2 projects
zadapter
JAX/Flax-native bottleneck FFN adapter with Activation Function Annealing (AFA), tensor-parallel sharding, and zenbit NF4 quantization integration. TPU-optimized PEFT alternative to LoRA.
zenbit
TPU sharding (data-parallel, tensor-parallel) and quantization (int8, NF4) utilities for PyTorch/XLA training
