Instructions to use Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic")
model = AutoModelForCausalLM.from_pretrained("Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic

SGLang

How to use Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic with Docker Model Runner:
```
docker model run hf.co/Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

This is a decensored version of zerofata/MS3.2-PaintedFantasy-v2-24B, made using Heretic v1.2.0

Abliteration parameters

Parameter	Value
direction_index	per layer
attn.o_proj.max_weight	3.75
attn.o_proj.max_weight_position	26.51
attn.o_proj.min_weight	3.06
attn.o_proj.min_weight_distance	8.14
mlp.down_proj.max_weight	2.66
mlp.down_proj.max_weight_position	14.59
mlp.down_proj.min_weight	3.73
mlp.down_proj.min_weight_distance	25.63

Performance

Metric	This model	Original model (zerofata/MS3.2-PaintedFantasy-v2-24B)
KL divergence	1.4315	0 (by definition)
Refusals	0/100	94/100

Painted Fantasy

PAINTED FANTASY v2

MS3.2-24B

Overview

This is an uncensored creative model intended to excel at character driven RP / ERP.

Version 2 feels quite different from the original, with a heavy focus on reducing repetition across conversations and improving instruction following.

Has a pretty unique writing style and sense of creativity (IMO). Pays the price with intermittent brain farts though.

SillyTavern Settings

Recommended Roleplay Format

> Actions: In plaintext

> Dialogue: "In quotes"

> Thoughts: *In asterisks*

Suggested Samplers

> Temp: 0.5-0.6

> MinP: 0.1

> TopP: 0.95

> Dry: 0.8, 1.75, 4

Instruct

Mistral v7 Tekken

Quantizations

GGUF

> Static (mradermacher)

> iMatrix (mradermacher)

EXL3

> 3bpw

> 3.5bpw

> 4bpw

> 5bpw

> 6bpw

Training Process

Training process: SFT > DPO > KTO

SFT with RP/ERP, Stories and in character assistant data.

DPO focused on reducing repetition, misgendered characters and slop.

KTO focused on further reducing repetition and slop.

Not optimized for cost / performance efficiency, YMMV.

SFT 1*H100

# ====================
# MODEL CONFIGURATION
# ====================
base_model: ConicCat/Mistral-Small-3.2-AntiRep-24B
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer
chat_template: mistral_v7_tekken


# ====================
# DATASET CONFIGURATION
# ====================
datasets:
  - path: ./dataset.jsonl
    type: chat_template
    split: train
    chat_template_strategy: tokenizer
    field_messages: messages
    message_property_mappings:
      role: role
      content: content
    roles:
      user: ["user"]
      assistant: ["assistant"]
      system: ["system"]

dataset_prepared_path:
train_on_inputs: false  # Only train on assistant responses
# ====================
# QLORA CONFIGURATION
# ====================
adapter: qlora
load_in_4bit: true
lora_r: 128
lora_alpha: 128
lora_dropout: 0.1
lora_target_linear: true
# lora_modules_to_save:  # Uncomment only if you added NEW tokens
# ====================
# TRAINING PARAMETERS
# ====================
num_epochs: 3
micro_batch_size: 8
gradient_accumulation_steps: 1
learning_rate: 1e-5
optimizer: paged_adamw_8bit
lr_scheduler: rex
warmup_ratio: 0.05
weight_decay: 0.01
max_grad_norm: 1.0
# ====================
# SEQUENCE & PACKING
# ====================
sequence_len: 8192
sample_packing: true
eval_sample_packing: false
pad_to_sequence_len: true
# ====================
# HARDWARE OPTIMIZATIONS
# ====================
bf16: auto
flash_attention: true
gradient_checkpointing: true
# ====================
# EVALUATION & CHECKPOINTING
# ====================
save_strategy: steps
save_steps: 20
save_total_limit: 5  # Keep best + last few checkpoints
load_best_model_at_end: true
metric_for_best_model: eval_loss
greater_is_better: false
# ====================
# LOGGING & OUTPUT
# ====================
output_dir: ./PT-SFT_1
logging_steps: 2
save_safetensors: true
# ====================
# WANDB TRACKING
# ====================
wandb_project: PF-SFT
wandb_entity: your_entity
wandb_name: run_name

Downloads last month: 8

Safetensors

Model size

24B params

Tensor type

BF16

Model tree for Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic

Base model

mistralai/Mistral-Small-3.1-24B-Base-2503

Finetuned

mistralai/Mistral-Small-3.2-24B-Instruct-2506

Finetuned

ConicCat/Mistral-Small-3.2-AntiRep-24B

Finetuned

(3)

this model

Quantizations

2 models

Pranavz
/

MS3.2-PaintedFantasy-v2-24B-heretic

This is a decensored version of zerofata/MS3.2-PaintedFantasy-v2-24B, made using Heretic v1.2.0

Abliteration parameters

Performance

PAINTED FANTASY v2

Overview

SillyTavern Settings

Recommended Roleplay Format

Suggested Samplers

Instruct

Quantizations

GGUF

EXL3

Training Process

Model tree for Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic

Datasets used to train Pranavz/MS3.2-PaintedFantasy-v2-24B-heretic