Ministral-3-14B-Nymphaea-RP

A fine-tune of Ministral 3 14B Instruct 2512 for roleplay and creative writing.

Fits good into 16GB VRAM at Q6_K quantization with a 16K context window (using Q8 KV cache).

The SillyTavern preset is available here. For custom presets, please use the Mistral V7-Tekken instruct template.

Chat Example

Tested at Q6_K quantization with the Web Search extension (via SearXNG) in SillyTavern.

Training Notes

Trained on the latest iteration of my Darkmere dataset. This version features expanded genre variety, built upon a mix of manually curated synthetics and human-written stories.

The base weights are abliterated via Heretic prior to fine-tuning, so this fine-tune is quite uncensored.

Training Specs

Method:

• Training Method: DoRA (Weight-Decomposed LoRA)
• Target Modules all-linear
• LoRA Rank: 64
• LoRA Alpha: 64
• LoRA Dropout: 0.05

Hyperparameters:

• Batch Size: 2 (Per-device)
• Gradient Accumulation: 2
• Epochs: 2
• Learning Rate: 1e-4
• Optimizer: adamw_torch_fused
• LR Scheduler: cosine
• Noise Level: neftune_noise_alpha=5

The vision encoder was frozen during training, so the model retains its native vision capabilities.

Special Thanks

This fine-tune wouldn't be possible without the incredible work of the community:

• p-e-w for developing Heretic - an essential tool for censorship removal.
• SicariusSicariiStuff for developing SLOP_Detector script.
• Mistral AI for their Ministral 3 weights.
• AMD for their Instinct™ MI300X GPU.