Skip to main content

Browse the archive

Revision fff34406c03ef4d8017c269aaa86ae3448256e2b authored by Gerard Maggiolino on 23 July 2019, 06:23:32 UTC, committed by Gerard Maggiolino on 24 July 2019, 17:48:54 UTC 
Fixed PyTorch 1.1.0 use, Seeding, Refactoring
 
Changes to fix PyTorch 1.1.0 incompatibilities, add seeding functionality,
deleting unused lines.
- See issue #42
- Change of attributes and added context manager use in data.load_data and
  model.Ganomaly.set_input
- Removed unused statements and redundant attributes throughout model.Ganomaly
- Added previously missing seeding functionality to data loading, model
  initialization, and training procedure based on opt args.
1 parent 76ea5e4
Raw File
train.py 
"""
TRAIN GANOMALY

. Example: Run the following command from the terminal.
    run train.py                             \
        --model ganomaly                        \
        --dataset UCSD_Anomaly_Dataset/UCSDped1 \
        --batchsize 32                          \
        --isize 256                         \
        --nz 512                                \
        --ngf 64                               \
        --ndf 64
"""


##
# LIBRARIES
from __future__ import print_function

from options import Options
from lib.data import load_data
from lib.model import Ganomaly

##
# def main():
""" Training
"""

##
# ARGUMENTS
opt = Options().parse()

##
# LOAD DATA
dataloader = load_data(opt)

##
# LOAD MODEL
model = Ganomaly(opt, dataloader)

##
# TRAIN MODEL
model.train()

# if __name__ == '__main__':
#     main()
back to top