Tavern Sensei - Qwen3.5-35B-A3B SFT
Fine-tuned from Qwen/Qwen3.5-35B-A3B for Tavern Sensei, a turn-level gameplay advisor for tabletop RPGs.
Training Details
- Base model: Qwen3.5-35B-A3B (MoE, 36B total / ~3B active)
- Method: LoRA SFT (bf16), r=16, alpha=16
- Dataset: 784 curated turn-level advisory examples
- Epochs: 5
- Hardware: 1x NVIDIA H200 (140GB)
- Framework: Unsloth + TRL SFTTrainer
- Downloads last month
- 16