- class insights.parsers.nvidia.NvidiaGPUInfo(index, model, uuid, memory_total)
Bases:
tuplenamedtuple: Represents the information parsed from
nvidia-smi --query-gpucommand output.- index
- memory_total
- model
- uuid
- class insights.parsers.nvidia.NvidiaSmiActiveClocksEventReasons(context)[source]
Bases:
Parser,listParser for the output of command /usr/bin/nvidia-smi --query-gpu=name,clocks_event_reasons.active --format=csv,noheader. This command lists each of the NVIDIA GPUs in the system, along with their names and bitmasks of active clock event reasons.
- Raises:
ParseException -- When run into an unparsable line
SkipComponent -- When content is empty
- Sample Content::
NVIDIA L4, 0x0000000000000001 NVIDIA A1, 0x0000000000000000 NVIDIA H1, 0x0000000000000084
- Examples::
>>> type(active_clocks_event_reasons) <class 'insights.parsers.nvidia.NvidiaSmiActiveClocksEventReasons'> >>> len(active_clocks_event_reasons) 3 >>> active_clocks_event_reasons[0]['applications_clocks_setting'] False >>> active_clocks_event_reasons[2]['hw_power_brake_slowdown'] True >>> active_clocks_event_reasons[2]['none'] False
- class insights.parsers.nvidia.NvidiaSmiL(context)[source]
Bases:
Parser,listWarning
This class is deprecated and will be removed from 3.8.0. Please use the
insights.parsers.nvidia.NvidiaSmiQueryGPUinstead.Prase for output of command /usr/bin/nvidia-smi -L. This command lists each of the NVIDIA GPUs in the system, along with their UUIDs.
The GPU info shown in each line will be parsed as follows:
model (string): The gpu model uuid (string): The gpu uuid
- Raises:
ParseException -- When run into unparsable gpu line
SkipComponent -- When content is empty or no parsable content
Sample Content:
GPU 0: NVIDIA A100-PCIE-40GB (UUID: GPU-63110aaa-3561-c8f5-e125-4ab40bbcf838) GPU 1: NVIDIA A100-PCIE-40GB (UUID: GPU-c9bd25dc-c0c4-3ab6-8f7f-3ad16d6bde4a)
- Examples::
>>> gpus.gpu_count 2 >>> "NVIDIA A100-PCIE-40GB" in gpus.gpu_models True >>> gpus[0] {'model': 'NVIDIA A100-PCIE-40GB', 'uuid': 'GPU-63110aaa-3561-c8f5-e125-4ab40bbcf838'}
- property gpu_count
Returns the GPU count
- Type:
str
- property gpu_models
Returns the GPUs model set
- Type:
str
- class insights.parsers.nvidia.NvidiaSmiQueryGPU(context)[source]
Bases:
Parser,listParser for the output of command /usr/bin/nvidia-smi --query-gpu=index,name,uuid,memory.total --format=csv,noheader. This command lists each of the NVIDIA GPUs info in the system.
The GPU info shown in each line will be parsed to a namedtuple
NvidiaGPUInfocontains the follows info:index (str): The gpu index model (str): The gpu model uuid (str): The gpu uuid memory_total(str): The gpu total memory
- Raises:
ParseException -- When run into an unparsable line
SkipComponent -- When content is empty
Sample Content:
0, NVIDIA L4, GPU-24598a07-f97d-fc79-86de-485c4c82d01c, 23034 MiB 1, NVIDIA A1, GPU-63110aaa-3561-c8f5-e125-4ab40bbcf838, 16280 MiB 2, NVIDIA H1, GPU-c9bd25dc-c0c4-3ab6-8f7f-3ad16d6bde4a, 24564 MiB
- Examples::
>>> gpus_info.gpu_count 3 >>> "NVIDIA L4" in gpus_info.gpu_models True >>> type(gpus_info[0]) <class 'insights.parsers.nvidia.NvidiaGPUInfo'> >>> gpus_info[0].uuid 'GPU-24598a07-f97d-fc79-86de-485c4c82d01c' >>> gpus_info[0].memory_total '23034 MiB'
- property gpu_count
Returns the GPU count
- Type:
str
- property gpu_models
Returns the GPUs model set
- Type:
str