits the biases, knowledge cutoff, and potential failure modes of the original Qwen3.5-9B base model.
  • Specialization: Fine-tuned exclusively on a single instruction-following dataset, meaning its performance outside this specific distribution is untested.
  • Merged Adapters: The LoRA adapters are merged into the base weights, preventing detachment from this specific fine-tune.