convert_model

Applicability

Product	Supported
Atlas A3 training series products/Atlas A3 inference series products	√
Atlas A2 training products/Atlas A2 inference products	√
Atlas 200I/500 A2 inference product	√
Atlas inference series products	√
Atlas training products	√

Description

Converts a TensorFlow model based on the user-defined quantization factors into a model serving for both accuracy simulation in the TensorFlow environment and inference on the Ascend AI Processor.

Prototype

convert_model(pb_model, outputs, record_file, save_path)

Parameters

Parameter	Input/Output	Description
pb_model	Input	Original .pb model file for adaptation. A string.
outputs	Input	List of output operators in a graph. A list.
record_file	Input	Path of the quantization factor record file (.txt) computed by the user. A string.
save_path	Input	Model save path. Must include the prefix of the model name, for example, **./quantized_model/model***. A string.

Returns

None

Restrictions

The user model must match the quantization factor record file. For example, if "Conv+BN" fusion is performed before computing the quantization factors of Conv, "Conv+BN" fusion should also be performed in advance in the TensorFlow model to be converted.
The format and content of the quantization factor record file must comply with the AMCT requirements defined in Record Files.
AMCT quantizes the following layers: Conv2D, MatMul, DepthwiseConv2dNative (dilation = 1), Conv2DBackpropInput (dilation = 1), and AvgPool.
This API supports the fusion of the "Conv+BN", "Depthwise_Conv+BN", and "Group_Conv+BN" in the user model. Per-layer fusion switch is supported.
Adaptation of only an original floating-point model is supported. The model cannot be adapted if the input model contains any of the following custom quantization layers: QuantIfmr, QuantArq, SearchN, AscendQuant, AscendDequant, AscendAntiQuant, and AscendWeightQuant.

Example

import amct_tensorflow as amct
convert_model(pb_model='./user_model.pb',
              outputs=["model/outputs"],
              record_file='./record_quantized.txt',
              save_path='./quantized_model/model')

Flush file: a .pb model file that can be used for accuracy simulation in the TensorFlow environment or offline inference on the Ascend AI Processor.

When adaptation is performed again, the preceding files output by the API will be overwritten.

Parent topic: Model Adaptation APIs