Introduction

To help users quickly call MindStudio functions during model inference, an integrated Ascend Model Inference Tool (AMIT) is provided to allow accesses to a range of model analysis, model conversion, offline model dump, accuracy comparison, and performance analysis tools through a single, unified portal, thus enhancing the development efficiency. The AMIT GUI also provides step-by-step guidance to facilitate users' understanding of the inference development process and make it easier to use.

Figure 1 Training process

• Environment setup: Prepare an inference server and install the CANN Toolkit and MindStudio.
• Data preparation: Prepare the sample project and original model required for inference.
• Model analysis: Analyze operators that are not supported by a model and allows users to query information about operators supported by the current CANN version and models supported by ModelZoo.
• Model conversion: Use the ATC tool to generate an offline OM model.
• Application project inference: Build and run the offline OM model.
• Offline model dump: Generate dump data based on the offline OM model.
• Accuracy comparison: Compare the accuracy difference of computation results of Huawei proprietary operators and third-party equivalents based on the dump data of the offline OM model.
• Performance analysis: Use the Profiling tool to collect profile data and perform simple profile data analysis.