if args.only_cpu:
ONLY_CPU = True
if args.minimize_mode:
OPTIMIZE_MODE = "min"
if __name__ == "__main__":
parser = argparse.ArgumentParser("Start bi model tuning with hyperspace and resiliency testing, "
"specify output csv file name.")
parser.add_argument("-o", "--out", required=True)
parser.add_argument("-m", "--model")
parser.add_argument("-t", "--trials")
parser.add_argument("-j", "--json")
parser.add_argument('-d', "--max_diff", action="store_true")
parser.add_argument('-r', '--minimize_resiliency', action="store_true")
parser.add_argument('-l', '--on_lambda', action="store_true")
parser.add_argument('-n', '--start_space')
parser.add_argument('-c', '--only_cpu', action='store_true')
parser.add_argument('-p', '--project_name', default="hyper_sensitive")
parser.add_argument('--minimize_mode', action="store_true")
args = parser.parse_args()
bitune_parse_arguments(args)
# print(PT_MODEL)
print(OPTIMIZE_MODE)
if args.on_lambda:
spaceray.run_experiment(args, multi_train, ray_dir="~/raylogs", cpu=8, start_space=int(args.start_space),
mode=OPTIMIZE_MODE)
else:
spaceray.run_experiment(args, multi_train, ray_dir="/lus/theta-fs0/projects/CVD-Mol-AI/mzvyagin/raylogs", cpu=8,
start_space=int(args.start_space), mode=OPTIMIZE_MODE, project_name=args.project_name,
group_name='bi_tune')
from hyper_resilient_experiments.alexnet_cifar import pytorch_alexnet, tensorflow_alexnet
import spaceray
from ray import tune
class Namespace:
def __init__(self, **kwargs):
self.__dict__.update(kwargs)
def objective(config):
acc, model = tensorflow_alexnet.cifar100_tf_objective(config)
search_results = {'average_res': acc}
tune.report(**search_results)
return search_results
if __name__ == "__main__":
args = Namespace(json='../standard.json',
trials=10,
out='tensorflow_benchmark.csv')
spaceray.run_experiment(
args,
objective,
ray_dir="/lus/theta-fs0/projects/CVD-Mol-AI/mzvyagin/raylogs",
cpu=8,
start_space=0,
mode="max")
if __name__ == "__main__":
parser = argparse.ArgumentParser(
"Start bi model tuning with hyperspace and resiliency testing, "
"specify output csv file name.")
parser.add_argument("-o", "--out", required=True)
parser.add_argument("-m", "--model")
parser.add_argument("-t", "--trials")
parser.add_argument("-j", "--json")
parser.add_argument('-d', "--max_diff", action="store_true")
parser.add_argument('-r', '--minimize_resiliency', action="store_true")
parser.add_argument('-l', '--on_lambda', action="store_true")
parser.add_argument('-f', '--framework', required=True)
parser.add_argument('-n', '--start_space')
args = parser.parse_args()
bitune_parse_arguments(args)
# print(PT_MODEL
if args.on_lambda:
spaceray.run_experiment(args,
double_train,
ray_dir="~/raylogs",
cpu=8,
start_space=int(args.start_space))
else:
spaceray.run_experiment(
args,
double_train,
ray_dir="/lus/theta-fs0/projects/CVD-Mol-AI/mzvyagin/raylogs",
cpu=8,
start_space=int(args.start_space))
import numpy as np
import argparse
import torch
import spaceray
from ray import tune
def objective(config):
search_results = {}
average_res = np.random.rand()
search_results['average_res'] = average_res
tune.report(**search_results)
return search_results
if __name__ == "__main__":
parser = argparse.ArgumentParser(
"Start bi model tuning with hyperspace and resiliency testing, "
"specify output csv file name.")
parser.add_argument("-o", "--out", required=True)
parser.add_argument("-m", "--model")
parser.add_argument("-t", "--trials")
parser.add_argument("-j", "--json")
# print(NUM_CLASSES)
args = parser.parse_args()
spaceray.run_experiment(
args,
objective,
cpu=8,
ray_dir="/lus/theta-fs0/projects/CVD-Mol-AI/mzvyagin/raylogs")
print("NOTE: Training using Max Diff approach")
if args.minimize_resiliency:
MIN_RESILIENCY = True
print("NOTE: Training using Min Resiliency approach")
if __name__ == "__main__":
parser = argparse.ArgumentParser(
"Start bi model tuning with hyperspace and resiliency testing, "
"specify output csv file name.")
parser.add_argument("-o", "--out", required=True)
parser.add_argument("-t", "--trials")
parser.add_argument("-j", "--json")
parser.add_argument('-d', "--max_diff", action="store_true")
parser.add_argument('-r', '--minimize_resiliency', action="store_true")
parser.add_argument('-l', '--on_lambda', action="store_true")
args = parser.parse_args()
bitune_parse_arguments(args)
# print(PT_MODEL)
if args.on_lambda:
spaceray.run_experiment(args,
segmentation_multi_train,
ray_dir="~/raylogs",
cpu=8)
else:
spaceray.run_experiment(
args,
segmentation_multi_train,
ray_dir="/lus/theta-fs0/projects/CVD-Mol-AI/mzvyagin/raylogs",
cpu=8)