Older runs

SewerML

F1normal (paper)	F2 CIW (paper)	F1normal (Leaderpaper)	F2 CIW (Leaderboard)	F1normal	F2 CIW	RunID	wandb Project	wandb name	Date of Submission	JOBID	Py file name	Input Dataset	Img-aug	Hyp-tuning	TL	Backbone	Epochs	freeze_epoch	BatchSize	set_seed	lr at start	cpus per task	GPU	GPU (GiB)	--mem=	--time=	actual running time	train runnining time	GPU0 (%)	GPU1 (%)	GPU2 (%)	GPU3 (%)	GPU0	GPU1	GPU2	GPU3	Remark
0.9094	0.5511	0.9268	0.6568	0.879649891	0.239015408	01	fastai_sewerml_v2	01_sample_00_v2_2022/10/07_20:23	10/7/2022	47139202	fastai_multi_label_v2.py	sample_00_v2	Resize((576, 720),method='squish', pad_mode='zeros')	None	TRUE	restnet50	5	4	32	42	1.20E-03	24	v100l:4	32	10G	0:30:00	0:16:52	0:13:56	86	0	0	0	27.52	0	0	0
0.9094	0.5511	0.9268	0.6568			02	fastai_sewerml_v2		10/7/2022	47163381	fastai_multi_label_v2.py	SewerML_Train_Val	Resize((576, 720),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	32	42		24	v100l:4	32	0	160:00:00
0.9094	0.5511	0.9268	0.6568			03	fastai_sewerml_v1		10/7/2022	47163397	fastai_multi_label_v1.py	SewerML_Train_v1	Resize((576, 720),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	32	42		24	v100l:4	32	0	160:00:00
0.9094	0.5511	0.9268	0.6568			04	fastai_sewer_cc		10/12/2022	47670560	fastai_distributed.py	sample_00	Resize(224, method='squish')	None	TRUE	restnet50	10	4	128	42		24	p100l:4	32	10G	00:30:00	0:10:04										srun python ./script_runner.py instead of accelerate launch
0.9094	0.5511	0.9268	0.6568			05	fastai_sewer_cc		10/12/2022	47672432	fastai_distributed.py	sample_00	Resize((576, 720),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	128	42		24	p100l:4	32	10G	00:30:00	fail										srun python ./script_runner.py instead of accelerate launch
0.9094	0.5511	0.9268	0.6568			06	fastai_sewer_cc		10/12/2022	47674139	fastai_distributed.py	sample_00	Resize((576, 720),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	128	42		24	p100l:4	32	10G	00:30:00	fail										use accelerate launch, add 'export NCCL_BLOCKING_WAIT=1' to the shell file
0.9094	0.5511	0.9268	0.6568			07	fastai_sewerml_v2		10/14/2022	47843676	fastai_multi_label_v2.py	SewerML_Train_Val	Resize((576, 720),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	32	42		24	v100l:4	32	110G	160:00:00
0.9094	0.5511	0.9268	0.6568	0.838574423	0.287488192	08	fastai_sewerml_v2		10/15/2022	47887637	fastai_multi_label_v2.py	sample_00_v2	Resize((288, 360),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	128	42		24	v100l:4	32	10G	0:20:00	0:10:21
0.9094	0.5511	0.9268	0.6568	0.81799591	0.391778293	09	fastai_sewerml_v2		10/15/2022	47888322	fastai_multi_label_v2.py	sample_00_v2	Resize((288, 360),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	144	42		24	v100l:4	32	10G	0:20:00	0:10:26
0.9094	0.5511	0.9268	0.6568	0.916343222	0.602978419	10	fastai_sewerml_v2		10/15/2022	47888978	fastai_multi_label_v2.py	SewerML_Train_Val	Resize((288, 360),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	144	42		24	v100l:4	32	110G	50:00:00

DNV

RunID	F1 normal	F1averag(w/o nd)	F2 avergae (w/o ND)	wandb Project	wandb name	Date of Submission	JOBID	Py file name	Input Dataset	Img-aug	Hyp-tuning	TL	Backbone	Epochs	freeze_epoch	BatchSize	set_seed	lr at start	cpus per task	GPU	GPU (GiB)	--mem=	--time=	actual running time	Remarks
29	99.33%	96.10%	94.58%	fastai_sewerml_v1		11/9/2022	50058424	fastai_multi_label_v1_470_700.py	azure_labels_Lynn_Valley_Oct_6_2021_1_12_14_20	Resize((470, 700),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	32	42		24	v100l:1	32	10G	1:00:00		80%-20% split
30	99.50%	94.67%	94.26%	fastai_sewerml_v1		11/9/2022	50066121	fastai_multi_label_v1_470_700.py	azure_labels_Lynn_Valley_Oct_6_2021 (33 videos)	Resize((470, 700),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	32	42		24	v100l:1	32	10G	3:00:00		80%-20% split
31	98.98%	93.57%	92.99%	fastai_sewerml_v1		11/14/2022	50504165	fastai_multi_label_v1_470_700.py	azure_labels_Lynn_Valley_March_3_2022 (39 videos)	Resize((470, 700),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	32	42		24	v100l:1	32	10G	3:00:00		80%-20% split
32	99.25%	97.36%	96.06%	fastai_sewerml_v1		11/14/2022	50513140	fastai_multi_label_v1_470_700.py	azure_33_39_videos_wo_B	Resize((470, 700),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	32	42		24	v100l:1	32	10G	5:00:00		80%-20% split
33	99.17%	94.88%	94.02%	fastai_sewerml_v2		11/16/2022	50711416	fastai_multi_label_v2_470_700.py	azure_33_39_videos	Resize((470, 700),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	32	42		24	v100l:1	32	10G	5:00:00		75%-15%-10% split 10% was not used in building the model
34	99.30%	98.03%	97.09%	fastai_sewerml_v2		11/16/2022	50711417	fastai_multi_label_v2_470_700.py	azure_33_39_videos_wo_B	Resize((470, 700),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	32	42		24	v100l:1	32	10G	5:00:00		75%-15%-10% split 10% was not used in building the model
35	99.52%	94.43%	93.92%	fastai_sewerml_v2		11/16/2022	50711426	fastai_multi_label_v2_470_700.py	azure_labels_Lynn_Valley_Oct_6_2021 (33 videos)	Resize((470, 700),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	32	42		24	v100l:1	32	10G	3:00:00		75%-15%-10% split 10% was not used in building the model
36	99.02%	93.50%	92.74%	fastai_sewerml_v2		11/16/2022	50711430	fastai_multi_label_v2_470_700.py	azure_labels_Lynn_Valley_March_3_2022 (39 videos)	Resize((470, 700),method='squish', pad_mode='zeros')	None	TRUE	restnet50	10	4	32	42		24	v100l:1	32	10G	3:00:00		75%-15%-10% split 10% was not used in building the model

Prediction run on test csv when doing 75-15-10 split

F1 normal	F1 average (no ND)	F2 average (no ND)	RunID	Date	JOBID	Test Data	cpus per task	--mem=	--time=	GPU	Actual Run time	Py file	saved_model
0.991249	0.950744	0.94364	33	44882	50746939	33_test_image.csv	24	10G	1:00:00	v100:1		fastai_multi_label_predict_DNV	33_73videos_75_15_10_split
0.992282	0.984584	0.976787	34	44882	50746965	34_test_image.csv	24	10G	1:00:00	v100:1		fastai_multi_label_predict_DNV	34_73videos_wo_B_75_15_10_split
0.99586	0.94002	0.932179	35	44882	50746851	35_test_image.csv	24	10G	1:00:00	v100:1		fastai_multi_label_predict_DNV	35_33videos_75_15_10_split
0.987027	0.924955	0.912248	36	44882	50747020	36_test_image.csv	24	10G	1:00:00	v100:1		fastai_multi_label_predict_DNV	36_39videos_75_15_10_split