Merge branch 'xs/upd_tsn_mem_used' into 'master'

add mem used in TSN datasets other than k400 and fix some bug in configs

See merge request open-mmlab/mmaction-lite!327

configs/recognition/tsn/README.md

@@ -31,27 +31,27 @@
 
 |config | pretrain | top1 acc| top5 acc | gpu_mem(M) | iter time(s) | ckpt | log|
 |-|-|-|-|-|-|-|-|
-|[tsn_r50_1x1x8_50e_sthv1_rgb](/configs/recognition/tsn/tsn_r50_1x1x8_50e_sthv1_rgb.py) | ResNet50 |18.55|44.80| x | x | [ckpt]() | [log]()|
-|[tsn_r50_1x1x16_50e_sthv1_rgb](/configs/recognition/tsn/tsn_r50_1x1x16_50e_sthv1_rgb.py) | ResNet50 |15.77|39.85| x | x | [ckpt]() | [log]()|
+|[tsn_r50_1x1x8_50e_sthv1_rgb](/configs/recognition/tsn/tsn_r50_1x1x8_50e_sthv1_rgb.py) | ResNet50 |18.55|44.80| 10978 | x | [ckpt]() | [log]()|
+|[tsn_r50_1x1x16_50e_sthv1_rgb](/configs/recognition/tsn/tsn_r50_1x1x16_50e_sthv1_rgb.py) | ResNet50 |15.77|39.85| 5691 | x | [ckpt]() | [log]()|
 
 ### Something-Something V2
 
 |config | pretrain | top1 acc| top5 acc | gpu_mem(M) | iter time(s) | ckpt | log|
 |-|-|-|-|-|-|-|-|
-|[tsn_r50_1x1x8_50e_sthv2_rgb](/configs/recognition/tsn/tsn_r50_1x1x8_50e_sthv2_rgb.py) | ResNet50 |32.41|64.05| x | x | [ckpt]() | [log]()|
-|[tsn_r50_1x1x16_50e_sthv2_rgb](/configs/recognition/tsn/tsn_r50_1x1x16_50e_sthv2_rgb.py) | ResNet50 |22.48|49.08| x | x | [ckpt]() | [log]()|
+|[tsn_r50_1x1x8_50e_sthv2_rgb](/configs/recognition/tsn/tsn_r50_1x1x8_50e_sthv2_rgb.py) | ResNet50 |32.41|64.05| 10978 | x | [ckpt]() | [log]()|
+|[tsn_r50_1x1x16_50e_sthv2_rgb](/configs/recognition/tsn/tsn_r50_1x1x16_50e_sthv2_rgb.py) | ResNet50 |22.48|49.08| 5698 | x | [ckpt]() | [log]()|
 
 ### Moments in Time
 
 |config | pretrain | top1 acc| top5 acc | gpu_mem(M) | iter time(s) | ckpt | log|
 |-|-|-|-|-|-|-|-|
-|[tsn_r50_1x1x6_100e_mit_rgb](/configs/recognition/tsn/tsn_r50_1x1x6_100e_mit_rgb.py) | ResNet50 |26.84|51.6| x | x | [ckpt]() | [log]()|
+|[tsn_r50_1x1x6_100e_mit_rgb](/configs/recognition/tsn/tsn_r50_1x1x6_100e_mit_rgb.py) | ResNet50 |26.84|51.6| 8339 | x | [ckpt]() | [log]()|
 
 ### Multi-Moments in Time
 
 |config | pretrain | mAP| gpu_mem(M) | iter time(s) | ckpt | log|
 |-|-|-|-|-|-|-|
-|[tsn_r101_1x1x5_50e_mmit_rgb](/configs/recognition/tsn/tsn_r101_1x1x5_50e_mmit_rgb.py) | ResNet101 |61.09| x | x | [ckpt]() | [log]()|
+|[tsn_r101_1x1x5_50e_mmit_rgb](/configs/recognition/tsn/tsn_r101_1x1x5_50e_mmit_rgb.py) | ResNet101 |61.09| 10467 | x | [ckpt]() | [log]()|
 
 For more details on data preparation, you can refer to [preparing_ucf101](/tools/data/ucf101/preparing_ucf101.md),
 [preparing_kinetics400](/tools/data/kinetics400/preparing_kinetics400.md), [preparing_sthv1](/tools/data/sthv1/preparing_sthv1.md),

configs/recognition/tsn/tsn_r50_1x1x6_100e_mit_rgb.py

@@ -80,7 +80,7 @@ test_pipeline = [
     dict(type='ToTensor', keys=['imgs'])
 ]
 data = dict(
-    videos_per_gpu=32,
+    videos_per_gpu=16,
     workers_per_gpu=4,
     train=dict(
         type=dataset_type,
@@ -98,7 +98,7 @@ data = dict(
         data_prefix=data_root_val,
         pipeline=test_pipeline))
 # optimizer
-optimizer = dict(type='SGD', lr=0.01, momentum=0.9, weight_decay=0.0001)
+optimizer = dict(type='SGD', lr=0.005, momentum=0.9, weight_decay=0.0001)
 optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
 # learning policy
 lr_config = dict(policy='step', step=[40, 80])

configs/recognition/tsn/tsn_r50_1x1x8_50e_sthv1_rgb.py

@@ -81,7 +81,7 @@ test_pipeline = [
     dict(type='ToTensor', keys=['imgs'])
 ]
 data = dict(
-    videos_per_gpu=8,
+    videos_per_gpu=16,
     workers_per_gpu=4,
     train=dict(
         type=dataset_type,
@@ -99,7 +99,7 @@ data = dict(
         data_prefix=data_root_val,
         pipeline=test_pipeline))
 # optimizer
-optimizer = dict(type='SGD', lr=0.01, momentum=0.9, weight_decay=0.0005)
+optimizer = dict(type='SGD', lr=0.02, momentum=0.9, weight_decay=0.0005)
 optimizer_config = dict(grad_clip=dict(max_norm=20, norm_type=2))
 # learning policy
 lr_config = dict(policy='step', step=[20, 40])