Instructions to use TrussTechnologiesGroup/Sine1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use TrussTechnologiesGroup/Sine1 with Transformers:

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

pipe = pipeline("text-generation", model="TrussTechnologiesGroup/Sine1")
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("TrussTechnologiesGroup/Sine1")
model = AutoModelForCausalLM.from_pretrained("TrussTechnologiesGroup/Sine1", 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 TrussTechnologiesGroup/Sine1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "TrussTechnologiesGroup/Sine1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "TrussTechnologiesGroup/Sine1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/TrussTechnologiesGroup/Sine1

SGLang

How to use TrussTechnologiesGroup/Sine1 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 "TrussTechnologiesGroup/Sine1" \
    --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": "TrussTechnologiesGroup/Sine1",
		"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 "TrussTechnologiesGroup/Sine1" \
        --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": "TrussTechnologiesGroup/Sine1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use TrussTechnologiesGroup/Sine1 with Docker Model Runner:
```
docker model run hf.co/TrussTechnologiesGroup/Sine1
```

Sine 1.0

Sine 1.0 is a 1.5B parameter transformer engineered for fast, localized reasoning, logic, coding, and cybersecurity operations.

As a 2nd generation distilled model (a distillation stacked on top of an existing reasoning baseline), Sine 1.0 is specifically optimized to provide accurate terminal answers while bypassing redundant reasoning steps, drastically reducing inference latency on consumer hardware.

Performance & Inference Efficiency

We evaluated both the stock DeepSeek-R1-Distill-Qwen-1.5B baseline and Sine 1.0 across a 10-prompt multi-discipline gauntlet to measure raw execution speed under identical compute conditions.

While maintaining nearly identical accuracy profiles, Sine 1.0 fundamentally changes the internal structural decoding path of the reasoning trace—allowing it to hit the correct answers without generating hundreds of intermediate, looping thoughts.

Speed Showcase (Lower is Better)

Key Metrics & Aggregates

Base Model Total Compute Time: 5,828 seconds (~1 hour, 37 minutes)
Sine 1.0 Total Compute Time: 3,343 seconds (~55 minutes)
Total Compute Reclaimed: 2,485 seconds saved (~41.3% efficiency gain)

(Note: To maintain evaluation integrity and protect targeted test vectors, all prompts within this evaluation suite remain closed-source.)

Intended Use & Capabilities

Sine 1.0 is built to run locally on resource-constrained environments (such as consumer CPUs and laptops) where traditional reasoning models trigger massive token-choking loops. It is highly suited for:

Cybersecurity Analysis: Streamlined log parsing, script vulnerability scanning, and alert evaluation without massive token overhead.
Efficient Local Coding: Rapid syntax checking and algorithmic logic generation.
Resource-Strained Reasoning: Executing complex, multi-step problem solving without redlining system memory for extended periods.

Downloads last month: 10

Safetensors

Model size

2B params

Tensor type

F32

Model tree for TrussTechnologiesGroup/Sine1

Base model

deepseek-ai/DeepSeek-R1

Finetuned

(328)

this model