`neural_compressor.ux.utils.templates.metric`¶

Metric class.

Module Contents¶

Classes¶

Metric represents data which is sent from Optimization and Benchmark.

class neural_compressor.ux.utils.templates.metric.Metric¶

Bases: neural_compressor.ux.utils.json_serializer.JsonSerializer

Metric represents data which is sent from Optimization and Benchmark.

property accuracy: float | None¶: Accuracy for model.

property latency: float | None¶: Latency for model.

property throughput: float | None¶: Throughput for model.

property acc_input_model: float | None¶: Accuracy for input model.

property acc_optimized_model: float | None¶: Accuracy for input model.

property size_input_model: float | None¶: Model size for float32.

property size_optimized_model: float | None¶: Get optimized model size.

property path_optimized_model: str | None¶: Get optimized model path.

property optimization_time: float | None¶: Optimization time.

property profiling_data: List[dict] | None¶: Get profiling data.

insert_data(attribute: str, value: str) → None¶: Set attribute value.

static calculate_throughput(value: float) → float¶

Calculate throughput based on latency.

Right now 1000 represents number of images in dataset. TODO: change 1000 to the batch size when Benchmark is ready