haifengl/smile

Statistical Machine Intelligence & Learning Engine

What it solves

SMILE is a high-performance machine learning framework designed for the JVM, providing a comprehensive suite of tools for data science, statistical analysis, and deep learning. It bridges the gap between the flexibility of data science tools and the the production-ready performance of Java, Scala, and Kotlin.

How it works

The framework is organized into several specialized modules:

  • Core ML: Implements a wide array of classification, regression, clustering, and manifold learning algorithms (e.g., Random Forest, SVM, t-SNE, UMAP).
  • Deep Learning & LLMs: Uses a LibTorch backend for GPU/CPU tensor operations and provides a full stack for LLaMA-3 inference, including BPE tokenizers and an OpenAI-compatible REST server.
  • NLP: Provides tools for text normalization, POS tagging, stemming, and relevance ranking.
  • Base Foundation: Handles the underlying math, linear algebra, and data structures like DataFrames.
  • SMILE Studio: An agentic IDE that allows users to interact with data using natural language via Python, Java, or Scala.

Who it’s for

It is primarily for developers and data scientists working within the JVM ecosystem (Java, Scala, Kotlin) who need a robust, integrated machine learning library that can be deployed in production environments or integrated with Apache Spark.

Highlights

  • JVM Native: Full support for Java, Scala, and Kotlin with idiomatic APIs.
  • LLM Capabilities: Built-in LLaMA-3 inference and an OpenAI-compatible inference server.
  • Comprehensive Algorithm Suite: Extensive coverage of traditional ML, deep learning, and NLP.
  • Agentic IDE: Includes SMILE Studio for natural language data interaction.
  • Visualization: Integrated Swing-based interactive plots and declarative Vega-Lite charts.

Related

  • Project
    huggingface/lighteval

    A comprehensive toolkit from Hugging Face for evaluating LLMs across multiple backends with support for over 1,000 benchmarks and custom metrics.

  • Project
    ModelTC/LightLLM

    LightLLM is a Python‑based, high‑performance inference and serving framework for large language models. It combines fast kernels (FlashAttention, Triton) with lightweight, token‑level KV‑cache management, supports multi‑GPU scaling, and adds research‑grade features like constrained decoding and SLA‑aware request scheduling.

  • Project
    UbiquitousLearning/mllm

    A lightweight multimodal LLM inference engine optimized for mobile and edge devices, supporting various NPUs and GPUs with advanced quantization.

  • Project
    EvolvingLMMs-Lab/lmms-engine

    A unified, high-performance training engine for multimodal models that integrates distributed training, kernel fusion, and advanced optimizers to enable scalable pretraining and fine-tuning.

  • Project
    xLLM-AI/xllm

    xLLM is an open‑source C++ inference framework that delivers high‑throughput, low‑latency LLM serving on Chinese AI accelerators (Ascend, Cambricon, Moore Threads, etc.). It separates a brpc‑based service layer from a hardware‑optimized engine, includes hybrid KV‑cache management via Mooncake, and provides ready‑to‑run deployment scripts for models such as GLM‑5.3‑Flash and MiniMax‑M3. The project is Apache‑2.0 licensed and backed by JD.com and several Chinese university labs.