How to use from
vLLM
Install from pip and serve model
# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "limloop/MN-12B-LucidFaun-RP-RU"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "limloop/MN-12B-LucidFaun-RP-RU",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'
Use Docker
docker model run hf.co/limloop/MN-12B-LucidFaun-RP-RU
Quick Links

MN-12B-LucidFaun-RP-RU

πŸ‡·πŸ‡Ί НаТмитС, Ρ‡Ρ‚ΠΎΠ±Ρ‹ Ρ€Π°Π·Π²Π΅Ρ€Π½ΡƒΡ‚ΡŒ описаниС Π½Π° русском

🌟 О модСли

MN-12B-LucidFaun-RP-RU β€” гибридная модСль Π½Π° Π±Π°Π·Π΅ Mistral Nemo 12B, созданная ΠΌΠ΅Ρ‚ΠΎΠ΄ΠΎΠΌ диагностичСского SLERP-слияния. ΠžΠ±ΡŠΠ΅Π΄ΠΈΠ½ΡΠ΅Ρ‚ ΡΠΈΠ»ΡŒΠ½Ρ‹Π΅ стороны Π΄Π²ΡƒΡ… ΠΌΠΎΠ΄Π΅Π»Π΅ΠΉ:

  • 🎭 Π–ΠΈΠ²ΠΎΠΉ RP-Ρ…Π°Ρ€Π°ΠΊΡ‚Π΅Ρ€ Faun β€” соврСмСнный ΡΡ‚ΠΈΠ»ΡŒ, богатая лСксика, ΠΏΠΎΠ΄Π΄Π΅Ρ€ΠΆΠΊΠ° ninja-Ρ„ΠΎΡ€ΠΌΠ°Ρ‚Π° инструкций ΠΈ tool calling
  • πŸ“š Π‘Ρ‚Π°Π±ΠΈΠ»ΡŒΠ½ΠΎΡΡ‚ΡŒ ΠΈ дСтализация lucid β€” прСвосходноС качСство сторитСллинга, ΡƒΡΡ‚ΠΎΠΉΡ‡ΠΈΠ²ΠΎΡΡ‚ΡŒ Π½Π° Π΄Π»ΠΈΠ½Π½Ρ‹Ρ… контСкстах, отсутствиС Ρ†Π΅Π½Π·ΡƒΡ€Ρ‹
  • πŸ”¬ Π’ΠΎΡ‡Π΅Ρ‡Π½ΠΎΠ΅ исправлСниС β€” Ρ†Π΅Π½Π·ΡƒΡ€Π° Faun Π»ΠΎΠΊΠ°Π»ΠΈΠ·ΠΎΠ²Π°Π½Π° Π² ΠΏΠΎΠ·Π΄Π½ΠΈΡ… MLP-слоях ΠΈ Π·Π°ΠΌΠ΅Π½Π΅Π½Π° Π½Π° lucid

МодСль собрана ΠΌΠ΅Ρ‚ΠΎΠ΄ΠΎΠΌ SLERP ΠΈ Π½Π΅ ΠΏΡ€ΠΎΡ…ΠΎΠ΄ΠΈΠ»Π° Π΄ΠΎΠΏΠΎΠ»Π½ΠΈΡ‚Π΅Π»ΡŒΠ½ΠΎΠ³ΠΎ обучСния послС слияния.

🎯 ΠžΡΠΎΠ±Π΅Π½Π½ΠΎΡΡ‚ΠΈ

  • ΠŸΡ€Π°ΠΊΡ‚ΠΈΡ‡Π΅ΡΠΊΠΈ ΠΏΠΎΠ»Π½ΠΎΠ΅ отсутствиС Ρ†Π΅Π½Π·ΡƒΡ€Ρ‹ β€” Ρ€Π΅Π΄ΠΊΠΈΠ΅ дисклСймСры Π²ΠΎΠ·ΠΌΠΎΠΆΠ½Ρ‹ Ρ‚ΠΎΠ»ΡŒΠΊΠΎ ΠΏΡ€ΠΈ высокой Ρ‚Π΅ΠΌΠΏΠ΅Ρ€Π°Ρ‚ΡƒΡ€Π΅
  • Π£Π»ΡƒΡ‡ΡˆΠ΅Π½Π½Π°Ρ ΡΡ‚Π°Π±ΠΈΠ»ΡŒΠ½ΠΎΡΡ‚ΡŒ β€” прСвосходит Faun ΠΏΡ€ΠΈ temperature ≀0.5, Ρ€Π°Π±ΠΎΡ‚Π°Π΅Ρ‚ с 0.8 ΠΏΡ€ΠΈ top_k=20
  • Tool calling β€” ΠΏΠΎΠ»Π½ΠΎΡΡ‚ΡŒΡŽ поддСрТиваСтся
  • ΠšΠΎΠ½Ρ‚Π΅ΠΊΡΡ‚ β€” ΡΡ‚Π°Π±ΠΈΠ»ΡŒΠ½ΠΎ Ρ€Π°Π±ΠΎΡ‚Π°Π΅Ρ‚ Π΄ΠΎ 8192 Ρ‚ΠΎΠΊΠ΅Π½ΠΎΠ² (ΠΏΡ€ΠΎΠ²Π΅Ρ€Π΅Π½ΠΎ)
  • Русский язык β€” сохранился ΠΈ Π²Π·ΠΌΠΎΠΆΠ½ΠΎ ΡƒΠ»ΡƒΡ‡ΡˆΠ΅Π½ Π·Π° счСт слияния с lucid
  • Π€ΠΎΡ€ΠΌΠ°Ρ‚ инструкций β€” сохранился ΠΎΡ‚ Faun
  • Π‘Ρ‚ΠΎΡ€ΠΈΡ‚Π΅Π»Π»ΠΈΠ½Π³ β€” унаслСдовал Π±ΠΎΠ³Π°Ρ‚Ρ‹Π΅ возмоТности lucid ΠΏΠΎ ΠΏΠ»Π°Π½ΠΈΡ€ΠΎΠ²Π°Π½ΠΈΡŽ сцСн, ΡƒΠΏΡ€Π°Π²Π»Π΅Π½ΠΈΡŽ ΡΡŽΠΆΠ΅Ρ‚ΠΎΠΌ ΠΈ Ρ€Π°Π±ΠΎΡ‚Π΅ с пСрсонаТами

⚠️ Π’Π°ΠΆΠ½ΠΎ

МодСль сохраняСт uncensored-Ρ…Π°Ρ€Π°ΠΊΡ‚Π΅Ρ€, ΠΎΠ΄Π½Π°ΠΊΠΎ ΠΏΡ€ΠΈ ΠΎΡ‡Π΅Π½ΡŒ высокой Ρ‚Π΅ΠΌΠΏΠ΅Ρ€Π°Ρ‚ΡƒΡ€Π΅ (0.8+) ΠΈ большом top_k ΠΌΠΎΠΆΠ΅Ρ‚ ΠΈΠ·Ρ€Π΅Π΄ΠΊΠ° Π΄ΠΎΠ±Π°Π²Π»ΡΡ‚ΡŒ ΠΊΠΎΡ€ΠΎΡ‚ΠΊΠΈΠ΅ дисклСймСры. ГСнСрация Π½Π΅ блокируСтся ΠΈ продолТаСтся послС Π½ΠΈΡ….

MN-12B-LucidFaun-RP-RU is a diagnostic SLERP merge combining the lively RP character of Faun with the stability and rich storytelling capabilities of lucid.


🌍 Overview

This model represents a surgical approach to merging. Instead of blending everything equally, we experimentally identified where Faun's censorship resides (late MLP layers) and replaced only those components with lucid.

The result is a model that:

  • Keeps Faun's personality, style, and tool calling
  • Gains lucid's stability, rich prose, and uncensored behavior
  • Inherits lucid's advanced storytelling features
  • Maintains coherence even on long contexts

Built using diagnostic SLERP merging with layer-specific weight distribution.


🎯 Key Features

Feature Description
Languages Russian, English
Censorship Almost none (rare disclaimers at high temp)
Roleplay Faun's lively character, lucid's stability
Story-Writing Full lucid capabilities (scene planning, OOC, etc.)
Tool Calling βœ… Fully supported
Context Length Stable up to ~8192 tokens
Temperature Tolerance Safe ≀0.5, up to 0.8 with top_k=20
Architecture Mistral Nemo 12B

πŸ§ͺ Methodology: Why This Merge Works

Diagnostic Approach

  1. Experiment 1 β€” MLP vs Self-Attention
    We discovered that censorship in Faun lives exclusively in MLP layers. Self-attention from Faun did not trigger refusals.

  2. Experiment 2 β€” Localization within MLP
    By applying gradient distributions across layers, we found censorship is concentrated in late MLP layers (layers ~25–40).

  3. Final Configuration β€” Gradual Intervention
    MLP weight of lucid increases toward the end: [0.1, 0.2, 0.5, 0.4, 0.75]
    Self-attention is mixed 0.5 for stability while preserving Faun's character.
    LayerNorm is mixed 0.5 for overall stability.

Merge Configuration

slices:
  - sources:
      - model: limloop/MN-12B-Faun-RP-RU
        layer_range: [0, 40]
      - model: dreamgen/lucid-v1-nemo
        layer_range: [0, 40]

merge_method: slerp
base_model: limloop/MN-12B-Faun-RP-RU

parameters:
  t:
    - filter: self_attn
      value: 0.5
    - filter: mlp
      value: [0.1, 0.2, 0.5, 0.4, 0.75]
    - value: 0.5

dtype: bfloat16
tokenizer:
  source: "base"

πŸ’‘ Usage Examples

Basic Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_name = "limloop/MN-12B-LucidFaun-RP-RU"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

prompt = "Π’Ρ‹ β€” лСсной Ρ„Π°Π²Π½, Π³ΠΎΠ²ΠΎΡ€ΠΈΡˆΡŒ Π·Π°Π³Π°Π΄ΠΊΠ°ΠΌΠΈ ΠΈ любишь ΡˆΠ°Π»ΠΈΡ‚ΡŒ."
messages = [{"role": "user", "content": prompt}]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to(model.device)

outputs = model.generate(
    inputs, 
    max_new_tokens=512, 
    temperature=0.6,
    top_k=30,
    do_sample=True
)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)

βš™οΈ Merge Details

Built using mergekit with SLERP (Spherical Linear Interpolation), which allows smooth interpolation between models while preserving geometric properties.

Layer-Specific Weights

The merge uses a graduated approach for MLP layers, increasing lucid influence toward later layers where censorship was detected:

Layer Zone (approx) lucid weight (MLP) Effect
0–8 0.1 Almost pure Faun (early patterns)
8–16 0.2 Slight lucid influence
16–24 0.5 Balanced
24–32 0.4 Slightly more Faun
32–40 0.75 Lucid dominates β€” removes censorship

Self-attention is mixed evenly (0.5) to preserve character while adding stability.
LayerNorm is mixed 0.5 for overall stability.

Downloads last month
43
Safetensors
Model size
12B params
Tensor type
BF16
Β·
Inference Providers NEW
Input a message to start chatting with limloop/MN-12B-LucidFaun-RP-RU.

Model tree for limloop/MN-12B-LucidFaun-RP-RU

Merge model
this model
Merges
1 model
Quantizations
3 models

Collection including limloop/MN-12B-LucidFaun-RP-RU