Skip to content

IWAE with KL penalty #52

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
maxtretiakov wants to merge 14 commits into
base: main
Choose a base branch
from
Open

IWAE with KL penalty #52

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
14 commits
Select commit Hold shift + click to select a range
267bec3
adapting iwae to the new code
maxtretiakov May 25, 2025
53ba964
added new configs
maxtretiakov May 25, 2025
7a94253
fixed init
maxtretiakov May 25, 2025
cd38adc
added inner vars for prob layer
maxtretiakov May 25, 2025
97b6e4d
adding beta
maxtretiakov May 25, 2025
534e9d2
fix
maxtretiakov May 25, 2025
cdb89cb
hyp fixes
maxtretiakov May 25, 2025
1c884a6
add mnist iwae
Yauhenii May 25, 2025
c50413f
log wandb
Yauhenii May 25, 2025
5874bdb
add experiments
Yauhenii May 25, 2025
41950c5
iwae using kl penalty
maxtretiakov May 26, 2025
644fc7f
addig kl penalty
maxtretiakov May 26, 2025
083480c
add experiments
Yauhenii May 26, 2025
ed099bf
added-configs
maxtretiakov May 27, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
63 changes: 63 additions & 0 deletions config/iwae-cifar-penalty/fclassic_conv15_sigma_0.01_kl_0.0001_lr_0.01.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,63 @@
bound:
delta: 0.025
delta_test: 0.01
dist_init:
seed: 110
factory:
bounds:
- kl
- mcallester
data_loader:
name: cifar10
params:
dataset_path: ./data/cifar10
losses:
- nll_loss
- scaled_nll_loss
- 01_loss
metrics:
- accuracy_micro_metric
- accuracy_macro_metric
- f1_micro_metric
- f1_macro_metric
model:
name: conv15
params:
dataset: cifar10
in_channels: 3
posterior_objective:
name: fclassic
params:
delta: 0.025
kl_penalty: 1.0
prior_objective:
name: fclassic
params:
delta: 0.025
kl_penalty: 0.0001
log_wandb: true
mcsamples: 1000
pmin: 5.0e-05
posterior:
training:
epochs: 1
lr: 0.001
momentum: 0.9
seed: 1135
prior:
training:
epochs: 100
lr: 0.01
momentum: 0.95
seed: 1135
sigma: 0.01
split_config:
batch_size: 250
dataset_loader_seed: 112
seed: 111
split_strategy:
prior_percent: 0.7
prior_type: learnt
self_certified: true
train_percent: 1.0
val_percent: 0.0
62 changes: 62 additions & 0 deletions config/iwae-cifar-penalty/iwae_conv15_sigma_0.001_kl_0.0001_lr_0.0001.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,62 @@
bound:
delta: 0.025
delta_test: 0.01
dist_init:
seed: 110
factory:
bounds:
- kl
- mcallester
data_loader:
name: cifar10
params:
dataset_path: ./data/cifar10
losses:
- nll_loss
- scaled_nll_loss
- 01_loss
metrics:
- accuracy_micro_metric
- accuracy_macro_metric
- f1_micro_metric
- f1_macro_metric
model:
name: conv15
params:
dataset: cifar10
in_channels: 3
posterior_objective:
name: bbb
params:
kl_penalty: 1.0
prior_objective:
name: iwae
params:
n: 10
kl_penalty: 0.0001
log_wandb: true
mcsamples: 1000
pmin: 5.0e-05
posterior:
training:
epochs: 1
lr: 0.001
momentum: 0.9
seed: 1135
prior:
training:
epochs: 100
lr: 0.0001
momentum: 0.95
seed: 1135
sigma: 0.001
split_config:
batch_size: 250
dataset_loader_seed: 112
seed: 111
split_strategy:
prior_percent: 0.7
prior_type: learnt
self_certified: true
train_percent: 1.0
val_percent: 0.0
62 changes: 62 additions & 0 deletions config/iwae-cifar-penalty/iwae_conv15_sigma_0.005_kl_0.00005_lr_0.0001.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,62 @@
bound:
delta: 0.025
delta_test: 0.01
dist_init:
seed: 110
factory:
bounds:
- kl
- mcallester
data_loader:
name: cifar10
params:
dataset_path: ./data/cifar10
losses:
- nll_loss
- scaled_nll_loss
- 01_loss
metrics:
- accuracy_micro_metric
- accuracy_macro_metric
- f1_micro_metric
- f1_macro_metric
model:
name: conv15
params:
dataset: cifar10
in_channels: 3
posterior_objective:
name: bbb
params:
kl_penalty: 1.0
prior_objective:
name: iwae
params:
n: 10
kl_penalty: 0.00005
log_wandb: true
mcsamples: 1000
pmin: 5.0e-05
posterior:
training:
epochs: 1
lr: 0.001
momentum: 0.9
seed: 1135
prior:
training:
epochs: 100
lr: 0.0001
momentum: 0.95
seed: 1135
sigma: 0.005
split_config:
batch_size: 250
dataset_loader_seed: 112
seed: 111
split_strategy:
prior_percent: 0.7
prior_type: learnt
self_certified: true
train_percent: 1.0
val_percent: 0.0
62 changes: 62 additions & 0 deletions config/iwae-cifar-penalty/iwae_conv15_sigma_0.005_kl_0.0001_lr_0.0001.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,62 @@
bound:
delta: 0.025
delta_test: 0.01
dist_init:
seed: 110
factory:
bounds:
- kl
- mcallester
data_loader:
name: cifar10
params:
dataset_path: ./data/cifar10
losses:
- nll_loss
- scaled_nll_loss
- 01_loss
metrics:
- accuracy_micro_metric
- accuracy_macro_metric
- f1_micro_metric
- f1_macro_metric
model:
name: conv15
params:
dataset: cifar10
in_channels: 3
posterior_objective:
name: bbb
params:
kl_penalty: 1.0
prior_objective:
name: iwae
params:
n: 10
kl_penalty: 0.0001
log_wandb: true
mcsamples: 1000
pmin: 5.0e-05
posterior:
training:
epochs: 1
lr: 0.001
momentum: 0.9
seed: 1135
prior:
training:
epochs: 100
lr: 0.0001
momentum: 0.95
seed: 1135
sigma: 0.005
split_config:
batch_size: 250
dataset_loader_seed: 112
seed: 111
split_strategy:
prior_percent: 0.7
prior_type: learnt
self_certified: true
train_percent: 1.0
val_percent: 0.0
62 changes: 62 additions & 0 deletions config/iwae-cifar-penalty/iwae_conv15_sigma_0.005_kl_0.001_lr_0.0001.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,62 @@
bound:
delta: 0.025
delta_test: 0.01
dist_init:
seed: 110
factory:
bounds:
- kl
- mcallester
data_loader:
name: cifar10
params:
dataset_path: ./data/cifar10
losses:
- nll_loss
- scaled_nll_loss
- 01_loss
metrics:
- accuracy_micro_metric
- accuracy_macro_metric
- f1_micro_metric
- f1_macro_metric
model:
name: conv15
params:
dataset: cifar10
in_channels: 3
posterior_objective:
name: bbb
params:
kl_penalty: 1.0
prior_objective:
name: iwae
params:
n: 10
kl_penalty: 0.001
log_wandb: true
mcsamples: 1000
pmin: 5.0e-05
posterior:
training:
epochs: 1
lr: 0.001
momentum: 0.9
seed: 1135
prior:
training:
epochs: 100
lr: 0.0001
momentum: 0.95
seed: 1135
sigma: 0.005
split_config:
batch_size: 250
dataset_loader_seed: 112
seed: 111
split_strategy:
prior_percent: 0.7
prior_type: learnt
self_certified: true
train_percent: 1.0
val_percent: 0.0
62 changes: 62 additions & 0 deletions config/iwae-cifar-penalty/iwae_conv15_sigma_0.01_kl_0.0001_lr_0.0001.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,62 @@
bound:
delta: 0.025
delta_test: 0.01
dist_init:
seed: 110
factory:
bounds:
- kl
- mcallester
data_loader:
name: cifar10
params:
dataset_path: ./data/cifar10
losses:
- nll_loss
- scaled_nll_loss
- 01_loss
metrics:
- accuracy_micro_metric
- accuracy_macro_metric
- f1_micro_metric
- f1_macro_metric
model:
name: conv15
params:
dataset: cifar10
in_channels: 3
posterior_objective:
name: bbb
params:
kl_penalty: 1.0
prior_objective:
name: iwae
params:
n: 10
kl_penalty: 0.0001
log_wandb: true
mcsamples: 1000
pmin: 5.0e-05
posterior:
training:
epochs: 1
lr: 0.001
momentum: 0.9
seed: 1135
prior:
training:
epochs: 100
lr: 0.0001
momentum: 0.95
seed: 1135
sigma: 0.01
split_config:
batch_size: 250
dataset_loader_seed: 112
seed: 111
split_strategy:
prior_percent: 0.7
prior_type: learnt
self_certified: true
train_percent: 1.0
val_percent: 0.0
Loading