import onnxruntime_genai as og

# Load the model

model = og.Model("haydso/SmolLM2-1.7B-ONNX")

tokenizer = og.Tokenizer(model)

# Generate text

prompt = "Your prompt here"

tokens = tokenizer.encode(prompt)

params = og.GeneratorParams(model)

params.set_search_options(max_length=200)

params.input_ids = tokens

generator = og.Generator(model, params)

while not generator.is_done():

    generator.compute_logits()

    generator.generate_next_token()

output_tokens = generator.get_sequence(0)

text = tokenizer.decode(output_tokens)

print(text)

Original Model Card

SmolLM2

Model Summary
Evaluation
Limitations
Training
License
Citation

Model Summary

SmolLM2 is a family of compact language models available in three size: 135M, 360M, and 1.7B parameters. They are capable of solving a wide range of tasks while being lightweight enough to run on-device. More details in our paper: https://arxiv.org/abs/2502.02737v1

The 1.7B variant demonstrates significant advances over its predecessor SmolLM1-1.7B, particularly in instruction following, knowledge, reasoning, and mathematics. It was trained on 11 trillion tokens using a diverse dataset combination: FineWeb-Edu, DCLM, The Stack, along with new mathematics and coding datasets that we curated and will release soon. We developed the instruct version through supervised fine-tuning (SFT) using a combination of public datasets and our own curated datasets. We then applied Direct Preference Optimization (DPO) using UltraFeedback.

The instruct model additionally supports tasks such as text rewriting, summarization and function calling thanks to datasets developed by Argilla such as Synth-APIGen-v0.1. You can find the SFT dataset here: https://huggingface.co/datasets/HuggingFaceTB/smoltalk and finetuning code in the alignement handbook.

For more details refer to: https://github.com/huggingface/smollm. You will find pre-training, post-training, evaluation and local inference code.

How to use

pip install transformers

Running the model on CPU/GPU/multi GPU

Using full precision

python

# pip install transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
checkpoint = "HuggingFaceTB/SmolLM2-1.7B"
device = "cuda" # for GPU usage or "cpu" for CPU usage
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
# for multiple GPUs install accelerate and do `model = AutoModelForCausalLM.from_pretrained(checkpoint, device_map="auto")`
model = AutoModelForCausalLM.from_pretrained(checkpoint).to(device)
inputs = tokenizer.encode("Gravity is", return_tensors="pt").to(device)
outputs = model.generate(inputs)
print(tokenizer.decode(outputs[0]))

Using torch.bfloat16

python

# pip install accelerate
# for fp16 use `torch_dtype=torch.float16` instead
model = AutoModelForCausalLM.from_pretrained(checkpoint, device_map="auto", torch_dtype=torch.bfloat16)
inputs = tokenizer.encode("Gravity is", return_tensors="pt").to("cuda")
outputs = model.generate(inputs)
print(tokenizer.decode(outputs[0]))

bash

>>> print(f"Memory footprint: {model.get_memory_footprint() / 1e6:.2f} MB")
Memory footprint: 3422.76 MB

Evaluation

In this section, we report the evaluation results of SmolLM2. All evaluations are zero-shot unless stated otherwise, and we use lighteval to run them.

Base Pre-Trained Model

Metric	SmolLM2-1.7B	Llama-1B	Qwen2.5-1.5B	SmolLM1-1.7B
HellaSwag	68.7	61.2	66.4	62.9
ARC (Average)	60.5	49.2	58.5	59.9
PIQA	77.6	74.8	76.1	76.0
MMLU-Pro (MCF)	19.4	11.7	13.7	10.8
CommonsenseQA	43.6	41.2	34.1	38.0
TriviaQA	36.7	28.1	20.9	22.5
Winogrande	59.4	57.8	59.3	54.7
OpenBookQA	42.2	38.4	40.0	42.4
GSM8K (5-shot)	31.0	7.2	61.3	5.5

Instruction Model

Metric	SmolLM2-1.7B-Instruct	Llama-1B-Instruct	Qwen2.5-1.5B-Instruct	SmolLM1-1.7B-Instruct
IFEval (Average prompt/inst)	56.7	53.5	47.4	23.1
MT-Bench	6.13	5.48	6.52	4.33
OpenRewrite-Eval (micro_avg RougeL)	44.9	39.2	46.9	NaN
HellaSwag	66.1	56.1	60.9	55.5
ARC (Average)	51.7	41.6	46.2	43.7
PIQA	74.4	72.3	73.2	71.6
MMLU-Pro (MCF)	19.3	12.7	24.2	11.7
BBH (3-shot)	32.2	27.6	35.3	25.7
GSM8K (5-shot)	48.2	26.8	42.8	4.62

Limitations

SmolLM2 models primarily understand and generate content in English. They can produce text on a variety of topics, but the generated content may not always be factually accurate, logically consistent, or free from biases present in the training data. These models should be used as assistive tools rather than definitive sources of information. Users should always verify important information and critically evaluate any generated content.

Training

Model

Architecture: Transformer decoder
Pretraining tokens: 11T
Precision: bfloat16

Hardware

GPUs: 256 H100

Software

Training Framework: nanotron

License

Apache 2.0

Citation