smile

The smile-deep module provides idiomatic Java API for deep learning on the JVM while still reaching CPU, CUDA, and MPS backends by wrapping the PyTorch / LibTorch C++ runtime. It also provides tiktoken BPE tokenizer, LLaMA-3 inference, EfficientNet-V2, and an image classification pipeline out of the box.

SMILE Studio is an IDE for machine learning and data science. It combines an interactive notebook and AI-powered agents in a single, modern desktop application. The AutoML Agent is capable of automating the end-to-end workflow of building, training, and deploying machine learning models. You may try it with the prebuilt package. You may also guide the agent to build ML/AI solutions step by step with natural language instructions. - Data loading from CSV, ARFF, JSON, Avro, Parquet, Iceberg, SQL - Advanced interactive data visualization - Model training, evaluation, and ensembling - Inference server management

SMILE Serve is a production-ready inference server built on [Quarkus](https://quarkus.io/) that brings together three complementary inference capabilities on the JVM: - **Classic ML**: `/api/v1/models` for serialized SMILE models (`.sml`) - **ONNX Runtime**: `/api/v1/onnx` for any model in the ONNX open format (`.onnx`) - **LLM Chat**: `/api/v1/chat` for Llama 3 chat completions A React-based web UI is bundled and served from the same process.

> I don't think it's right to recommend that new users move away from the package because of licensing issues I was going to chime in to agree but then I saw how this was done - a completely innocuous looking commit: https://github.com/haifengl/smile/commit/6f22097b233a3436519... And literally no mention in the release notes: https://github.com/haifengl/smile/releases/tag/v3.0.0 I think if you are going to change license, especially in a way that makes it less permissive, you need to be super open and clear about both the fact you are doing it and your reasons for that. This is done so silently as to look like it is intentionally trying to mislead and trick people. So maybe I wouldn't say to move away because of the specific license, but it's legitimate to avoid something when it's so clearly driven by a single entity and that entity acts

> Too bad there's not really a ML/AI ecosystem around (besides Spark, which seems to be losing favor partially because of its embarrassing performance), because using Python is such a drag. I think it's not that bad. Sure, it loses to Python, but that's some tough competition. Personally, I find that smile[1] covers most use cases. Breeze[2] also has a lot of love, but I'm not much of a fan of the `UFunc` abstraction. I'm also really excited about polynote[3]. It's still a bit green and only supports Scala 2.11/2.12, but it's such a joy to be able to use Scala in notebooks :). You also get Python interop, in case you need to use some specific python library. P.S.: I hope that Scala's ML/AI libraries other than Spark keep growing, because "embarassing" is a nice way to describe it's performance on everything that's not a "really-huge datas

Looks very interesting! Another Java Library for ML is Smile(https://github.com/haifengl/smile). I deeply recommend it!

It looks to be a custom function for parsing JSON arrays: https://github.com/haifengl/smile/blob/5bf4075b2978af68b31c7... The benefit is that the type will be known to be JSArray without need for runtime casting (at least, outside the scope of the function). For comparison, `json` returns a JSObject. I honestly haven't seen this before working with Scala, and it's defined just for the smile repository.

I wish you knew the history of H2O's attempts to make a decent product. H2O, the open-source project, was largely created by H2O the company's technical co-founder, Cliff Click, who was forced out of the company more than a year ago for internal political reasons. With him, H2O lost much of its ability to develop and extend its code base, particularly since Cliff built many things in house that might have better been dependencies. That's why Steam never happened. Now H2O is mostly marketing, an overfunded startup in slow decline that was forced to fire its sales team last year because, well, nobody wants their stuff. https://venturebeat.com/2016/09/24/machine-learning-startup-... H2O is not nearly as fast as SMILE by most metrics: http://haifengl.github.io/smile/ https://github.com/haifengl/smile But what I'm saying is that true gains are

Hi, skymind cofounder here. To offer some context on our book there: We have appendixes covering some of the fundamental concepts such as linear algebra and statistics. For other machine learning libraries in java: https://github.com/haifengl/smile http://knime.org http://rapidminer.com