`neural_compressor.tensorflow.quantization.utils.graph_rewriter.qdq.insert_qdq_pattern`

Insert QDQ pattern Graph Rewriter.

Module Contents

Classes

GenerateGraphWithQDQPattern

Insert Q/DQ pairs before quantizable ops.

class neural_compressor.tensorflow.quantization.utils.graph_rewriter.qdq.insert_qdq_pattern.GenerateGraphWithQDQPattern(model, calibration_data, op_wise_config, fake_quant, fp32_ops, bf16_ops, quantized_nodes, device, performance_only, itex_mode, llm_weight_minmax)[source]: Insert Q/DQ pairs before quantizable ops.