'kwargs': {
'lstm3d_spec': [
(w, 'forward'),
] * N,
}
}
} for N in range(2, 6)]
search_space += [{
'model': {
'kwargs': {
'lstm3d_spec': [
(w, 'backward'),
] * N,
}
}
} for N in range(1, 6)]
search_space += [{
'model': {
'kwargs': {
'lstm3d_spec': [
(w, 'bidirectional'),
] * N,
}
}
} for N in range(1, 6)]
speval(lambda x: create_and_train_model(**config, extra_kwargs=x),
search_space,
os.path.join(ROOT_OUTDIR, config['outdir'], "trials.db"),
timeout=10 * 60 * 60)
'kwargs': {
'l': 0.001
},
},
'schedule': {
'name': 'standard',
'kwargs': {
'monitor': 'val_loss',
'factor': 0.5,
'patience': 5,
'cooldown': 0
},
},
'seed': 0,
'steps_per_epoch': 250,
'test_size': 0.2,
'weights': None,
# Args:
'vars_mod_png2d': None,
'vars_mod_png3d': None,
'vars_mod_slice': None,
'outdir': 'dune/numu/01_rnne_v1/',
})
parse_concurrency_cmdargs(config)
logger = setup_logging(
log_file=os.path.join(ROOT_OUTDIR, config['outdir'], "train.log"))
create_and_train_model(**config)
