Release v0.2.0

Added

• Quantized layers with different acceleration options
  • QConv (binary, quantized) - CPU, Cutlass
  • QLinear (binary, quantized, mixed bit-width) - CUDA, Cutlass, MPS
  • QEmbedding (binary)
• Optimizer(s) for quantized layers
  • Hybrid optimizer diode_beta based on Diode v1 (binary) and AdamW (quantized) for memory-efficient training
  • Initial support for galore projection
• Examples
  • MNIST training script with and without PyTorch Lightning