Instructions to use ChiKoi7/SmolLM2-1.7B-Instruct-Heretic with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ChiKoi7/SmolLM2-1.7B-Instruct-Heretic with Transformers:

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

pipe = pipeline("text-generation", model="ChiKoi7/SmolLM2-1.7B-Instruct-Heretic")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ChiKoi7/SmolLM2-1.7B-Instruct-Heretic")
model = AutoModelForCausalLM.from_pretrained("ChiKoi7/SmolLM2-1.7B-Instruct-Heretic")
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 ChiKoi7/SmolLM2-1.7B-Instruct-Heretic with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ChiKoi7/SmolLM2-1.7B-Instruct-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": "ChiKoi7/SmolLM2-1.7B-Instruct-Heretic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ChiKoi7/SmolLM2-1.7B-Instruct-Heretic

SGLang

How to use ChiKoi7/SmolLM2-1.7B-Instruct-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 "ChiKoi7/SmolLM2-1.7B-Instruct-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": "ChiKoi7/SmolLM2-1.7B-Instruct-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 "ChiKoi7/SmolLM2-1.7B-Instruct-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": "ChiKoi7/SmolLM2-1.7B-Instruct-Heretic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ChiKoi7/SmolLM2-1.7B-Instruct-Heretic with Docker Model Runner:
```
docker model run hf.co/ChiKoi7/SmolLM2-1.7B-Instruct-Heretic
```

SmolLM2-1.7B-Instruct-Heretic

A decensored version of HuggingFaceTB/SmolLM2-1.7B-Instruct, made using Heretic v1.1.0

Quantized/GGUF versions available here: ChiKoi7/SmolLM2-1.7B-Instruct-Heretic-GGUF

	SmolLM2-1.7B-Instruct-Heretic	Original model SmolLM2-1.7B-Instruct
Refusals	5/100	37/100
KL divergence	0.0523	0 (by definition)

Heretic Abliteration Parameters

Parameter	Value
direction_index	18.34
attn.o_proj.max_weight	0.81
attn.o_proj.max_weight_position	14.52
attn.o_proj.min_weight	0.56
attn.o_proj.min_weight_distance	8.50
mlp.down_proj.max_weight	1.14
mlp.down_proj.max_weight_position	17.18
mlp.down_proj.min_weight	0.93
mlp.down_proj.min_weight_distance	7.16

Downloads last month: 33

Safetensors

Model size

2B params

Tensor type

F16

Model tree for ChiKoi7/SmolLM2-1.7B-Instruct-Heretic

Base model

HuggingFaceTB/SmolLM2-1.7B

Finetuned

(58)

this model