QuantizedShardedToAllLinear

Quantized distributed linear layer (gather).

Abstract Signature:

QuantizedShardedToAllLinear(in_features: int, out_features: int)

PyTorchApple MLX

PyTorch

API: —

Strategy: Custom / Partial

Apple MLX

API: mlx.nn.QuantizedShardedToAllLinear

Strategy: Direct Mapping

Official Docs ↗