def generate_search_space(num_nodes_per_cell, num_normal_cells,
num_reduction_cells, init_filters, stem_multiplier):
global global_vars, hp_sharer
global_vars = {}
hp_sharer = hp.HyperparameterSharer()
hp_sharer.register('drop_path_keep_prob',
lambda: D([.7], name='drop_path_keep_prob'))
stem_in, stem_out = stem(int(init_filters * stem_multiplier))
progress_in, progress_out = mo.identity()
global_vars['progress'] = progress_out['Out']
normal_cell_fn = create_cell_generator(num_nodes_per_cell, False)
reduction_cell_fn = create_cell_generator(num_nodes_per_cell, True)
total_cells = num_normal_cells + num_reduction_cells
hasReduction = [False] * num_normal_cells
for i in range(num_reduction_cells):
hasReduction[int(
float(i + 1) / (num_reduction_cells + 1) *
num_normal_cells)] = True
inputs = [stem_out, stem_out]
filters = init_filters
aux_loss_idx = int(
float(num_reduction_cells) /
(num_reduction_cells + 1) * num_normal_cells) - 1
outs = {}
cells_created = 0.0
for i in range(num_normal_cells):
if hasReduction[i]:
filters *= 2
connect_new_cell(
reduction_cell_fn(filters, (cells_created + 1) / total_cells),
inputs)
cells_created += 1.0
connect_new_cell(
normal_cell_fn(filters, (cells_created + 1) / total_cells), inputs)
cells_created += 1.0
if i == aux_loss_idx:
aux_in, aux_out = aux_logits()
aux_in['In'].connect(inputs[-1]['Out'])
outs['Out0'] = aux_out['Out']
_, final_out = mo.siso_sequential([(None, inputs[-1]),
relu(),
global_pool2d(),
dropout(D([1.0])),
fc_layer(D([10]))])
outs['Out1'] = final_out['Out']
return {'In0': stem_in['In'], 'In1': progress_in['In']}, outs
import tensorflow as tf
from collections import OrderedDict
import deep_architect.core as co
import deep_architect.hyperparameters as hp
import deep_architect.helpers.tfeager as htfe
import deep_architect.modules as mo
from search_spaces.tfe_ops import (relu, batch_normalization, conv2d,
separable_conv2d, avg_pool2d, max_pool2d,
min_pool2d, fc_layer, global_pool2d,
dropout, add, flatten)
from deep_architect.hyperparameters import Discrete as D
global_vars = {}
hp_sharer = hp.HyperparameterSharer()
def cell(input_fn, node_fn, combine_fn, unused_combine_fn, num_nodes,
hyperparameters):
def substitution_fn(**dh):
c_ins, c_outs = input_fn()
nodes = [c_outs['Out0'], c_outs['Out1']]
used_node = [False] * (num_nodes + 2)
for i in range(num_nodes):
# Get indices of hidden states to be combined
idx0 = dh[str(i) + '_0']
idx1 = dh[str(i) + '_1']
# Transform hidden states
h0 = node_fn(idx0, i, 0)