QuantizedEmbedding

Embedding layer with quantized weights.

Abstract Signature:

QuantizedEmbedding(num_embeddings: int, embedding_dim: int)

PyTorchApple MLX

PyTorch

API: —

Strategy: Custom / Partial

Apple MLX

API: mlx.nn.QuantizedEmbedding

Strategy: Direct Mapping

Official Docs ↗