forked from LINC-BIT/legodnn
-
Notifications
You must be signed in to change notification settings - Fork 0
/
legodnn_fcn_r18_test.py
89 lines (80 loc) · 4.97 KB
/
legodnn_fcn_r18_test.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
import sys
sys.setrecursionlimit(100000)
import torch
import copy
from legodnn import BlockRetrainer, BlockProfiler, LatencyEstimator, ScalingOptimizer
from legodnn.common.utils.dl.common.model import get_model_size, set_module
from legodnn.common.utils.dl.common.env import set_random_seed
set_random_seed(0)
from legodnn.common.detection.model_topology_extraction import topology_extraction
from legodnn.common.manager.block_manager.auto_block_manager import AutoBlockManager
from legodnn.common.detection.common_detection_manager_1204_new import CommonDetectionManager
from legodnn.common.utils.gen_series_legodnn_models import gen_series_legodnn_models
from legodnn.common.manager.model_manager.common_semantic_segmentation_model_manager_v2 import CommonSemanticSegmentationModelManager
from cv_task.semantic_segmentation.mmseg_models.legodnn_configs import get_fcn_r18_d8_512x512_b16_30k_voc2012_config
from cv_task.semantic_segmentation.mmseg_tools import mmseg_init_model
from cv_task.datasets.semantic_segmentation import mmseg_build_dataloader
from mmcv.parallel import MMDataParallel
if __name__=='__main__':
cv_task = 'semantic_segmentation'
dataset_name = 'cityscapes'
model_name = 'mmseg_fcn_r18_d8'
# compress_layer_max_ratio = 0.25
compress_layer_max_ratio = 0.125
device = 'cuda'
model_input_size = (1, 3, 800, 800)
train_batch_size = 64
test_batch_size = 64
block_sparsity = [0.0, 0.3, 0.6, 0.8]
root_path = os.path.join('results/legodnn', cv_task, model_name+'_'+dataset_name + '_' + str(compress_layer_max_ratio).replace('.', '-'))
compressed_blocks_dir_path = root_path + '/compressed'
trained_blocks_dir_path = root_path + '/trained'
descendant_models_dir_path = root_path + '/descendant'
block_training_max_epoch = 20
test_sample_num = 100
model_config = get_fcn_r18_d8_512x512_b16_30k_voc2012_config()
teacher_pt_file = None
checkpoint = None
print('\033[1;36m--------------------------------> BUILD LEGODNN GRAPH\033[0m')
jit_detector = mmseg_init_model(model_config, None, mode='lego_jit', device=device)
model_graph = topology_extraction(jit_detector, model_input_size, device=device)
model_graph.print_ordered_node()
print('\033[1;36m--------------------------------> START BLOCK DETECTION\033[0m')
detection_manager = CommonDetectionManager(model_graph, max_ratio=compress_layer_max_ratio)
detection_manager.detection_all_blocks()
detection_manager.print_all_blocks()
model_manager = CommonSemanticSegmentationModelManager()
block_manager = AutoBlockManager(block_sparsity, detection_manager, model_manager)
if teacher_pt_file is not None:
teacher_detector = mmseg_init_model(config=model_config, checkpoint=None, mode='mmdet_test', device=device)
raw_teacher = torch.load(teacher_pt_file).to(device)
for name, module in raw_teacher.named_modules():
if len(list(module.children()))>0:
continue
else:
set_module(teacher_detector, name, copy.deepcopy(module))
else:
teacher_detector = mmseg_init_model(config=model_config, checkpoint=checkpoint, mode='mmdet_test', device=device)
print('\033[1;36m--------------------------------> START BLOCK EXTRACTION\033[0m')
block_manager.extract_all_blocks(teacher_detector, compressed_blocks_dir_path, model_input_size, device)
print('\033[1;36m--------------------------------> START BLOCK TRAIN\033[0m')
train_loader, test_loader = mmseg_build_dataloader(cfg=model_config)
parallel_teacher_detector = MMDataParallel(teacher_detector.cuda(0), device_ids=[0])
block_trainer = BlockRetrainer(parallel_teacher_detector, block_manager, model_manager, compressed_blocks_dir_path,
trained_blocks_dir_path, block_training_max_epoch, train_loader, device=device)
block_trainer.train_all_blocks()
# exit(0)
server_block_profiler = BlockProfiler(teacher_detector, block_manager, model_manager,
trained_blocks_dir_path, test_loader, model_input_size, device)
server_block_profiler.profile_all_blocks()
edge_block_profiler = LatencyEstimator(block_manager, model_manager, trained_blocks_dir_path,
test_sample_num, model_input_size, device)
edge_block_profiler.profile_all_blocks()
# exit(0)
optimal_runtime = ScalingOptimizer(trained_blocks_dir_path, model_input_size,
block_manager, model_manager, device)
model_size_min = get_model_size(torch.load(os.path.join(compressed_blocks_dir_path, 'model_frame.pt')))/1024**2
model_size_max = get_model_size(teacher_detector)/1024**2 + 1
gen_series_legodnn_models(deadline=100, model_size_search_range=[model_size_min, model_size_max], target_model_num=50, optimal_runtime=optimal_runtime, descendant_models_save_path=descendant_models_dir_path, device=device)