def test_select_per_layer_comp_ratios_with_spatial_svd_pruner(self):
pruner = SpatialSvdPruner()
eval_func = unittest.mock.MagicMock()
rounding_algo = unittest.mock.MagicMock()
eval_func.side_effect = [
10, 20, 30, 40, 50, 60, 70, 80, 90, 11, 21, 31, 35, 40, 45, 50, 55,
60
]
rounding_algo.round.side_effect = [
0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.1, 0.2, 0.3, 0.4,
0.5, 0.6, 0.7, 0.8, 0.9
]
model = mnist_torch_model.Net()
layer_db = LayerDatabase(model, input_shape=(1, 1, 28, 28))
selected_layers = [
layer for layer in layer_db if isinstance(layer.module, nn.Conv2d)
]
layer_db.mark_picked_layers(selected_layers)
# Instantiate child
greedy_algo = comp_ratio_select.GreedyCompRatioSelectAlgo(
layer_db,
pruner,
SpatialSvdCostCalculator(),
eval_func,
20,
CostMetric.mac,
Decimal(0.4),
10,
True,
None,
rounding_algo,
False,
bokeh_session=None)
layer_comp_ratio_list, stats = greedy_algo.select_per_layer_comp_ratios(
)
original_cost = SpatialSvdCostCalculator.compute_model_cost(layer_db)
for layer in layer_db:
if layer not in selected_layers:
layer_comp_ratio_list.append(LayerCompRatioPair(layer, None))
compressed_cost = SpatialSvdCostCalculator.calculate_compressed_cost(
layer_db, layer_comp_ratio_list, CostMetric.mac)
actual_compression_ratio = compressed_cost.mac / original_cost.mac
self.assertTrue(
math.isclose(Decimal(0.3), actual_compression_ratio, abs_tol=0.8))
print('\n')
for pair in layer_comp_ratio_list:
print(pair)
def test_per_layer_eval_scores(self):
url, process = start_bokeh_server_session(8006)
bokeh_session = BokehServerSession(url=url, session_id="compression")
pruner = unittest.mock.MagicMock()
eval_func = unittest.mock.MagicMock()
model = mnist_torch_model.Net().to('cpu')
layer_db = LayerDatabase(model, input_shape=(1, 1, 28, 28))
layer1 = layer_db.find_layer_by_name('conv1')
layer_db.mark_picked_layers([layer1])
eval_func.side_effect = [90, 80, 70, 60, 50, 40, 30, 20, 10]
# Instantiate child
greedy_algo = comp_ratio_select.GreedyCompRatioSelectAlgo(
layer_db,
pruner,
SpatialSvdCostCalculator(),
eval_func,
20,
CostMetric.mac,
0.5,
10,
True,
None,
None,
False,
bokeh_session=None)
progress_bar = ProgressBar(1,
"eval scores",
"green",
bokeh_session=bokeh_session)
data_table = DataTable(num_columns=3,
num_rows=1,
column_names=[
'0.1', '0.2', '0.3', '0.4', '0.5', '0.6',
'0.7', '0.8', '0.9'
],
row_index_names=[layer1.name],
bokeh_session=bokeh_session)
pruner.prune_model.return_value = layer_db
eval_dict = greedy_algo._compute_layerwise_eval_score_per_comp_ratio_candidate(
data_table, progress_bar, layer1)
self.assertEqual(90, eval_dict[Decimal('0.1')])
bokeh_session.server_session.close("test complete")
os.killpg(os.getpgid(process.pid), signal.SIGTERM)
def test_eval_scores_with_spatial_svd_pruner(self):
pruner = SpatialSvdPruner()
eval_func = unittest.mock.MagicMock()
eval_func.side_effect = [
90, 80, 70, 60, 50, 40, 30, 20, 10, 91, 81, 71, 61, 51, 41, 31, 21,
11
]
model = mnist_torch_model.Net()
# Create a layer database
layer_db = LayerDatabase(model, input_shape=(1, 1, 28, 28))
layer1 = layer_db.find_layer_by_name('conv1')
layer2 = layer_db.find_layer_by_name('conv2')
layer_db.mark_picked_layers([layer1, layer2])
# Instantiate child
greedy_algo = comp_ratio_select.GreedyCompRatioSelectAlgo(
layer_db,
pruner,
SpatialSvdCostCalculator(),
eval_func,
20,
CostMetric.mac,
0.5,
10,
True,
None,
None,
True,
bokeh_session=None)
dict = greedy_algo._compute_eval_scores_for_all_comp_ratio_candidates()
print()
print(dict)
self.assertEqual(90, dict['conv1'][Decimal('0.1')])
self.assertEqual(51, dict['conv2'][Decimal('0.5')])
self.assertEqual(21, dict['conv2'][Decimal('0.8')])
def test_comp_ratio_select_tar(self):
compute_model_cost = unittest.mock.MagicMock()
pruner = unittest.mock.MagicMock()
eval_func = unittest.mock.MagicMock()
eval_func.side_effect = [
0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.45, 0.5, 0.55, 0.6, 0.65,
0.7, 0.75, 0.8, 0.85, 0.9, 0.95, 0.97, 1.0, 0.1, 0.15, 0.2, 0.25,
0.3, 0.35, 0.4, 0.45, 0.5, 0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.85,
0.9, 0.95, 0.97, 1.0, 0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.45,
0.5, 0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.85, 0.9, 0.95, 0.97, 1.0,
0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.45, 0.5, 0.55, 0.6, 0.65,
0.7, 0.75, 0.8, 0.85, 0.9, 0.95, 0.97, 1.0
]
compute_model_cost.return_value = (500, 500)
compute_network_cost = unittest.mock.MagicMock()
compute_network_cost.return_value = (500, 500)
model = mnist_torch_model.Net().to('cpu')
layer_db = LayerDatabase(model, input_shape=(1, 1, 28, 28))
layer1 = layer_db.find_layer_by_name('conv2')
layer_db.mark_picked_layers([layer1])
layer2 = layer_db.find_layer_by_name('fc2')
layer_db.mark_picked_layers([layer2])
layer3 = layer_db.find_layer_by_name('fc1')
layer_db.mark_picked_layers([layer3])
# Instantiate child
tar_algo = comp_ratio_select.TarRankSelectAlgo(
layer_db=layer_db,
pruner=pruner,
cost_calculator=WeightSvdCostCalculator(),
eval_func=eval_func,
eval_iterations=20,
cost_metric=CostMetric.mac,
num_rank_indices=20,
use_cuda=False,
pymo_utils_lib=pymo_utils)
tar_algo._svd_lib_ref = create_autospec(pymo.Svd, instance=True)
tar_algo._svd_lib_ref.SetCandidateRanks = unittest.mock.MagicMock()
tar_algo._svd_lib_ref.SetCandidateRanks.return_value = 20
tar_algo._num_rank_indices = 20
with unittest.mock.patch(
'aimet_common.cost_calculator.CostCalculator.calculate_comp_ratio_given_rank'
) as calculate_comp_ratio_given_rank:
calculate_comp_ratio_given_rank.side_effect = [
0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 0.1, 0.2,
0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 0.1, 0.2, 0.3, 0.4,
0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6,
0.7, 0.8, 0.9, 1.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8,
0.9, 1.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0
]
layer_comp_ratio_list, stats = tar_algo.select_per_layer_comp_ratios(
)
self.assertEqual(layer_comp_ratio_list[2].eval_score, 0.97)
self.assertEqual(layer_comp_ratio_list[2].comp_ratio, 1.0)
def test_select_per_layer_comp_ratios(self):
pruner = unittest.mock.MagicMock()
eval_func = unittest.mock.MagicMock()
rounding_algo = unittest.mock.MagicMock()
rounding_algo.round.side_effect = [
0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.1, 0.2, 0.3, 0.4,
0.5, 0.6, 0.7, 0.8, 0.9
]
eval_func.side_effect = [
10, 20, 30, 40, 50, 60, 70, 80, 90, 11, 21, 31, 35, 40, 45, 50, 55,
60
]
model = mnist_torch_model.Net()
layer_db = LayerDatabase(model, input_shape=(1, 1, 28, 28))
layer1 = layer_db.find_layer_by_name('conv1')
layer2 = layer_db.find_layer_by_name('conv2')
selected_layers = [layer1, layer2]
layer_db.mark_picked_layers([layer1, layer2])
try:
os.remove('./data/greedy_selection_eval_scores_dict.pkl')
except OSError:
pass
# Instantiate child
greedy_algo = comp_ratio_select.GreedyCompRatioSelectAlgo(
layer_db,
pruner,
SpatialSvdCostCalculator(),
eval_func,
20,
CostMetric.mac,
Decimal(0.6),
10,
True,
None,
rounding_algo,
False,
bokeh_session=None)
layer_comp_ratio_list, stats = greedy_algo.select_per_layer_comp_ratios(
)
original_cost = SpatialSvdCostCalculator.compute_model_cost(layer_db)
for layer in layer_db:
if layer not in selected_layers:
layer_comp_ratio_list.append(LayerCompRatioPair(layer, None))
compressed_cost = SpatialSvdCostCalculator.calculate_compressed_cost(
layer_db, layer_comp_ratio_list, CostMetric.mac)
rounding_algo.round.side_effect = [
0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.1, 0.2, 0.3, 0.4,
0.5, 0.6, 0.7, 0.8, 0.9
]
actual_compression_ratio = compressed_cost.mac / original_cost.mac
self.assertTrue(
math.isclose(Decimal(0.6), actual_compression_ratio, abs_tol=0.05))
self.assertTrue(
os.path.isfile('./data/greedy_selection_eval_scores_dict.pkl'))
print('\n')
for pair in layer_comp_ratio_list:
print(pair)
# lets repeat with a saved eval_dict
greedy_algo = comp_ratio_select.GreedyCompRatioSelectAlgo(
layer_db,
pruner,
SpatialSvdCostCalculator(),
eval_func,
20,
CostMetric.mac,
Decimal(0.6),
10,
True,
'./data/greedy_selection_eval_scores_dict.pkl',
rounding_algo,
False,
bokeh_session=None)
layer_comp_ratio_list, stats = greedy_algo.select_per_layer_comp_ratios(
)
original_cost = SpatialSvdCostCalculator.compute_model_cost(layer_db)
for layer in layer_db:
if layer not in selected_layers:
layer_comp_ratio_list.append(LayerCompRatioPair(layer, None))
compressed_cost = SpatialSvdCostCalculator.calculate_compressed_cost(
layer_db, layer_comp_ratio_list, CostMetric.mac)
actual_compression_ratio = compressed_cost.mac / original_cost.mac
self.assertTrue(
math.isclose(Decimal(0.6), actual_compression_ratio, abs_tol=0.05))
print('\n')
for pair in layer_comp_ratio_list:
print(pair)