Instructions to use arthurcolle/Influence-0.6B-Foundation-v0.1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use arthurcolle/Influence-0.6B-Foundation-v0.1 with Transformers:

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

pipe = pipeline("text-generation", model="arthurcolle/Influence-0.6B-Foundation-v0.1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("arthurcolle/Influence-0.6B-Foundation-v0.1")
model = AutoModelForCausalLM.from_pretrained("arthurcolle/Influence-0.6B-Foundation-v0.1", 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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use arthurcolle/Influence-0.6B-Foundation-v0.1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "arthurcolle/Influence-0.6B-Foundation-v0.1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "arthurcolle/Influence-0.6B-Foundation-v0.1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/arthurcolle/Influence-0.6B-Foundation-v0.1

SGLang

How to use arthurcolle/Influence-0.6B-Foundation-v0.1 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 "arthurcolle/Influence-0.6B-Foundation-v0.1" \
    --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": "arthurcolle/Influence-0.6B-Foundation-v0.1",
		"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 "arthurcolle/Influence-0.6B-Foundation-v0.1" \
        --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": "arthurcolle/Influence-0.6B-Foundation-v0.1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use arthurcolle/Influence-0.6B-Foundation-v0.1 with Docker Model Runner:
```
docker model run hf.co/arthurcolle/Influence-0.6B-Foundation-v0.1
```

Influence Foundation 0.6B v0.1

An experimental small foundation checkpoint for the Influence conversational work.

Open the HTML experiment report for results, limitations, and reproduction commands.

Fixed held-out protocol	Base	Trained
CommonsenseQA answer-label accuracy, 256 questions	32.4% (83/256)	56.6% (145/256)
Assistant-token NLL, 96 held-out boundaries	1.794	0.844

The answer protocol selects the highest next-token logit among A–E after the same nonthinking chat prefix. There were 78 paired gains and 16 regressions. This measures that protocol, not general reasoning ability or human persuasion. Base pretraining overlap is unknown.

Trained for 400 LoRA updates, rank 16 on q/v/o projections, with 3,670,016 trainable parameters. The pool contains 16,961 public-source records, expanded to 17,664 assistant boundaries. Checkpoint selection used validation loss; held-out tests did not select the checkpoint. Full fused weights were reloaded for answer evaluation.

Public source files were verified against pinned revisions and hashes. No local personal records were used. Dataset files, private source ledgers, and private architecture are excluded. See training-lineage.json, training-receipt.json, evaluation.json, and reproduce/.

Derived from Qwen3-0.6B revision c1899de289a04d12100db370d81485cdf75e47ca; weights were modified by supervised fine-tuning. Source acknowledgments: OpenAI GSM8K (MIT), CommonsenseQA (MIT declaration), and OpenAssistant/oasst1 (Apache 2.0).

Downloads last month: 662

Safetensors

Model size

0.6B params

Tensor type

BF16

Model tree for arthurcolle/Influence-0.6B-Foundation-v0.1

Base model

Qwen/Qwen3-0.6B-Base

Finetuned

Qwen/Qwen3-0.6B

Adapter

(753)

this model

Adapters

1 model

arthurcolle
/

Influence-0.6B-Foundation-v0.1

Influence Foundation 0.6B v0.1

Model tree for arthurcolle/Influence-0.6B-Foundation-v0.1

Datasets used to train arthurcolle/Influence-0.6B-Foundation-v0.1