type=int,
help="Offset where to start processing")
parser.add_argument('--results_dir',
default='results',
help="Name of results directory")
opts = parser.parse_args()
assert opts.o is None or len(
opts.datasets
) == 1, "Cannot specify result filename with more than one dataset"
for dataset_path in opts.datasets:
assert os.path.isfile(
check_extension(dataset_path)), "File does not exist!"
dataset_basename, ext = os.path.splitext(
os.path.split(dataset_path)[-1])
if opts.o is None:
results_dir = os.path.join(opts.results_dir, "pctsp",
dataset_basename)
os.makedirs(results_dir, exist_ok=True)
out_file = os.path.join(
results_dir, "{}{}{}-{}{}".format(
dataset_basename, "offs{}".format(opts.offset)
if opts.offset is not None else "",
"n{}".format(opts.n) if opts.n is not None else "",
opts.method, ext))
else [opts.data_distribution]
}
for problem, distributions in problems.items():
for distribution in distributions or [None]:
for graph_size in opts.graph_sizes:
datadir = os.path.join(opts.data_dir, problem)
os.makedirs(datadir, exist_ok=True)
if opts.filename is None:
filename = os.path.join(datadir, "{}{}{}_{}_seed{}.pkl".format(
problem,
"_{}".format(distribution) if distribution is not None else "",
graph_size, opts.name, opts.seed))
else:
filename = check_extension(opts.filename)
assert opts.f or not os.path.isfile(check_extension(filename)), \
"File already exists! Try running with -f option to overwrite."
np.random.seed(opts.seed)
if problem == 'vrp':
dataset = generate_vrp_data(opts.dataset_size, graph_size)
elif problem == 'vrptw':
dataset = generate_vrptw_data(opts.dataset_size, graph_size)
else:
assert False, "Unknown problem: {}".format(problem)
save_dataset(dataset, filename)
type=int,
default=10000,
help="Size of the dataset")
parser.add_argument(
'--graph_size',
type=int,
nargs='+',
default=20,
help="Sizes of problem instances (default 20, 50, 100)")
parser.add_argument("-f",
action='store_true',
help="Set true to overwrite")
parser.add_argument('--seed', type=int, default=4321, help="Random seed")
opts = parser.parse_args()
datadir = os.path.join(opts.data_dir, opts.problem)
os.makedirs(datadir, exist_ok=True)
if opts.filename is None:
filename = os.path.join(
datadir, "{}{}_{}_seed{}.pkl".format(opts.problem, opts.graph_size,
opts.name, opts.seed))
else:
filename = check_extension(opts.filename)
np.random.seed(opts.seed)
dataset = generate_tsp_data(opts.dataset_size, opts.graph_size)
print(dataset[0])
save_dataset(dataset, filename)
parser.add_argument("-f", action='store_true', help="Set true to overwrite")
parser.add_argument("-o", default=None, help="Name of the results file to write")
parser.add_argument("--cpus", type=int, help="Number of CPUs to use, defaults to all cores")
parser.add_argument('--disable_cache', action='store_true', help='Disable caching')
parser.add_argument('--progress_bar_mininterval', type=float, default=0.1, help='Minimum interval')
parser.add_argument('-n', type=int, help="Number of instances to process")
parser.add_argument('--offset', type=int, help="Offset where to start processing")
parser.add_argument('--results_dir', default='results', help="Name of results directory")
opts = parser.parse_args()
assert opts.o is None or len(opts.datasets) == 1, "Cannot specify result filename with more than one dataset"
for dataset_path in opts.datasets:
assert os.path.isfile(check_extension(dataset_path)), "File does not exist!"
dataset_basename, ext = os.path.splitext(os.path.split(dataset_path)[-1])
if opts.o is None:
results_dir = os.path.join(opts.results_dir, "pctsp", dataset_basename)
os.makedirs(results_dir, exist_ok=True)
out_file = os.path.join(results_dir, "{}{}{}-{}{}".format(
dataset_basename,
"offs{}".format(opts.offset) if opts.offset is not None else "",
"n{}".format(opts.n) if opts.n is not None else "",
opts.method, ext
))
else:
out_file = opts.o
def run(opts):
# disable sync
os.environ['WANDB_MODE'] = 'dryrun'
# initialize wandb
wandb.init(project='Network1')
# load all arguments to config to save as hyperparameters
wandb.config.update(opts)
# Pretty print the run args
pp.pprint(vars(opts))
# Set the random seed
torch.manual_seed(opts.seed)
# Optionally configure tensorboard
tb_logger = None
if not opts.no_tensorboard:
tb_logger = TbLogger(
os.path.join(opts.log_dir, "{}_{}".format(opts.problem,
opts.graph_size),
opts.run_name))
os.makedirs(opts.save_dir)
# Save arguments so exact configuration can always be found
with open(os.path.join(opts.save_dir, "args.json"), 'w') as f:
json.dump(vars(opts), f, indent=True)
# Set the device
opts.device = torch.device("cuda" if opts.use_cuda else "cpu")
# load created dataset from path
train_dataset = load_dataset(filename=opts.train_dataset)
# For now, val and train dataset are same
val_dataset = load_dataset(filename=opts.val_dataset)
# initialize model(need to be modified for regression case)
model = Net1(n_features=train_dataset[0].num_features,
embed_dim=opts.embed_dim,
out_features=opts.num_output).to(opts.device)
# code for multiple gpu model(disabled for now)
# enable once the model runs successfully for single GPU
# if opts.use_cuda and torch.cuda.device_count() > 1:
# print("No. of GPUs:", torch.cuda.device_count())
# model = torch.nn.DataParallel(model)
# initialize optimizer
optimizer = torch.optim.Adam(model.parameters(),
lr=0.01,
weight_decay=5e-4)
# save pytorch model and track all of the gradients and optionally parameters
wandb.watch(model, log='all') # "gradients", "parameters", "all", or None.
# start training
training_status = train(model, optimizer, train_dataset, val_dataset,
tb_logger, opts)
print(training_status)
# Get the embeddings and save to create data for Network2
scores, embeddings = compute_embeddings(model, opts, data=train_dataset)
# save embeddings to create dataset for Network2
filename = os.path.join(
opts.data_dir, "op{}{}_{}N{}_seed{}_label_{}_embeddings.pkl".format(
"_{}".format(opts.data_distribution)
if opts.data_distribution is not None else "", opts.graph_size,
opts.name, opts.dataset_size, opts.dataset_seed, opts.label_type))
assert opts.f or not os.path.isfile(check_extension(filename)), \
"File already exists! Try running with -f option to overwrite."
print(filename)
save_dataset([embeddings, scores], filename)
print("Embeddings Computed, shape:{}".format(embeddings.shape))
print("Scores Computed, shape:{}".format(scores.shape))
