obiwan.dplearn.cnn

Trains a CNN on fake and real galaxy images using TensorFlow.

Adapted from https://github.com/ageron/handson-ml

Functions

BatchGen(brick, indir[, batch_size])
bookmark_fn(outdir) Single line text file storing the epoch,brick,batch number of last ckpt
get_bookmark(outdir)
get_bricks([fn])
get_checkpoint(epoch, brick, outdir)
get_indir([nersc]) Returns path to dr5_testtrain directory
get_logdir(outdir)
get_outdir([nersc, knl]) Where to write ckpt and log files
get_xtrain_fns(brick, indir)
obiwan.dplearn.cnn.bookmark_fn(outdir)[source]

Single line text file storing the epoch,brick,batch number of last ckpt

obiwan.dplearn.cnn.get_indir(nersc=False)[source]

Returns path to dr5_testtrain directory

obiwan.dplearn.cnn.get_outdir(nersc=False, knl=False)[source]

Where to write ckpt and log files