#
# This flag is ignored if no preemptible instance types are
# configured, and has no effect on top-level containers.
- UsePreemptibleInstances: true
+ AlwaysUsePreemptibleInstances: true
# PEM encoded SSH key (RSA, DSA, or ECDSA) used by the
# cloud dispatcher for executing containers on worker VMs.
# %C number of VCPUs
# %M memory in MB
# %T tmp in MB
+ # %G number of GPU devices (runtime_constraints.cuda.device_count)
#
# Use %% to express a literal %. The %%J in the default will be changed
# to %J, which is interpreted by bsub itself.
# from /tmp, or adjust the "-o" and "-e" arguments accordingly.
BsubArgumentsList: ["-o", "/tmp/crunch-run.%%J.out", "-e", "/tmp/crunch-run.%%J.err", "-J", "%U", "-n", "%C", "-D", "%MMB", "-R", "rusage[mem=%MMB:tmp=%TMB] span[hosts=1]", "-R", "select[mem>=%MMB]", "-R", "select[tmp>=%TMB]", "-R", "select[ncpus>=%C]"]
+ # Arguments that will be appended to the bsub command line
+ # when submitting Arvados containers as LSF jobs with
+ # runtime_constraints.cuda.device_count > 0
+ BsubCUDAArguments: ["-gpu", "num=%G"]
+
# Use sudo to switch to this user account when submitting LSF
# jobs.
#
AddedScratch: 0
Price: 0.1
Preemptible: false
+ # Include this section if the node type includes GPU (CUDA) support
+ CUDA:
+ DriverVersion: "11.0"
+ HardwareCapability: "9.0"
+ DeviceCount: 1
StorageClasses: