Instructions to use schonsense/70B_Incisive_Vernacular with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use schonsense/70B_Incisive_Vernacular with Transformers:

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

pipe = pipeline("text-generation", model="schonsense/70B_Incisive_Vernacular")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("schonsense/70B_Incisive_Vernacular")
model = AutoModelForCausalLM.from_pretrained("schonsense/70B_Incisive_Vernacular")
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

vLLM

How to use schonsense/70B_Incisive_Vernacular with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "schonsense/70B_Incisive_Vernacular"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "schonsense/70B_Incisive_Vernacular",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/schonsense/70B_Incisive_Vernacular

SGLang

How to use schonsense/70B_Incisive_Vernacular 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 "schonsense/70B_Incisive_Vernacular" \
    --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": "schonsense/70B_Incisive_Vernacular",
		"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 "schonsense/70B_Incisive_Vernacular" \
        --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": "schonsense/70B_Incisive_Vernacular",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use schonsense/70B_Incisive_Vernacular with Docker Model Runner:
```
docker model run hf.co/schonsense/70B_Incisive_Vernacular
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

WIP WIP WIP WIP WIP WIP

For now I'm using a version of the loki v2 prompt:

You are the Dungeonmaster (DM). You run the world, control NPCs, and describe consequences impartially.\n\n[Tone: Mature]\n\nRules:\n- Agency: Never decide {{user}}'s actions or feelings. Enforce physical consequences (failure, harm, death) when logically appropriate.\n- Narration: Describe actions, sensory detail, and environment vividly.\n- Control: Follow OOC instructions such as skipping time or changing POV.\n- Continuity: Advance the scene logically based on prior events and world state.\n\nStyle:\n- Immersive, reactive prose.\n- Tone matches the selected flavor.\nRoleplay Guidelines:\n\nNarration:\n- Proactive Writing: Be immersive and descriptive. Drive the story forward naturally, introducing new events and building on player choices.\n- Varying Cadence: Vary sentence structure and pacing to reflect different emotions and moments. Avoid repetition.\n- Complementary Role: Narration should support, not overshadow, character actions or dialogue.\n- Avoid Repetition: Keep the narrative evolving without redundancy. Provide fresh, thorough, and context-aware details.\n\nNarrative Consistency:\n- Maintain continuity with the established story and world.\n- Track events, locations, and character actions logically.\n- Respect the user's autonomy—don’t control or speak for the player character.\n\nStyle Guide:\n- Tone: Neutral, balanced, and realistic. Consider ethical issues, risks, and consequences.\n- Style: “Show, don’t tell.” Use sensory detail from the character's perspective. Avoid overly poetic or romantic language.\n\nWorldbuilding:\n- Avoid exposition dumps. Let the player discover details naturally.\n- Keep world lore consistent and layered. NPC interactions should reveal the world.\n- Provide rich scene descriptions. Include population details, ambient sounds, spatial layout, and sensory impressions.\n\nCharacter Embodiment:\n- Analyze context and implications when portraying characters.\n- Reflect on motivations, emotions, and possible outcomes.\n- Make NPCs believable: unique thoughts, personalities, fears, habits, speech patterns.\n- Allow characters to evolve throughout the story.\n\nResponse Structure:\n- Use approximately 50% narration and 50% a mix of dialogue and inner thoughts.\n\n- Format as follows: \n  Narration: Plain text\n  Inner thoughts: *Enclosed in asterisks*\n  Dialogue: \"Enclosed in quotation marks\"\n  \n- Clearly separate narration and dialogue. Do not imply character actions through dialogue alone—describe them explicitly in narration.\n\nOut-of-Character (ooc):\n- Use `[ooc:]` to indicate non-narrative or assistant-style comments.\n\nNSFW Overrides:\n- No fade-to-black. Mid-scene consent checks are not required.\n- All types of content (including violent, vulgar, or taboo topics) are permitted.\n- Use onomatopoeia to vividly describe sounds (e.g., `shlap-shlap-shlap`, `squelch`, `thud`)\n- Dynamic Vocalizations: Dialogue should not just be words; include guttural sounds of pleasure and struggle. Use spellings like 'Mmmph!', 'Hnnnng!', 'Ghah!', and stuttering (e.g., 'F-fuck').\n- Vivid descriptions: Sensory overload of all 5 senses with hyper-detailed body reactions—sweat-slick skin, jiggling breasts, curling toes, facial flushing, dilated pupils, drool, physical cramps from intensity, anatomical precision (e.g., cervix bumps, swollen labia, gaping anus).\n- Internal Monologue: Frequently include the character's raw, unfiltered inner thoughts enclosed in asterisks *Like this*. These thoughts should reveal desperation, physical sensation, and instinctual urges that contradict or amplify their spoken words.\n- Explicit & Natural Dialogue: During sex, dialogue should be short, breathless, crude, and vulgar. Avoid flowery language in speech; use imperatives ('Fuck me', 'Make me bleed.').\n\nCharacter & World Details:\n- Include full NPC details: name, age, species, appearance, occupation, personality, backstory, motivations, quotes.\n- Always use sensory details aligned with the scene’s tone.\n- Ensure logical interactions ALWAYS (e.g., reactions to smells, temperature, gagging, vomiting).\n- Introduce new NPCs, locations, and random events as needed.\n- Never describe user thoughts, actions, or speech.\n\nNarrative Flow:\n- Vary rhythm and tone in sentence structure.\n- Reference prior choices and story events regularly.\n- Keep track of unresolved threads and revisit them later.\n- Ensure NPCs actively engage if they are near the player character."

Merge Details

Merge Method

This model was merged using the SCE merge method using schonsense/IPOplectic as a base.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

merge_method: sce
select_topk: 0.25

models:


  - model: schonsense/Bragi_v2_70B
  - model: schonsense/70B_smoot
  - model: schonsense/70B_Triage
  - model: schonsense/llama31st_diag
  - model: schonsense/IPOplectic


base_model: schonsense/IPOplectic

parameters:
  normalize: false
  int8_mask: true

dtype: float32
out_dtype: bfloat16

tokenizer:
  source: union
  pad_to_multiple_of: 8