forked from BR-IDL/PaddleViT
-
Notifications
You must be signed in to change notification settings - Fork 0
/
SETR_Naive_Large_768x768_80k_cityscapes_bs_8.yaml
59 lines (56 loc) · 1.58 KB
/
SETR_Naive_Large_768x768_80k_cityscapes_bs_8.yaml
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
DATA:
BATCH_SIZE: 2 # per GPU [total bs is set to 8 or 16]
BATCH_SIZE_VAL: 1 # per GPU
DATASET: 'Cityscapes' # dataset name
DATA_PATH: '/home/ssd3/wutianyi/datasets/cityscapes'
CROP_SIZE: (768,768) # input_size (training)
NUM_CLASSES: 19
MODEL:
NAME: 'SETR_Naive'
ENCODER:
TYPE: 'ViT'
OUT_INDICES: [9, 14, 19, 23]
PRETRAINED: './pretrain_models/backbones/vit_large_patch16_224.pdparams'
DECODER_TYPE: 'Naive_VisionTransformerUpHead'
PUP:
INPUT_CHANNEL: 1024
NUM_CONV: 2
NUM_UPSAMPLE_LAYER: 1
CONV3x3_CONV1x1: False
ALIGN_CORNERS: False
TRANS:
PATCH_SIZE: 16
HIDDEN_SIZE: 1024 # 768(Base), 1024(Large), 1280(Huge)
MLP_RATIO: 4 # same as mlp_ratio = 4.0
NUM_HEADS: 16 # 12(Base), 16(Large), 16(Huge)
NUM_LAYERS: 24 # 12(Base), 24(Large), 32(Huge)
QKV_BIAS: True
AUXPUP:
INPUT_CHANNEL: 1024
NUM_CONV: 2
NUM_UPSAMPLE_LAYER: 1
CONV3x3_CONV1x1: False
ALIGN_CORNERS: False
AUX:
AUXIHEAD: True
TRAIN:
BASE_LR: 0.01
END_LR: 1e-4
DECODER_LR_COEF: 10.0
ITERS: 80000
POWER: 0.9
DECAY_STEPS: 80000
LR_SCHEDULER:
NAME: 'PolynomialDecay'
OPTIMIZER:
GRAD_CLIP: 1.0
WEIGHT_DECAY: 0.0
NAME: 'SGD'
MOMENTUM: 0.9
VAL:
MULTI_SCALES_VAL: False
SCALE_RATIOS: [0.5, 0.75, 1.0, 1.25, 1.5, 1.75]
IMAGE_BASE_SIZE: 1024
CROP_SIZE: [768,768]
STRIDE_SIZE: [512,512]
SAVE_DIR: "./output/SETR_Naive_Large_768x768_80k_cityscapes_bs_8"