Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Commit 2e439c1

Browse files
bl0bl0
committed
refactor
1 parent 4a431f3 commit 2e439c1

25 files changed

+210
-190
lines changed

‎README.md‎

Lines changed: 9 additions & 11 deletions
Original file line numberDiff line numberDiff line change
@@ -31,18 +31,16 @@ If you need run on NUSWIDE_81 and COCO, we recommend you to follow https://githu
3131
## Models
3232
### TODO
3333

34-
- [ ] Pretrain model of Alexnet
35-
- [ ] pretrained G model
36-
- [ ] eval frequence & eval at last iter
37-
- [ ] refactor all
38-
- [ ] use config instead of constant
39-
- [ ] use no split
34+
- [ ] code refactor
4035
- [ ] evaluate mode
41-
- [ ] output dir which contains images, models, logs
42-
- [ ] mkdir automatically
43-
- [ ] training longger
44-
- [ ] resume training
45-
- [ ] rerun all process on a fresh machine
36+
- [ ] resume training
37+
- [ ] tensorboard
38+
- [ ] experiment
39+
- [ ] training longger
40+
- [ ] code release
41+
- [ ] pretrained G model
42+
- [ ] Pretrain model of Alexnet
43+
- [ ] rerun all process on a fresh machine
4644

4745
Configuration for th models is specified in a list of constants at the top of
4846
the file, you can use the following command to run it:

‎config.py‎

Lines changed: 10 additions & 16 deletions
Original file line numberDiff line numberDiff line change
@@ -1,19 +1,20 @@
1-
from yacs.config import CfgNode as CN
1+
from yacs.config import CfgNode
22
import os.path as osp
33

4-
_C = CN()
4+
_C = CfgNode()
55

6-
_C.MODEL = CN()
7-
_C.MODEL.PRETRAINED_MODEL_PATH = ""
6+
_C.MODEL = CfgNode()
87
_C.MODEL.ARCHITECTURE = "NORM" # GOOD, NORM
9-
_C.MODEL.DIM_G = 128 # Generator dimensionality
8+
_C.MODEL.DIM_G = 128 # generator dimensionality
109
_C.MODEL.DIM_D = 128 # Critic dimensionality
11-
_C.MODEL.DIM = 64 # DIM for good Generator and Discriminator
10+
_C.MODEL.DIM = 64 # DIM for good generator and discriminator
1211
_C.MODEL.HASH_DIM = 64
12+
_C.MODEL.PRETRAINED_MODEL_PATH = ""
1313

14-
_C.DATA = CN()
14+
_C.DATA = CfgNode()
1515
_C.DATA.USE_DATASET = "cifar10" # "cifar10", "nuswide81", "coco"
16-
_C.DATA.DATA_ROOT = "./data/cifar10" # "cifar10", "nuswide81", "coco"
16+
_C.DATA.LIST_ROOT = "./data/cifar10"
17+
_C.DATA.DATA_ROOT = "./data_list/cifar10"
1718
_C.DATA.LABEL_DIM = 10
1819
_C.DATA.DB_SIZE = 54000
1920
_C.DATA.TEST_SIZE = 1000
@@ -26,7 +27,7 @@
2627
_C.DATA.MODEL_DIR = osp.join(_C.DATA.OUTPUT_DIR, "models")
2728
_C.DATA.LOG_DIR = osp.join(_C.DATA.OUTPUT_DIR, "logs")
2829

29-
_C.TRAIN = CN()
30+
_C.TRAIN = CfgNode()
3031
_C.TRAIN.USE_PRETRAIN = False
3132
_C.TRAIN.BATCH_SIZE = 64
3233
_C.TRAIN.ITERS = 100000
@@ -44,10 +45,3 @@
4445
_C.TRAIN.FAKE_RATIO = 1.0
4546

4647
config = _C
47-
48-
49-
def update_and_inference(cfg_file):
50-
config.merge_from_file(args.cfg)
51-
52-
config.freeze()
53-
return c

‎config_yaml/cifar_step_1.yaml‎

Lines changed: 30 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,2 +1,31 @@
1+
MODEL:
2+
ARCHITECTURE: "NORM"
13
DATA:
2-
USE_DATASET: "cifar10"
4+
USE_DATASET: "cifar10" # "cifar10", "nuswide81", "coco"
5+
LABEL_DIM: 10
6+
DB_SIZE: 54000
7+
TEST_SIZE: 1000
8+
WIDTH_HEIGHT: 32
9+
OUTPUT_DIM: 3072 # Number of pixels (32*32*3)
10+
MAP_R: 54000
11+
LIST_ROOT: "./data_list/cifar10"
12+
DATA_ROOT: "./data/cifar10"
13+
OUTPUT_DIR: "./output/cifar10_step_1"
14+
IMAGE_DIR: "./output/cifar10_step_1/images"
15+
MODEL_DIR: "./output/cifar10_step_1/models"
16+
LOG_DIR: "./output/cifar10_step_1/logs"
17+
18+
TRAIN:
19+
USE_PRETRAIN: True
20+
BATCH_SIZE: 64
21+
ITERS: 100000
22+
CROSS_ENTROPY_ALPHA: 5
23+
LR: 1e-4 # Initial learning rate
24+
G_LR: 1e-4 # 1e-4
25+
DECAY: True # Whether to decay LR over learning
26+
N_CRITIC: 5 # Critic steps per generator steps
27+
SAVE_FREQUENCY: 20000 # How frequently to save model
28+
ACGAN_SCALE: 1.0
29+
ACGAN_SCALE_G: 0.1
30+
WGAN_SCALE: 1.0
31+
WGAN_SCALE_G: 1.0

‎config_yaml/cifar_step_2.yaml‎

Lines changed: 3 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -8,7 +8,9 @@ DATA:
88
TEST_SIZE: 1000
99
WIDTH_HEIGHT: 32
1010
OUTPUT_DIM: 3072 # Number of pixels (32*32*3)
11-
MAP_R: 54000
11+
MAP_R: 54000
12+
LIST_ROOT: "./data_list/cifar10"
13+
DATA_ROOT: "./data/cifar10"
1214
OUTPUT_DIR: "./output/cifar10_finetune"
1315
IMAGE_DIR: "./output/cifar10_finetune/images"
1416
MODEL_DIR: "./output/cifar10_finetune/models"

data/cifar10/database.txt renamed to data_list/cifar10/database.txt

File renamed without changes.

data/cifar10/database_nolabel.txt renamed to data_list/cifar10/database_nolabel.txt

File renamed without changes.

data/cifar10/test.txt renamed to data_list/cifar10/test.txt

File renamed without changes.

data/cifar10/train.txt renamed to data_list/cifar10/train.txt

File renamed without changes.

data/coco/database.txt renamed to data_list/coco/database.txt

File renamed without changes.

data/coco/database_nolabel.txt renamed to data_list/coco/database_nolabel.txt

File renamed without changes.

0 commit comments

Comments
 (0)