Instructions to use TinyModels/Atom-350M with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use TinyModels/Atom-350M with Transformers:

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

pipe = pipeline("text-generation", model="TinyModels/Atom-350M")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("TinyModels/Atom-350M")
model = AutoModelForCausalLM.from_pretrained("TinyModels/Atom-350M")
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 TinyModels/Atom-350M with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "TinyModels/Atom-350M"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "TinyModels/Atom-350M",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/TinyModels/Atom-350M

SGLang

How to use TinyModels/Atom-350M 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 "TinyModels/Atom-350M" \
    --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": "TinyModels/Atom-350M",
		"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 "TinyModels/Atom-350M" \
        --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": "TinyModels/Atom-350M",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use TinyModels/Atom-350M with Docker Model Runner:
```
docker model run hf.co/TinyModels/Atom-350M
```

Atom-350M / README.md

QuantaSparkLabs

Update README.md

a9a3078 verified 1 day ago

preview code

raw

history blame contribute delete

9.4 kB

	---
	license: apache-2.0
	language:
	- en
	tags:
	- assistant
	- chat
	- general-purpose
	- smollm2
	- tiny
	- efficient
	- tinymodels
	pipeline_tag: text-generation
	library_name: transformers
	---

	<div align="center">
	<svg width="100" height="100" viewBox="0 0 100 100" xmlns="http://www.w3.org/2000/svg" style="animation: float 4s ease-in-out infinite;">
	<circle cx="50" cy="50" r="45" fill="none" stroke="url(#grad1)" stroke-width="3"/>
	<circle cx="50" cy="50" r="30" fill="none" stroke="url(#grad2)" stroke-width="2" stroke-dasharray="8,6"/>
	<circle cx="50" cy="50" r="12" fill="url(#grad3)"/>
	<defs>
	<linearGradient id="grad1" x1="0%" y1="0%" x2="100%" y2="100%">
	<stop offset="0%" style="stop-color:#ff0080;stop-opacity:1" />
	<stop offset="100%" style="stop-color:#7928ca;stop-opacity:1" />
	</linearGradient>
	<linearGradient id="grad2" x1="100%" y1="0%" x2="0%" y2="100%">
	<stop offset="0%" style="stop-color:#7928ca;stop-opacity:1" />
	<stop offset="100%" style="stop-color:#ff0080;stop-opacity:1" />
	</linearGradient>
	<radialGradient id="grad3" cx="30%" cy="30%" r="70%">
	<stop offset="0%" style="stop-color:#ff0080;stop-opacity:1" />
	<stop offset="100%" style="stop-color:#7928ca;stop-opacity:1" />
	</radialGradient>
	</defs>
	</svg>
	</div>

	<h1 align="center" style="font-family: 'Orbitron', sans-serif; font-size: 3rem; letter-spacing: 2px; margin-top: 10px; background: linear-gradient(135deg, #ff0080, #7928ca); -webkit-background-clip: text; -webkit-text-fill-color: transparent;">ATOM‑350M</h1>
	<p align="center" style="font-family: 'Rajdhani', sans-serif; font-size: 1.5rem; color: #b0b0b0; margin-top: -10px;">Tiny models, <strong style="color: #ff0080;">big</strong> ideas.</p>

	<div align="center" style="margin: 30px 0;">
	<span style="background: rgba(255,0,128,0.1); border: 1px solid rgba(255,0,128,0.3); border-radius: 50px; padding: 6px 20px; font-family: 'Rajdhani', sans-serif; font-weight: 600; font-size: 0.9rem; color: #ff0080; letter-spacing: 1px;">
	⚡ BUILT BY TINYMODELS
	</span>
	</div>

	<p align="center" style="max-width: 700px; margin: 0 auto; font-family: 'Rajdhani', sans-serif; font-size: 1.1rem; color: #aaa; line-height: 1.6;">
	A lightning‑fast, open‑source AI assistant forged in the heart of the TinyModels community. Hand‑picked data, real‑world training, and a personality that doesn't feel like a corporate robot. This is our take on what a compact, genuinely useful model should be.
	</p>

	<div align="center" style="margin: 40px 0;">
	<a href="#quick-start" style="background: linear-gradient(135deg, #ff0080, #7928ca); color: white; padding: 14px 36px; border-radius: 50px; font-family: 'Orbitron', sans-serif; font-weight: 600; text-decoration: none; font-size: 1rem; box-shadow: 0 4px 20px rgba(255,0,128,0.4); transition: 0.2s;">
	⬇ GET STARTED
	</a>
	</div>

	<div style="display: flex; justify-content: center; gap: 40px; margin: 60px 0; flex-wrap: wrap;">
	<div style="text-align: center; padding: 24px; border: 1px solid rgba(255,255,255,0.08); border-radius: 20px; background: rgba(20,20,30,0.6); backdrop-filter: blur(10px); width: 160px;">
	<div style="font-family: 'Orbitron', sans-serif; font-size: 2rem; color: #ff0080;">360M</div>
	<div style="color: #888; font-family: 'Rajdhani', sans-serif;">Parameters</div>
	</div>
	<div style="text-align: center; padding: 24px; border: 1px solid rgba(255,255,255,0.08); border-radius: 20px; background: rgba(20,20,30,0.6); backdrop-filter: blur(10px); width: 160px;">
	<div style="font-family: 'Orbitron', sans-serif; font-size: 2rem; color: #7928ca;">724 MB</div>
	<div style="color: #888; font-family: 'Rajdhani', sans-serif;">Size (FP16)</div>
	</div>
	<div style="text-align: center; padding: 24px; border: 1px solid rgba(255,255,255,0.08); border-radius: 20px; background: rgba(20,20,30,0.6); backdrop-filter: blur(10px); width: 160px;">
	<div style="font-family: 'Orbitron', sans-serif; font-size: 2rem; color: #ff0080;">1 Epoch</div>
	<div style="color: #888; font-family: 'Rajdhani', sans-serif;">Focused Training</div>
	</div>
	<div style="text-align: center; padding: 24px; border: 1px solid rgba(255,255,255,0.08); border-radius: 20px; background: rgba(20,20,30,0.6); backdrop-filter: blur(10px); width: 160px;">
	<div style="font-family: 'Orbitron', sans-serif; font-size: 2rem; color: #7928ca;">Apache 2.0</div>
	<div style="color: #888; font-family: 'Rajdhani', sans-serif;">License</div>
	</div>
	</div>

	---

	<div id="quick-start"></div>
	<h2 style="font-family: 'Orbitron', sans-serif; color: #fff; font-size: 1.8rem;">🚀 QUICK START</h2>

	```python
	from transformers import AutoModelForCausalLM, AutoTokenizer

	model = AutoModelForCausalLM.from_pretrained(
	"TinyModels/Atom-350M",
	torch_dtype="auto",
	device_map="auto",
	)
	tokenizer = AutoTokenizer.from_pretrained("TinyModels/Atom-350M")

	messages = [
	{"role": "user", "content": "Explain how a bicycle stays upright in simple terms."}
	]
	prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
	inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

	outputs = model.generate(**inputs, max_new_tokens=150, temperature=0.7)
	response = tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True)
	print(response)
	```

	---

	<h2 style="font-family: 'Orbitron', sans-serif; color: #fff; font-size: 1.8rem;">🔧 OUR RECIPE</h2>

	<div style="display: grid; grid-template-columns: repeat(auto-fit, minmax(250px, 1fr)); gap: 20px; margin: 30px 0;">
	<div style="background: rgba(20,20,30,0.7); border: 1px solid rgba(255,0,128,0.2); border-radius: 16px; padding: 20px;">
	<h3 style="font-family: 'Orbitron', sans-serif; color: #ff0080; font-size: 1rem;">BASE MODEL</h3>
	<p style="color: #bbb; font-family: 'Rajdhani', sans-serif; margin: 0;">SmolLM2‑360M‑Instruct</p>
	</div>
	<div style="background: rgba(20,20,30,0.7); border: 1px solid rgba(255,0,128,0.2); border-radius: 16px; padding: 20px;">
	<h3 style="font-family: 'Orbitron', sans-serif; color: #ff0080; font-size: 1rem;">FINE‑TUNING</h3>
	<p style="color: #bbb; font-family: 'Rajdhani', sans-serif; margin: 0;">QLoRA (4‑bit, rank‑16)</p>
	</div>
	<div style="background: rgba(20,20,30,0.7); border: 1px solid rgba(255,0,128,0.2); border-radius: 16px; padding: 20px;">
	<h3 style="font-family: 'Orbitron', sans-serif; color: #ff0080; font-size: 1rem;">TRAINING DATA</h3>
	<p style="color: #bbb; font-family: 'Rajdhani', sans-serif; margin: 0;">SmolTalk (smol‑magpie‑ultra)</p>
	</div>
	<div style="background: rgba(20,20,30,0.7); border: 1px solid rgba(255,0,128,0.2); border-radius: 16px; padding: 20px;">
	<h3 style="font-family: 'Orbitron', sans-serif; color: #ff0080; font-size: 1rem;">FRAMEWORK</h3>
	<p style="color: #bbb; font-family: 'Rajdhani', sans-serif; margin: 0;">Unsloth + Hugging Face</p>
	</div>
	<div style="background: rgba(20,20,30,0.7); border: 1px solid rgba(255,0,128,0.2); border-radius: 16px; padding: 20px;">
	<h3 style="font-family: 'Orbitron', sans-serif; color: #ff0080; font-size: 1rem;">HARDWARE</h3>
	<p style="color: #bbb; font-family: 'Rajdhani', sans-serif; margin: 0;">Kaggle T4 (free tier!)</p>
	</div>
	<div style="background: rgba(20,20,30,0.7); border: 1px solid rgba(255,0,128,0.2); border-radius: 16px; padding: 20px;">
	<h3 style="font-family: 'Orbitron', sans-serif; color: #ff0080; font-size: 1rem;">EPOCHS</h3>
	<p style="color: #bbb; font-family: 'Rajdhani', sans-serif; margin: 0;">1 — lean & efficient</p>
	</div>
	</div>

	---

	<h2 style="font-family: 'Orbitron', sans-serif; color: #fff; font-size: 1.8rem;">🤍 THIS IS OURS</h2>

	<p style="font-family: 'Rajdhani', sans-serif; color: #bbb; font-size: 1.1rem; line-height: 1.6; max-width: 800px;">
	<strong style="color: #ff0080;">We didn't just download a model and slap a name on it.</strong> TinyModels hand‑picked the training data, configured the QLoRA adapters, and ran the entire training pipeline ourselves. Atom‑350M is <em>our</em> interpretation of a small, helpful, open assistant — built with pride, released with no strings attached.
	</p>

	<p style="font-family: 'Rajdhani', sans-serif; color: #aaa; font-size: 1rem; margin-top: 10px;">
	If you do something cool with it, just give us a shout‑out. We'd love to see what you build.
	</p>

	---

	<h2 style="font-family: 'Orbitron', sans-serif; color: #fff; font-size: 1.8rem;">📜 LICENSE</h2>

	<p style="font-family: 'Rajdhani', sans-serif; color: #bbb; font-size: 1.1rem;">
	Apache 2.0 — free for research, commercial use, and even intergalactic travel.
	</p>

	<div align="center" style="margin: 60px 0 20px;">
	<p style="font-family: 'Rajdhani', sans-serif; color: #555; font-size: 0.9rem;">
	Crafted by <strong style="color: #ff0080;">TinyModels</strong> — small models, big ambitions.
	</p>
	</div>

	<style>
	@import url('https://fonts.googleapis.com/css2?family=Orbitron:wght@500;600;700&family=Rajdhani:wght@400;500;600&display=swap');

	body {
	background: #0c0c14;
	color: #ccc;
	}

	@keyframes float {
	0% { transform: translateY(0); }
	50% { transform: translateY(-8px); }
	100% { transform: translateY(0); }
	}

	a {
	color: #ff0080;
	text-decoration: none;
	}
	</style>