-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathjob_prov.sh
116 lines (61 loc) · 7.25 KB
/
job_prov.sh
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
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
parallel_processes=0 # parallelizing provenance
check_prov_cache=True
# ## =========================================== Job 1: Provenance Main Paper Results ===========================================
# key_start='For-Main-Training-'
# num_clients=100
# clients_per_round=10
# dirichlet_alpha="0.1,0.2,0.3"
# text_models="openai-community/openai-gpt,google-bert/bert-base-cased"
# text_dataset="dbpedia_14,yahoo_answers_topics"
# num_rounds=25
# python -m tracefl.main --multirun check_prov_cache=$check_prov_cache parallel_processes=$parallel_processes do_provenance=True exp_key=$key_start model.name=$text_models dataset.name=$text_dataset num_clients=$num_clients num_rounds=$num_rounds clients_per_round=$clients_per_round dirichlet_alpha=$dirichlet_alpha dry_run=False hydra/launcher=joblib | tee -a logs/prov_text_classification_exp.log
# img_models="resnet18,densenet121"
# img_datasets="mnist,cifar10"
# num_rounds=50
# python -m tracefl.main --multirun check_prov_cache=$check_prov_cache parallel_processes=$parallel_processes do_provenance=True exp_key=$key_start model.name=$img_models dataset.name=$img_datasets num_clients=$num_clients num_rounds=$num_rounds clients_per_round=$clients_per_round dirichlet_alpha=$dirichlet_alpha dry_run=False hydra/launcher=joblib | tee -a logs/prov_image_classification_exp.log
# img_models="resnet18,densenet121"
# img_datasets="pathmnist,organamnist"
# num_rounds=25
# python -m tracefl.main --multirun check_prov_cache=$check_prov_cache parallel_processes=$parallel_processes do_provenance=True exp_key=$key_start model.name=$img_models dataset.name=$img_datasets num_clients=$num_clients num_rounds=$num_rounds clients_per_round=$clients_per_round dirichlet_alpha=$dirichlet_alpha dry_run=False hydra/launcher=joblib | tee -a logs/prov_med_image_classification_exp.log
# wait
# ## =========================================== Job 2: Provenance Analysis Differential Privacy ===========================================
# echo " ****************** Differential Privacy Provenance ******************"
# num_rounds=15
# dp_noise="0.0001,0.0003,0.0007,0.0009,0.001,0.003"
# dp_clip="15,50"
# num_clients=100
# clients_per_round=10
# check_prov_cache=True
# python -m tracefl.main --multirun check_prov_cache=$check_prov_cache parallel_processes=$parallel_processes do_provenance=True exp_key=DP-text model.name="openai-community/openai-gpt" dataset.name="dbpedia_14" num_clients=$num_clients num_rounds=$num_rounds clients_per_round=$clients_per_round dirichlet_alpha=0.3 noise_multiplier=$dp_noise clipping_norm=$dp_clip dry_run=False hydra/launcher=joblib | tee -a logs/prov_dp_exp_text.log
# python -m tracefl.main --multirun check_prov_cache=$check_prov_cache parallel_processes=$parallel_processes do_provenance=True exp_key=DP-image model.name="densenet121" dataset.name="mnist" num_clients=$num_clients num_rounds=$num_rounds clients_per_round=$clients_per_round dirichlet_alpha=0.3 noise_multiplier=$dp_noise clipping_norm=$dp_clip dry_run=False hydra/launcher=joblib | tee -a logs/prov_dp_exp_image2.log
# python -m tracefl.main --multirun check_prov_cache=$check_prov_cache parallel_processes=$parallel_processes do_provenance=True exp_key=DP-image model.name="densenet121" dataset.name="pathmnist" num_clients=$num_clients num_rounds=$num_rounds clients_per_round=$clients_per_round dirichlet_alpha=0.3 noise_multiplier=$dp_noise clipping_norm=$dp_clip dry_run=False hydra/launcher=joblib | tee -a logs/prov_dp_exp_image2.log
# echo " ****************** 2 Differential Privacy Analysis ******************"
# num_rounds=15
# # dp_noise="0.0001,0.0003,0.0007,0.0009,0.001,0.003"
# # dp_clip="15,50"
# num_clients=100
# clients_per_round=10
# dp_noise=0.003
# dp_clip=15
# python -m tracefl.main --multirun do_provenance=True do_training=True exp_key=DP-text model.name="openai-community/openai-gpt" dataset.name="dbpedia_14" num_clients=$num_clients num_rounds=$num_rounds clients_per_round=$clients_per_round dirichlet_alpha=0.3 noise_multiplier=$dp_noise clipping_norm=$dp_clip dry_run=False | tee -a logs/train_dp_exp_text2.log
# dp_noise=0.006
# dp_clip=10
# python -m tracefl.main --multirun do_provenance=True do_training=True exp_key=DP-text model.name="openai-community/openai-gpt" dataset.name="dbpedia_14" num_clients=$num_clients num_rounds=$num_rounds clients_per_round=$clients_per_round dirichlet_alpha=0.3 noise_multiplier=$dp_noise clipping_norm=$dp_clip dry_run=False | tee -a logs/train_dp_exp_text2.log
# dp_noise=0.012
# dp_clip=15
# python -m tracefl.main --multirun do_provenance=True do_training=True exp_key=DP-text model.name="openai-community/openai-gpt" dataset.name="dbpedia_14" num_clients=$num_clients num_rounds=$num_rounds clients_per_round=$clients_per_round dirichlet_alpha=0.3 noise_multiplier=$dp_noise clipping_norm=$dp_clip dry_run=False | tee -a logs/train_dp_exp_text2.log
# ## =========================================== Job 3: Provenance Analysis Scalablity Experiments ===========================================
# echo " ****************** Scalablity Experiments ******************"
# num_rounds=15
# scaling_clients="200,400,600,800,1000"
# python -m tracefl.main --multirun check_prov_cache=$check_prov_cache parallel_processes=$parallel_processes do_provenance=True exp_key=Scaling model.name="openai-community/openai-gpt" dataset.name="dbpedia_14" num_clients=$scaling_clients num_rounds=$num_rounds clients_per_round=10 dirichlet_alpha=0.3 dry_run=False hydra/launcher=joblib | tee -a logs/prov_scaling_exp_text_total_clients.log
# per_round_clients="20,30,40,50"
# python -m tracefl.main --multirun check_prov_cache=$check_prov_cache parallel_processes=$parallel_processes do_provenance=True exp_key=Scaling model.name="openai-community/openai-gpt" dataset.name="dbpedia_14" num_clients=400 num_rounds=$num_rounds clients_per_round=$per_round_clients dirichlet_alpha=0.3 dry_run=False hydra/launcher=joblib | tee -a logs/prov_scaling_exp_text_clients_per_round.log
# # scalablity with number of rounds
# python -m tracefl.main --multirun check_prov_cache=$check_prov_cache parallel_processes=$parallel_processes do_provenance=True exp_key=Scaling model.name="openai-community/openai-gpt" dataset.name="dbpedia_14" num_clients=400 num_rounds=100 clients_per_round=10 dirichlet_alpha=0.3 dry_run=False hydra/launcher=joblib | tee -a logs/prov_scaling_exp_num_rounds.log
echo "*** Dirchlet Alpha Experiments ***"
num_clients=100
num_rounds=15
dirichlet_alpha="0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1"
python -m tracefl.main --multirun check_prov_cache=$check_prov_cache parallel_processes=$parallel_processes do_provenance=True exp_key=Dirichlet-Alpha model.name="openai-community/openai-gpt" dataset.name="yahoo_answers_topics,dbpedia_14" num_clients=$num_clients num_rounds=$num_rounds clients_per_round=10 dirichlet_alpha=$dirichlet_alpha dry_run=False hydra/launcher=joblib| tee logs/train_dirichlet_alpha.log
python -m tracefl.main --multirun check_prov_cache=$check_prov_cache parallel_processes=$parallel_processes do_provenance=True exp_key=Dirichlet-Alpha model.name="densenet121" dataset.name="pathmnist,organamnist,mnist,cifar10" num_clients=$num_clients num_rounds=$num_rounds clients_per_round=10 dirichlet_alpha=$dirichlet_alpha dry_run=False hydra/launcher=joblib| tee logs/train_dirichlet_alpha.log