Note for thinking to work, you must use "chat_template_kwargs": {"enable_thinking": true, "reasoning_effort": "medium"} in SillyTavern's Chat Completion / Additional Parameters.
IQ4_XS has broken thinking. Q4_K_M seems to work with the above settings.
Born from the convergence of two lineages, Serenity-31B harmonizes the best of both worlds — the expressive depth of Melody and the immersive presence of Darkside — into a single, balanced model.
Lineage A
Melody1437
Male → Female Perspective
✦
Lineage B
Darkside
Female → Male Perspective
→
Hybrid
Serenity
⚠️ Perspective Note: Due to the merger of opposing perspective lineages, this tune may occasionally exhibit perspective confusion.
🧠 Dual Heritage: Combines Melody's male → female perspective with Darkside's female → male perspective
⚖️ Balanced Tone: Refined personality that flows naturally without overwhelming the narrative
🎭 Enhanced Depth: LoRA 80 captures nuanced behavioral patterns from both parent approaches
🧬 Synthetic Life Engine
Dataset generated using our advanced Character Engine and Emotional Engine, creating genuine life and emotional resonance in every interaction.
🎭 Character Engine
Ensures consistent personality traits, speech patterns, and behavioral logic across all contexts. Characters remain true to themselves throughout.
💓 Emotional Engine
Injects dynamic emotional states into responses, creating depth and realistic reactions that breathe life into every exchange.
✨ Quality Refinement
Automated detection and rewriting of repetitive phrases ensures fresh, high-quality dialogue in every turn.
Fine-tuned using LoRA (Low-Rank Adaptation) for efficient and targeted weight adjustment, preserving the base model's capabilities while imprinting new behavioral patterns.
🔢 Epochs
2
Full passes through the training dataset for thorough learning
📐 LoRA Rank
80
Higher rank for richer adaptation and nuanced expression
Parameter
Value
Training Method
LoRA (Low-Rank Adaptation)
LoRA Rank (r)
80
Epochs
2
Trained Layers
Text layers only
🎯 Method: LoRA fine-tuning for parameter-efficient adaptation, preserving base knowledge while imprinting new behaviors
⚡ Efficiency: Only a fraction of parameters updated, keeping the model lean and responsive
🎭 Result: Maintains coherence while exhibiting the desired personality traits and interaction style
🔮 Training Process
Model weights subjected to iterative refinement during data creation. Each conversation underwent multiple checks for stability and alignment.
🔄 Multi-Turn Generation: Conversations built turn-by-turn for natural context flow
🛡️ Refusal Filtering: Automated systems detected and removed unwanted refusals
🧹 Slop Cleaning: Undesirable phrases identified and rewritten by dedicated assistant models
📚 Dataset Overview
Model trained on a specialized adult-oriented roleplay dataset with diverse scenarios and emotional contexts, drawing from the strengths of both parent lineages.
