From 64dfde52d20ce40fe5edf8b30cb1170591cbb048 Mon Sep 17 00:00:00 2001 From: Simon Guilbault Date: Wed, 17 Apr 2024 15:22:14 -0400 Subject: [PATCH] L4 does not support FP64 (#25) --- slurm-job-exporter.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/slurm-job-exporter.py b/slurm-job-exporter.py index 6189447..0577e71 100644 --- a/slurm-job-exporter.py +++ b/slurm-job-exporter.py @@ -161,7 +161,7 @@ def __init__(self, dcgm_update_interval=10): device = pydcgm.dcgm_agent.dcgmGetDeviceAttributes(self.handle.handle, gpu_id) name = device.identifiers.deviceName print('Detected gpu {} with ID {}'.format(name, gpu_id)) - if name in ['NVIDIA RTX A6000']: + if name in ['NVIDIA RTX A6000', 'NVIDIA L4']: # This GPU does not supports fp64, we don't support a mix of fp64 and non-fp64 GPUs in the same node print('Removing fp64 metrics since {} does not support fp64'.format(name)) del self.fieldIds_dict[dcgm_fields.DCGM_FI_PROF_PIPE_FP64_ACTIVE]