fig = rpt.make_plot(pd)
fig = rpt.make_plot(pd_h, h_figure=fig, axes_index=0)
return fig
# * New model, suspecting something is off.
models = ['2020-02-13-22.48.36', '2020-01-30-14.12.49']
# * Stacked LSTM 256 HUBER LOSS, L2 1024 LSTM, LSTM 512 HUBER LOSS
#models = ['2020-01-19-22.00.11', '2020-01-17-02.06.38', '2020-01-30-11.17.16']#, '2020-01-17-13.35.31']
# models = ['2020-01-08-13.54.40', ]
perf_classes = []
for model in models:
# * Locate the model directory
paths = hf.find_files(model)
for path in paths:
if path.split('/')[-1] == model:
break
perf_class_path = path + '/data/EnergyPerformance.pickle'
perf_class = pickle.load(open(perf_class_path, "rb"))
perf_classes.append(perf_class)
path = get_project_root() + '/plots/polar_L2_vs_sqr_angle.png'
title = 'Energy: Stacked 256 LSTM-size Huber (blue), 1028 LSTM L2 (orange)'
fig = energy_plot(models, perf_classes, title=title) #, savefig=path)
type=float,
help='Sets the minimum learning rate for scan inspection.')
parser.add_argument('--min_yrange',
default=np.inf,
type=float,
help='Sets the minimum learning rate for scan inspection.')
args = parser.parse_args()
if __name__ == '__main__':
if args.path != None:
model_dir = args.path
if 'model_dir' not in locals():
raise ValueError('No path supplied!')
# * Locate the model directory
paths = find_files(model_dir)
for path in paths:
if path.split('/')[-1] == model_dir:
model = path
break
from_lr, to_lr = args.min_lr, args.max_lr
lrs = pickle.load(open(model + '/lr.pickle', 'rb'))
losses = pickle.load(open(model + '/loss_vals.pickle', 'rb'))
indices = [
index for index in range(len(lrs)) if from_lr <= lrs[index] <= to_lr
]
chosen_lrs = np.array(lrs)[indices]
chosen_losses = np.array(losses)[indices]
if args.max_yrange != np.inf:
from src.modules.main_funcs import (calc_predictions_pickle,
load_model_pars, get_project_root)
description = 'Runs all experiments saved in "/experiments" starting with the oldest.'
parser = argparse.ArgumentParser(description=description)
parser.add_argument('-p', '--path', nargs='+', metavar='', type=str, help='Paths to model directories')
args = parser.parse_args()
if __name__ == '__main__':
for model in args.path:
# * Locate the model directory
paths = find_files(model)
for path in paths:
if path.split('/')[-1] == model:
break
hyper_pars, data_pars, arch_pars, meta_pars = load_model_pars(path)
WANDB_DIR = get_project_root()+'/models'
PROJECT = meta_pars['project']
WANDB_ID = path.split('/')[-1]
wandb.init(resume=True, id=WANDB_ID, dir=WANDB_DIR, project=PROJECT)
# * Load the model and check all predictions are there
perf_class_path = path +'/data/Performance.pickle'
perf_class = pickle.load( open( perf_class_path, "rb" ) )
perf_class.update_onenumber_performance()
perf = perf_class.onenumber_performance
print(get_time(), 'Onenumber performance: %.3f'%(perf))