Merge Reports

After collecting the raw profile data, you can use the Merge Reports function to parse and display it in MindStudio. To merge reports, perform the following steps:

1. On the top menu bar, choose Ascend > System Profiler > Merge Reports. The Merge Reports configuration window is displayed.
2. Configure the input directory, output directory, and output .json file name. See Figure 1.
   Figure 1 Merge Reports configuration
   

   Table 1 Merge Reports parameters

   Parameter	Description
   Input directory	Data source path. Set it to the data output path (parent directory of the PROF_XXX directory) specified by the msprof command or directly to the PROF_XXX directory.
   Output directory	Output path.
   Output Json name	Name of the output .json file, which is user-defined. The .json file is automatically named in the following format after being opened using Merge Reports: {user-defined_json_file_name}_{device_id}_{model_id}_{iter_id}.json, in which {device_id} indicates the device ID, {model_id} indicates the model ID, and {iter_id} indicates the ID of an iteration.

   

   • For details about how to use the msprof command, see "Advanced Features > Profile Data Collection with the msprof Command Line" in the Profiling Instructions.
   • If the msprof command is run in a multi-device environment, how many devices there are, how many PROF_XXX files will be generated in the data output path.
   • If you execute the msprof command for multiple times in the same path, multiple PROF_XXX files will be generated in the data output path.
   • If the size of the .json file exceeds the maximum cache for automatic result display of Merge Reports, the file may fail to be automatically displayed. In this case, manually open it.
   • Profile data (the parent directory of PROF_XXX is specified) in the training cluster scenario can be imported and displayed. For details about the display interface, see Analysis Summary, Cluster Iteration Analysis, and Data Preparation. For details about profile data collection in cluster scenarios, see "Advanced Functions > Profiling in Cluster Scenarios" in the Profiling Instructions.