Software

See also this list of ML related software.

• Pytorch
  • Huggingface
  • LLMs
    • Unsloth github Incredibly fast. As of March 2025, this is the fastest single GPU framework for LLMs
    • llama.cpp
    • LLVM
  • Multi-GPU training
    • Note: recommendations can change frequently (every few months it seems)
    • As of March 2025: Nilay recommends DeepSpeed, which is integrated with HuggingFace Accelerate. See here: HuggingFace Accelerate - DeepSpeed

• Fairseq (pytorch)
  • Tutorial on how to add a new model to fairseq: Tutorial: Simple LSTM Nilay used this tutorial when adding his model to fairseq
  • Translation example
  • CLI Documentation
• OpenNMT (pytorch)
• NLP Architect (Intel's NLP software)
• MiniTorch (educational)
• JAX + Flax
• Tensorflow
  • Tensorflow - Fine-tuning a BERT model
• Keras
  • Fine-tune a pretrained Huggingface model in Keras and Huggingface: Fine-tuning a model with Keras
• Export formats
  • ONNX The industry de-facto standard. See this video for how to import into Mathematica. Videos: PyTorch to Tensorflow using ONNX PyTorch, TensorFlow, Keras Model File Conversion

See MT Software.

• Nematus (Tensorflow) Rico Sennrich's NMT system
• Marian paper (C++) Optimized implementations of RNNs and Transformer
• Sockeye (MXNet) Amazon's NMT system

• Weights and biases
• Tensorboard PyTorch tutorial PyTorch Recipe
• Data Version Control (DVC)

• Awesome Machine Learning A curated list of awesome machine learning frameworks, libraries and software.

• Hugging Face
• Transformers - Software
• NLP - Software
• PyTorch
• Tensorflow