fusion_op (Operator Fusion Information)

The data about operator fusion (before and after) does not contain the timeline information. The summary information is summarized in the fusion_op_*.csv file.

This profile data file does not exist in single-operator scenarios (such as the PyTorch scenario).

Applicability

For details about the mapping between AI processors and Ascend products, see Ascend Product Form Description.

AI Processor Type	Supported
Atlas 350 Accelerator Card	√
Atlas A3 training product/Atlas A3 inference product	√
Atlas A2 training product/Atlas A2 inference product	√
Atlas 200I/500 A2 inference product	√
Atlas inference product	√
Atlas training product	√
Atlas 350 Accelerator Card	√

fusion_op_*.csv File

The following shows the file content format.

Figure 1 fusion_op_*.csv

**Table 1** Field description
Field	Description
Device_id	Device ID. For data collected from the host, the value is host.
Model Name	Model name.
Model ID	Model ID.
Fusion Op	Fused operator name.
Original Ops	Names of the base operators.
Memory Input(KB)	Input tensor memory size, in KB.
Memory Output(KB)	Output tensor memory size, in KB.
Memory Weight(KB)	Weight memory size, in KB.
Memory Workspace(KB)	Workspace memory size, in KB.
Memory Total(KB)	Total memory size, in KB, calculated as the sum of Memory Input, Memory Output, Memory Weight, and Memory Workspace.

Parent topic: Profile Data File References