mirrors
/
fast-reid
mirror of https://github.com/JDAI-CV/fast-reid.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity
fast-reid/configs/Base-SBS.yml

64 lines
912 B
YAML
Raw Normal View History

update fastreid V1.0
2021-01-18 11:36:38 +08:00
_BASE_: Base-bagtricks.yml
style(configs): put all config files together put all config files into one place for easily control, and add tools for put train_net.py which almost the same in different projects
2020-04-29 16:18:54 +08:00
update all config files
2020-07-06 16:48:14 +08:00
MODEL:
update fastreid V1.0
2021-01-18 11:36:38 +08:00
FREEZE_LAYERS: [ backbone ]
style(configs): put all config files together put all config files into one place for easily control, and add tools for put train_net.py which almost the same in different projects
2020-04-29 16:18:54 +08:00
BACKBONE:
chore(configs): update Market1501 training config
2020-04-29 21:06:57 +08:00
WITH_NL: True
style(configs): put all config files together put all config files into one place for easily control, and add tools for put train_net.py which almost the same in different projects
2020-04-29 16:18:54 +08:00
HEADS:
update fastreid V1.0
2021-01-18 11:36:38 +08:00
NECK_FEAT: after
add configurable decorator & linear loss decouple (#441) Summary: Add configurable decorator which can call `Baseline` with `Baseline(cfg)` or `Baseline(cfg, heads=heads, ...)` Decouple linear and loss computation for partial-fc support. Reviewed By: l1aoxingyu
2021-03-23 12:10:06 +08:00
POOL_LAYER: GeneralizedMeanPoolingP
CLS_LAYER: CircleSoftmax
chore(configs): update all training config
2020-05-01 09:04:51 +08:00
SCALE: 64
MARGIN: 0.35
style(configs): put all config files together put all config files into one place for easily control, and add tools for put train_net.py which almost the same in different projects
2020-04-29 16:18:54 +08:00
LOSSES:
NAME: ("CrossEntropyLoss", "TripletLoss",)
CE:
EPSILON: 0.1
SCALE: 1.0
v0.3 update Summary: 1. change DPP training in apex way; 2. make warmup scheduler by iter and lr scheduler by epoch; 3. replace random erasing with torchvision implementation; 4. naming modification in config file
2020-12-07 14:19:20 +08:00
style(configs): put all config files together put all config files into one place for easily control, and add tools for put train_net.py which almost the same in different projects
2020-04-29 16:18:54 +08:00
TRI:
MARGIN: 0.0
HARD_MINING: True
NORM_FEAT: False
SCALE: 1.0
INPUT:
update fastreid V1.0
2021-01-18 11:36:38 +08:00
SIZE_TRAIN: [ 384, 128 ]
SIZE_TEST: [ 384, 128 ]
v0.3 update Summary: 1. change DPP training in apex way; 2. make warmup scheduler by iter and lr scheduler by epoch; 3. replace random erasing with torchvision implementation; 4. naming modification in config file
2020-12-07 14:19:20 +08:00
support classification in fastreid (#443) Summary: support classification and refactor build_dataloader which can support explicit parameters passing
2021-03-26 20:17:39 +08:00
AUTOAUG:
ENABLED: True
PROB: 0.1
style(configs): put all config files together put all config files into one place for easily control, and add tools for put train_net.py which almost the same in different projects
2020-04-29 16:18:54 +08:00
DATALOADER:
NUM_INSTANCE: 16
SOLVER:
support classification in fastreid (#443) Summary: support classification and refactor build_dataloader which can support explicit parameters passing
2021-03-26 20:17:39 +08:00
AMP:
ENABLED: True
update fastreid V1.0
2021-01-18 11:36:38 +08:00
OPT: Adam
v0.3 update Summary: 1. change DPP training in apex way; 2. make warmup scheduler by iter and lr scheduler by epoch; 3. replace random erasing with torchvision implementation; 4. naming modification in config file
2020-12-07 14:19:20 +08:00
MAX_EPOCH: 60
style(configs): put all config files together put all config files into one place for easily control, and add tools for put train_net.py which almost the same in different projects
2020-04-29 16:18:54 +08:00
BASE_LR: 0.00035
WEIGHT_DECAY: 0.0005
IMS_PER_BATCH: 64
update fastreid V1.0
2021-01-18 11:36:38 +08:00
SCHED: CosineAnnealingLR
v0.3 update Summary: 1. change DPP training in apex way; 2. make warmup scheduler by iter and lr scheduler by epoch; 3. replace random erasing with torchvision implementation; 4. naming modification in config file
2020-12-07 14:19:20 +08:00
DELAY_EPOCHS: 30
update fastreid V1.0
2021-01-18 11:36:38 +08:00
ETA_MIN_LR: 0.0000007
style(configs): put all config files together put all config files into one place for easily control, and add tools for put train_net.py which almost the same in different projects
2020-04-29 16:18:54 +08:00
v0.3 update Summary: 1. change DPP training in apex way; 2. make warmup scheduler by iter and lr scheduler by epoch; 3. replace random erasing with torchvision implementation; 4. naming modification in config file
2020-12-07 14:19:20 +08:00
WARMUP_FACTOR: 0.1
make lr warmup by iter Summary: change warmup way by iter not by epoch, which will make it more flexible when training small epochs
2021-01-22 11:17:21 +08:00
WARMUP_ITERS: 2000
style(configs): put all config files together put all config files into one place for easily control, and add tools for put train_net.py which almost the same in different projects
2020-04-29 16:18:54 +08:00
update fastreid V1.0
2021-01-18 11:36:38 +08:00
FREEZE_ITERS: 1000
v0.3 update Summary: 1. change DPP training in apex way; 2. make warmup scheduler by iter and lr scheduler by epoch; 3. replace random erasing with torchvision implementation; 4. naming modification in config file
2020-12-07 14:19:20 +08:00
CHECKPOINT_PERIOD: 20
style(configs): put all config files together put all config files into one place for easily control, and add tools for put train_net.py which almost the same in different projects
2020-04-29 16:18:54 +08:00
TEST:
update fastreid V1.0
2021-01-18 11:36:38 +08:00
EVAL_PERIOD: 10
update all config files
2020-07-06 16:48:14 +08:00
IMS_PER_BATCH: 128
style(configs): put all config files together put all config files into one place for easily control, and add tools for put train_net.py which almost the same in different projects
2020-04-29 16:18:54 +08:00
CUDNN_BENCHMARK: True