SmolLM 135M
par HuggingFaceTB
Open source · 148k downloads · 255 likes
3.0
(255 avis)ChatAPI & LocalÀ propos
SmolLM 135M est un petit modèle de langage performant, conçu pour offrir une alternative efficace et accessible aux grands modèles. Entraîné sur un corpus éducatif varié et de haute qualité, il excelle dans la compréhension et la génération de texte en anglais, tout en restant léger et optimisé pour fonctionner sur des configurations matérielles modestes. Ses capacités couvrent des tâches comme la réponse à des questions, la synthèse d’informations ou la rédaction de contenus simples, bien qu’il doive être utilisé avec discernement en raison de possibles inexactitudes ou biais. Idéal pour des applications nécessitant un équilibre entre performance et ressources limitées, il se distingue par sa rapidité et son adaptabilité, notamment grâce à des versions quantifiées pour une utilisation encore plus légère.
Documentation
SmolLM
Table of Contents
Model Summary
SmolLM is a series of state-of-the-art small language models available in three sizes: 135M, 360M, and 1.7B parameters. These models are built on Cosmo-Corpus, a meticulously curated high-quality training dataset. Cosmo-Corpus includes Cosmopedia v2 (28B tokens of synthetic textbooks and stories generated by Mixtral), Python-Edu (4B tokens of educational Python samples from The Stack), and FineWeb-Edu (220B tokens of deduplicated educational web samples from FineWeb). SmolLM models have shown promising results when compared to other models in their size categories across various benchmarks testing common sense reasoning and world knowledge. For detailed information on training, benchmarks and performance, please refer to our full blog post.
This is the SmolLM-135M
Generation
Bash
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/SmolLM-135M"
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("def print_hello_world():", return_tensors="pt").to(device)
outputs = model.generate(inputs)
print(tokenizer.decode(outputs[0]))
Bash
>>> print(f"Memory footprint: {model.get_memory_footprint() / 1e6:.2f} MB")
Memory footprint: 12624.81 MB
- Using
torch.bfloat16
Python
# pip install accelerate
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
checkpoint = "HuggingFaceTB/SmolLM-135M"
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
# for fp16 use `torch_dtype=torch.float16` instead
model = AutoModelForCausalLM.from_pretrained(checkpoint, device_map="auto", torch_dtype=torch.bfloat16)
inputs = tokenizer.encode("def print_hello_world():", 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: 269.03 MB
Quantized Versions through bitsandbytes
- Using 8-bit precision (int8)
Python
# pip install bitsandbytes accelerate
from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
# to use 4bit use `load_in_4bit=True` instead
quantization_config = BitsAndBytesConfig(load_in_8bit=True)
checkpoint = "HuggingFaceTB/SmolLM-135M"
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
model = AutoModelForCausalLM.from_pretrained(checkpoint, quantization_config=quantization_config)
inputs = tokenizer.encode("def print_hello_world():", 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")
# load_in_8bit
Memory footprint: 162.87 MB
# load_in_4bit
>>> print(f"Memory footprint: {model.get_memory_footprint() / 1e6:.2f} MB")
Memory footprint: 109.78 MB
Limitations
While SmolLM models have been trained on a diverse dataset including educational content and synthetic texts, they have limitations. The 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. For a more comprehensive discussion of the models' capabilities and limitations, please refer to our full blog post..
This repository contains a converted version of our latest trained model. We've noticed a small performance difference between this converted checkpoint (transformers) and the original (nanotron). We're currently working to resolve this issue.
Training
Model
- Architecture: For architecture detail, see the blog post.
- Pretraining steps: 600k
- Pretraining tokens: 600B
- Precision: bfloat16
- Tokenizer: HuggingFaceTB/cosmo2-tokenizer
Hardware
- GPUs: 64 H100
Software
- Training Framework: Nanotron
License
Citation
Bash
@misc{allal2024SmolLM,
title={SmolLM - blazingly fast and remarkably powerful},
author={Loubna Ben Allal and Anton Lozhkov and Elie Bakouch and Leandro von Werra and Thomas Wolf},
year={2024},
}
Liens & Ressources