def loadmodels(args):
    models = util.loadmodels(args, "cfg/classification")
    if args.mode == "top":
        for name in models.keys():
            models[name]['trainer'].max_steps = 10000  # training on top 500 never seems to need more than 10000 steps, since it's a small training set
    for cfg in models.itervalues():
        cfg["model"].conv_seq[0].fsize = 24     # Override default max motif length
    return models
Esempio n. 2
0
def loadmodels(args):
    models = util.loadmodels(args, "cfg/classification")
    if args.mode == "top":
        for name in models.keys():
            models[name][
                'trainer'].max_steps = 10000  # training on top 500 never seems to need more than 10000 steps, since it's a small training set
    for cfg in models.itervalues():
        cfg["model"].conv_seq[
            0].fsize = 24  # Override default max motif length
    return models
Esempio n. 3
0
def loadmodels(args):
    models = util.loadmodels(args, "cfg/regression/maxpool")
    for cfg in models.itervalues():
        cfg["model"].conv_seq[0].fsize = 24  # Override default max motif length
    return models
Esempio n. 4
0
def loadmodels(args):
    models = util.loadmodels(args, "cfg/classification")
    return models
Esempio n. 5
0
def loadmodels(args):
    models = util.loadmodels(args, "cfg/regression/maxpool")
    for cfg in models.itervalues():
        cfg["model"].conv_seq[
            0].fsize = 24  # Override default max motif length
    return models
Esempio n. 6
0
def loadmodels(args):
    models = util.loadmodels(args, "cfg/classification")
    return models
Esempio n. 7
0
def loadmodels(args):
    models = util.loadmodels(args, modeldir="cfg/regression/allpool")
    return models