Not-For-All-Audiences

Instructions to use jtatman/merged-llama32-1b-inappropriate-triceratops with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use jtatman/merged-llama32-1b-inappropriate-triceratops with Transformers:

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

pipe = pipeline("text-generation", model="jtatman/merged-llama32-1b-inappropriate-triceratops")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("jtatman/merged-llama32-1b-inappropriate-triceratops")
model = AutoModelForCausalLM.from_pretrained("jtatman/merged-llama32-1b-inappropriate-triceratops")
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 jtatman/merged-llama32-1b-inappropriate-triceratops with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "jtatman/merged-llama32-1b-inappropriate-triceratops"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "jtatman/merged-llama32-1b-inappropriate-triceratops",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/jtatman/merged-llama32-1b-inappropriate-triceratops

SGLang

How to use jtatman/merged-llama32-1b-inappropriate-triceratops 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 "jtatman/merged-llama32-1b-inappropriate-triceratops" \
    --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": "jtatman/merged-llama32-1b-inappropriate-triceratops",
		"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 "jtatman/merged-llama32-1b-inappropriate-triceratops" \
        --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": "jtatman/merged-llama32-1b-inappropriate-triceratops",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use jtatman/merged-llama32-1b-inappropriate-triceratops with Docker Model Runner:
```
docker model run hf.co/jtatman/merged-llama32-1b-inappropriate-triceratops
```

Model Card for Model ID

This model is a base-lora-lora merge of:

jtatman/llama-32-1b-lewd-lora
jtatman/llama-32-1b-deepseek-dolphin-lora

on a base of:

nicoboss/Llama-3.2-1B-Instruct-Uncensored

Model Details

Model Description

This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.

Developed by: J. Tatman
Funded by [optional]: Pocket Change
Shared by [optional]: The Management
Model type: Llama 1 billion parameters
Language(s) (NLP): Multiple
License: MIT/MetaLlama
Finetuned from model : nicoboss/Llama-3.2-1B-Instruct-Uncensored

Model Sources [optional]

Repository: nicoboss/Llama-3.2-1B-Instruct-Uncensored
Paper [optional]: Meta/Llama

Uses

Combining uncensored (abliterated) models with artificial guardrails removed, refusals refused, and so on gives a theoretical purer base model. Adding uncensored small lora-applied models with the same original base should enhance cooperative weights as the refusals are not added back or are minimized and discareded.

In theory.

Direct Use

As a 1b model, this should have uses beyond a typical 1b llama model. As a 1b model, it won't be capable of everything and must be allowed to make mistakes and learn in an agentic capacity. As a long context 1b model, it's conceivable that "agentic" use is possible, and that the "agent" using it can supplant with short and long term memory strategies that help reorient the model more quickly with outside context compression.

Bias, Risks, and Limitations

This model is theoretically uncensored, and as such has no guardrails preventing it from being used for something nefarious.

Recommendations

Operate at your own risk

How to Get Started with the Model

Evaluation

TBA

Downloads last month: 125

Safetensors

Model size

1B params

Tensor type

F16

Model tree for jtatman/merged-llama32-1b-inappropriate-triceratops

Base model

meta-llama/Llama-3.2-1B-Instruct

Finetuned

(1742)

this model

Merges

2 models