def fresh_config():
return Config(
executors=[
HighThroughputExecutor(
label='theta_local_htex_multinode',
max_workers=1,
address=address_by_hostname(),
provider=CobaltProvider(
queue=user_opts['theta']['queue'],
account=user_opts['theta']['account'],
launcher=AprunLauncher(overrides="-d 64"),
walltime='00:10:00',
nodes_per_block=2,
init_blocks=1,
max_blocks=1,
# string to prepend to #COBALT blocks in the submit
# script to the scheduler eg: '#COBALT -t 50'
scheduler_options='',
# Command to be run before starting a worker, such as:
# 'module load Anaconda; source activate parsl_env'.
worker_init=user_opts['theta']['worker_init'],
cmd_timeout=120,
),
)
], )
def parsl_config(name: str) -> Tuple[Config, int]:
"""Make the compute resource configuration
Args:
name: Name of the diesred configuration
Returns:
- Parsl compute configuration
- Number of compute slots: Includes execution slots and pre-fetch buffers
"""
if name == 'local':
return Config(
executors=[
HighThroughputExecutor(max_workers=16, prefetch_capacity=1)
]
), 64
elif name == 'theta-debug':
return Config(
retries=16,
executors=[HighThroughputExecutor(
address=address_by_hostname(),
label="debug",
max_workers=64,
prefetch_capacity=64,
cpu_affinity='block',
provider=CobaltProvider(
account='redox_adsp',
queue='debug-flat-quad',
nodes_per_block=8,
scheduler_options='#COBALT --attrs enable_ssh=1',
walltime='00:60:00',
init_blocks=0,
max_blocks=1,
cmd_timeout=360,
launcher=AprunLauncher(overrides='-d 64 --cc depth -j 1'),
worker_init='''
module load miniconda-3
conda activate /lus/theta-fs0/projects/CSC249ADCD08/edw/env''',
),
)]
), 64 * 8 * 4
else:
raise ValueError(f'Configuration not defined: {name}')
max_workers=8, # One task per node
provider=CobaltProvider(
cmd_timeout=120,
nodes_per_block=8,
account='CSC249ADCD08',
queue='debug-cache-quad',
walltime="1:00:00",
init_blocks=1,
max_blocks=1,
launcher=SimpleLauncher(), # Places worker on the launch node
scheduler_options='#COBALT --attrs enable_ssh=1',
worker_init='''
module load miniconda-3
export PATH=~/software/psi4/bin:$PATH
conda activate /lus/theta-fs0/projects/CSC249ADCD08/colmena/env
# NWChem settings
export PATH="/home/lward/software/nwchem-6.8.1/bin/LINUX64:$PATH"
module load atp
export MPICH_GNI_MAX_EAGER_MSG_SIZE=16384
export MPICH_GNI_MAX_VSHORT_MSG_SIZE=10000
export MPICH_GNI_MAX_EAGER_MSG_SIZE=131072
export MPICH_GNI_NUM_BUFS=300
export MPICH_GNI_NDREG_MAXSIZE=16777216
export MPICH_GNI_MBOX_PLACEMENT=nic
export MPICH_GNI_LMT_PATH=disabled
export COMEX_MAX_NB_OUTSTANDING=6
export LD_LIBRARY_PATH=/opt/intel/compilers_and_libraries_2018.0.128/linux/compiler/lib/intel64_lin:$LD_LIBRARY_PATH
''',
),
),
executors=[
HighThroughputExecutor(
label='theta-htex',
max_workers=WORKERS_PER_NODE * MY_COMPUTE_NODES *
MY_COMPUTE_BLOCKS,
worker_debug=True,
address=address_by_hostname(),
provider=CobaltProvider(
queue=MY_QUEUE,
account=MY_ALLOCATION,
launcher=AprunLauncher(overrides="-d 64"),
walltime=MY_TIME,
nodes_per_block=MY_COMPUTE_NODES,
init_blocks=1,
min_blocks=1,
max_blocks=MY_COMPUTE_BLOCKS,
# string to prepend to #COBALT blocks in the submit
# script to the scheduler eg: '#COBALT -t 50'
scheduler_options='',
# Command to be run before starting a worker, such as:
worker_init='module load miniconda-3; export PATH=$PATH:{}'.
format(MY_USER_PATH),
cmd_timeout=120,
),
),
ThreadPoolExecutor(label='login-node', max_threads=8),
],
monitoring=MonitoringHub(
hub_address=address_by_hostname(),
hub_port=55055,
monitoring_debug=False,
HighThroughputExecutor(
address=address_by_hostname(),
label="htex",
max_workers=4,
prefetch_capacity=1,
provider=CobaltProvider(
queue='CVD_Research',
account='CVD_Research',
launcher=AprunLauncher(overrides="-d 256 --cc depth -j 4"),
walltime='3:00:00',
nodes_per_block=4,
init_blocks=1,
min_blocks=1,
max_blocks=4,
scheduler_options='#COBALT --attrs enable_ssh=1',
worker_init=f'''
module load miniconda-3
module load java
source activate /home/lward/exalearn/covid/toxicity-prediction/opera/env
export KMP_AFFINITY=disabled
which python
# Set the environment variables
{envs}
''',
cmd_timeout=120,
),
),
],
strategy=None,
)
def theta_persistent(log_dir: str,
nodes_per_nwchem: int = 1,
qc_nodes: int = 8,
ml_nodes: int = 8,
ml_prefetch: int = 0) -> Config:
"""Configuration where the application is persistent and sits on the Theta login node.
Nodes will be requested from Cobalt using separate jobs for ML and QC tasks.
Args:
nodes_per_nwchem: Number of nodes per NWChem computation
log_dir: Path to store monitoring DB and parsl logs
qc_nodes: Number of nodes dedicated to QC tasks
ml_prefetch: Number of tasks for ML workers to prefetch for inference
Returns:
(Config) Parsl configuration
"""
return Config(
retries=8,
executors=[
HighThroughputExecutor(
address=address_by_hostname(),
label="qc",
max_workers=qc_nodes // nodes_per_nwchem,
prefetch_capacity=ml_prefetch,
provider=CobaltProvider(
account='CSC249ADCD08',
queue='debug-cache-quad' if qc_nodes <= 8 else None,
walltime='00:60:00',
nodes_per_block=qc_nodes,
init_blocks=0,
max_blocks=1,
launcher=SimpleLauncher(),
cmd_timeout=360,
worker_init='''
module load miniconda-3
conda activate /lus/theta-fs0/projects/CSC249ADCD08/edw/env
export OMP_NUM_THREADS=64
export KMP_INIT_AT_FORK=FALSE
export PYTHONPATH=$PYTHONPATH:$(pwd)
export PATH="/lus/theta-fs0/projects/CSC249ADCD08/software/nwchem-6.8.1/bin/LINUX64:$PATH"
mkdir -p scratch # For the NWChem tasks
pwd
which nwchem
hostname
module load atp
export MPICH_GNI_MAX_EAGER_MSG_SIZE=16384
export MPICH_GNI_MAX_VSHORT_MSG_SIZE=10000
export MPICH_GNI_MAX_EAGER_MSG_SIZE=131072
export MPICH_GNI_NUM_BUFS=300
export MPICH_GNI_NDREG_MAXSIZE=16777216
export MPICH_GNI_MBOX_PLACEMENT=nic
export MPICH_GNI_LMT_PATH=disabled
export COMEX_MAX_NB_OUTSTANDING=6
export LD_LIBRARY_PATH=/opt/intel/mkl/lib/intel64_lin/:/opt/intel/compilers_and_libraries_2020.0.166/linux/compiler/lib/intel64_lin:$LD_LIBRARY_PATH
''',
),
),
HighThroughputExecutor(
address=address_by_hostname(),
label="ml",
max_workers=1,
prefetch_capacity=ml_prefetch,
provider=CobaltProvider(
account='CSC249ADCD08',
queue='debug-flat-quad',
nodes_per_block=ml_nodes,
scheduler_options='#COBALT --attrs enable_ssh=1',
walltime='00:60:00',
init_blocks=0,
max_blocks=1,
cmd_timeout=360,
launcher=AprunLauncher(
overrides='-d 256 --cc depth -j 4'
), # Places worker on the compute node
worker_init='''
module load miniconda-3
conda activate /lus/theta-fs0/projects/CSC249ADCD08/edw/env''',
),
)
],
run_dir=log_dir,
strategy='simple',
max_idletime=15.)
