`neural_compressor.adaptor.tf_utils.graph_rewriter.int8.fuse_matmul_redundant_dequantize`

Fuse QuantizedMatMul with redundant Dequantize Graph Rewriter.

Module Contents

Classes

FuseMatMulRedundantDequantizeTransformer

Fuse _QuantizedMatMul with the successor Dequantize Op.

class neural_compressor.adaptor.tf_utils.graph_rewriter.int8.fuse_matmul_redundant_dequantize.FuseMatMulRedundantDequantizeTransformer(model, device='cpu')[source]: Fuse _QuantizedMatMul with the successor Dequantize Op.