Common Parameters

The following table describes commands' common parameters.

Level-1 Parameter

Level-2 Parameter

Mandatory

Description

Supported Function

[-dg, --dg, --diagnosis]

[-dg, --dg, --diagnosis]

Yes

Performs a fault diagnosis of the entire card.

If multiple level-2 parameters such as -i and -r need to be appended to ascend-dmi --dg, the sequence of these parameters is flexible. The sequence does not affect the command output.

Fault diagnosis:

  • One-click diagnosis
  • CANN-driver compatibility diagnosis
  • Chip diagnosis
  • Network health diagnosis
  • On-chip memory diagnosis
  • Driver health diagnosis
  • Eye diagram diagnosis
  • Bandwidth diagnosis
  • NIC diagnosis
  • PRBS stream diagnosis
  • Computing power diagnosis
  • AI Core stress test
  • AI Core diagnosis
  • AI CPU stress test
  • Power consumption stress test
  • One-click on-chip memory stress test
  • On-chip memory stress test
  • On-chip memory high-risk address stress test
  • P2P stress test
  • DSA stress test

[-se, --scene, --se]

No

Specifies a diagnosis scenario. Currently, the following scenarios are supported:

  • healthCheck
  • performanceCheck
  • stressTest

-

[-i, --items]

No

Specifies the diagnosis check item.

  • You can specify one or more items among driver, cann, device, network, bandwidth, aiflops, hbm/chipMemory, signalQuality, and random. Separate the items with commas (,).
  • If this parameter is not specified, check items except AI Core, PRBS, EDP, TDP, AI CPU, and NIC are diagnosed by default.
  • If -i specifies aicpu, aicore, prbs, edp, tdp, and random, other diagnostic items cannot be used.

Fault diagnosis:

  • CANN-driver compatibility diagnosis
  • Chip diagnosis
  • Network health diagnosis
  • On-chip memory diagnosis
  • Driver health diagnosis
  • Eye diagram diagnosis
  • Bandwidth diagnosis
  • NIC diagnosis
  • PRBS stream diagnosis
  • Computing power diagnosis
  • AI Core stress test
  • AI Core diagnosis
  • AI CPU stress test
  • Power consumption stress test
  • One-click on-chip memory stress test
  • On-chip memory stress test
  • On-chip memory high-risk address stress test
  • P2P stress test
  • DSA stress test

[-r, --result]

No

Specifies the path for saving fault diagnosis results and information collection results, for example, /test. The specified path must meet security requirements and cannot contain the wildcard (*).

  • If you specify a path for saving the result file, create an ascend_check folder in the specified path. The path specified by the root user is created in the root directory, and the path specified by a non-root user is created in the $HOME directory. If no path is specified, the result file is saved in the default path. For the root user, the path is /var/log/ascend_check. For a non-root user, the path is $HOME/var/log/ascend_check.
  • To prevent the permission on the result saving directory from being modified, you can set the permission on ascend_check to 700 for security purposes.

Fault diagnosis:

  • One-click diagnosis
  • CANN-driver compatibility diagnosis
  • Chip diagnosis
  • Network health diagnosis
  • On-chip memory diagnosis
  • Driver health diagnosis
  • Eye diagram diagnosis
  • Bandwidth diagnosis
  • NIC diagnosis
  • PRBS stream diagnosis
  • Computing power diagnosis
  • AI Core stress test
  • AI Core diagnosis
  • AI CPU stress test
  • Power consumption stress test
  • One-click on-chip memory stress test
  • On-chip memory stress test
  • On-chip memory high-risk address stress test
  • P2P stress test
  • DSA stress test

[-s, --stress]

No

Performs a stress test.

This parameter is mandatory when the functions listed in the right column are enabled.

Fault diagnosis:

  • AI Core stress test
  • AI CPU stress test
  • Power consumption stress test
  • One-click on-chip memory stress test
  • On-chip memory stress test
  • On-chip memory high-risk address stress test
  • P2P stress test

[-p, --path]

-

No

Requires the installation path of the software package.

  • If you do not use the default installation path when installing the target package, this parameter must be set to the actual installation path.
  • The specified path must meet security requirements and cannot contain the wildcard (*).
  • If this parameter is not specified and the package is installed by the root user, the default path /usr/local/Ascend is used.
  • If the check items specified by [-i, --items] do not contain cann, leave this parameter unspecified.

Information query:

  • Software-hardware compatibility test

Fault diagnosis:

  • One-click diagnosis
  • CANN-driver compatibility diagnosis

[-d, --device]

-

No

Specifies the ID of the device to be diagnosed. The device ID is the logic ID of the Ascend NPU.

  • Except for bandwidth and compute power tests, which can only specify a single device ID, other scenarios allow specifying one or more device IDs. When multiple are specified, they should be separated by commas (',').
  • If no device ID is specified, the bandwidth and computing power test results of device 0 are returned in H2D/D2H/D2D mode for products except the Atlas A3 training product and Atlas A3 inference products. In other scenarios, the diagnosis results of all devices are returned.
  • In this document, the input or output device ID is the logic ID. You can run the npu-smi info -m command to obtain the logic ID by viewing the value of Chip Logic ID on the GUI. The NPU ID is the physical chip ID.

Performance test:

  • Bandwidth test
  • SuperPoD P2P bandwidth test
  • Computing power test
  • Eye diagram test
  • Stream test (one-click/custom stream generation)

Fault diagnosis:

  • Chip diagnosis
  • Network health diagnosis
  • On-chip memory diagnosis
  • Eye diagram diagnosis
  • Bandwidth diagnosis
  • NIC diagnosis
  • PRBS stream diagnosis
  • Computing power diagnosis
  • AI Core stress test
  • AI Core diagnosis
  • AI CPU stress test
  • One-click on-chip memory stress test
  • On-chip memory stress test
  • On-chip memory high-risk address stress test
  • P2P stress test
  • DSA stress test

NPU environment restoration

[-fmt, --fmt, --format]

-

No

Specifies the output format. The value can be normal or json.

  • If this parameter is not specified, the default value normal is used.
  • If the json format is specified by [-fmt, --fmt, --format], the fault diagnosis result is saved in the ascend_check/environment_check_before.txt file. If the json format is not specified, the fault diagnosis result is not saved.
  • If a diagnosis item fails, the returned JSON result is displayed as JSON Example Returned When Diagnosis Fails.

All functions except the power consumption test

[-q, --quiet]

-

No

  • If this parameter is specified, no foolproof message is displayed. By default, this operation is allowed. To use --diagnosis, this parameter must be used together with bandwidth, aiflops, hbm, aicore, prbs, tdp, edp, aicpu, nic, and random specified by -i.
  • If this parameter is not specified, you need to enter Y or N (y or n) to confirm whether to perform the test.

Performance test:

  • Bandwidth test
  • SuperPoD P2P bandwidth test
  • Power consumption test
  • Computing power test
  • Stream test (one-click/custom stream generation)
  • Latency test

Fault diagnosis:

  • One-click diagnosis
  • Bandwidth diagnosis
  • NIC diagnosis
  • PRBS stream diagnosis
  • Computing power diagnosis
  • AI Core stress test
  • AI Core diagnosis
  • AI CPU stress test
  • Power consumption stress test
  • One-click on-chip memory stress test
  • On-chip memory stress test
  • On-chip memory high-risk address stress test
  • P2P stress test
  • DSA stress test

NPU environment restoration

[-h, --help]

-

No

Displays the help information of a specified function of Ascend DMI.

All