Apache SINGA is an open-source distributed deep learning library for training machine learning and deep learning models across multiple GPUs. It handles core deep learning functionality including automatic gradient calculation, computational graph optimization, distributed training, and standard optimizers (Adam, SGD with momentum, RMSProp, AdaGrad), with support for ONNX format interoperability and domain-specific deep learning models.
Real-world deployment across both academic and enterprise settings – the National University of Singapore, Alibaba, and Citigroup are all listed users – is what separates SINGA from an academic proof-of-concept that never left the lab. As an Apache Top Level Project on version 5.1.0 with active GitHub development, it’s a genuinely maintained framework, not abandoned open-source infrastructure someone forgot to archive.
This is a code library for teams building and training their own deep learning models, not a hosted product with a dashboard or signup – install via Pip, Conda, Docker, or source compilation and integrate directly into your own ML infrastructure. It’s free and open-source under the Apache license.
