def main():
# Construct argument parser
parser = argparse.ArgumentParser(description='DSPACE Model training.')
parser.add_argument('model_fn', help='model h5 file')
parser.add_argument('--shards', nargs='+')
args = parser.parse_args()
outputs = load_outputs("outputs.txt")
# Train model
evaluate_model(args.model_fn, args.shards, outputs)
return
def main():
# Construct argument parser
parser = argparse.ArgumentParser(description='DSPACE Model training.')
parser.add_argument(
'-w', '--datadir', type=str, default="data", help="Main data directory for all outputs"
)
parser.add_argument(
'-o', '--output', default='/output', help='output path to write output files to'
)
parser.add_argument('-e', '--epochs', type=int, default=10, help='Number of epochs')
args = parser.parse_args()
# Construct outputs
outputs = load_outputs('outputs.txt')
# Train model
train_model(args.epochs, args.datadir, args.output, outputs)
def main():
# Construct argument parser
parser = argparse.ArgumentParser(description='DSPACE Model training.')
parser.add_argument('-w',
'--datadir',
type=str,
default="data",
help="Main data directory for all outputs")
parser.add_argument('-o',
'--output',
default='/output',
help='output path to write output files to')
parser.add_argument('-e',
'--epochs',
type=int,
default=10,
help='Number of epochs')
parser.add_argument(
'--tigrfam_filter',
type=str,
help=
'Skip records containing at least one of the listed tifgrfams. (comma separated)'
)
args = parser.parse_args()
# Construct outputs
outputs = load_outputs('outputs.txt')
# Construct classflags
classflags = None
if args.tigrfam_filter:
classflags = {"tigrfams": args.tigrfam_filter.split(',')}
# Train model
train_model(args.epochs,
args.datadir,
args.output,
outputs,
classflags=classflags)
def main():
"""For testing purposes
"""
parser = argparse.ArgumentParser()
parser.add_argument('shard_files', nargs='+')
parser.add_argument('--batch_id', default=1, type=int)
parser.add_argument('--batch_size', default=8, type=int)
args = parser.parse_args()
outputs = load_outputs('outputs.txt')
batch = BatchProcessor(args.shard_files, args.batch_size, outputs)
# this will always be the same for a given shard_file input
for b in batch:
pass # print(batch[args.batch_id])
# these will change everytime
batch.on_epoch_end()
for b in batch:
pass # print(batch[args.batch_id])
return
def __init__(self, model, onehot_count, multihot_thresh, oprecision, eprecision, vector_form=False, skip_tasks=[], from_embed=False):
if type(model) is str:
self.model = load_model(
model, custom_objects={'precision': precision,
'recall': recall,
'fmeasure': fmeasure}
)
else:
self.model = model
self.eauto_index = None
self.model.outputs.append(self.model.get_layer('embedding').output)
self.embedding_index = len(self.model.outputs) - 1
self.model.outputs.append(self.model.get_layer('eauto3d').output)
self.eauto_index = len(self.model.outputs) - 1
self.onehot_count = onehot_count
self.multihot_thresh = multihot_thresh
self.oprecision = oprecision
self.eprecision = eprecision
self.vector_form = vector_form
self.skip_tasks = skip_tasks
self.from_embed = from_embed
self.outputs = load_outputs('outputs.txt')
if self.from_embed:
inputs, outputs = build_network(self.outputs, from_embed=True)
model2 = Model(inputs=inputs, outputs=outputs)
for layer in model2.layers:
try:
layer.set_weights(self.model.get_layer(layer.name).get_weights())
except:
pass
self.model = model2
self.eauto_index = len(self.model.outputs) - 1