Skip to content
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

Jump to bottom

Real galaxies #9

Merged
merged 72 commits into from
Mar 12, 2024
Merged

Real galaxies #9

merged 72 commits into from
Mar 12, 2024

Conversation

swagnercarena
Copy link
Owner

Merge branch into main given that this branch is currently behaving like the default main branch.

Sebastian Wagner-Carena and others added 30 commits March 16, 2023 21:22
added the ability to include random rotations in the image generation…
ee8d437 
…. This allows for random rotations despite a fixed image set.
typo in the license
9d686d6
can't have the rotation angle be an optional keyword with vmap
0c52df2
added the ability to specify configuration paths instead of defaultin…
941e19c 
…g to the same path.
bug in config loader
f879ea2
config for lsst data
8b567df
correct exposure times
853826d
variable schedule in the training config
77f95d2
should not use is literal
3c8c254
changed cycling in fixed number of random examples to be drawn from r…
8e3c3f1 
…andom permutations, and added the first portion of the hierarchical inference code. Still need to finish with tests.
finished hierarchical inference code update with tests, added utility…
2b0d7bf 
… function for random permutations, and changed the train script to use random permutations instead of a default cycle over images
added script to run hierarchical inference
ef45601
made it possible to disentangle the normalization configuration from …
152c20f 
…the configuration from which examples are drawn.
added normalization config
42d7d99
wrong part of config gets used in case of Nonre for normalizing_config
2c64fb5
added snpe_c loss to train and tested the loss.
ddc39f2
training script for snpe. Very rough, will need some cleanup later.
b25913c
didn't add get_outputs change
45e2573
should still be same total number of training steps for learning rate
1f749b2
need to do a deepcopy of the config to make sure that we don't overwr…
3e30721 
…ite it during iterative retraining
added linear learning rate and switched from jax.random.shuffle to ja…
e9ba87e 
…x.random.permutation.
small bug in learning rate
0c910ab
@swagnercarena
made folder for training configs and added a number
975ecff
@swagnercarena
added some changes to hierarchical inference
28a6bcd
added much smaller models to force learning plateau
e8b0442
added training procedure with sgd
9dea323
@swagnercarena
new lens models that express everything in Cartesian coordinates
3415c39
@swagnercarena
modified the way the learning rate is treated for snpe and added new …
07a464f 
…configs for each of the start options
@swagnercarena
expanded number of dimensions for the prior and initial proposal dist…
4555c99 
…ribution
@swagnercarena
hardcoded normalization in train_snpe
43f6bf6
Sebastian Wagner-Carena and others added 28 commits September 11, 2023 21:47
added numba and removed flax version
9057ac3
modifications for newer versions of python
9583807
@swagnercarena
added feature for reversing the normalization function. For now just …
6e86f11 
…a handy function for visualization purposes, but should probably be integrated more broadly in the future.
@swagnercarena
added the ability to support multiple Gaussian distributions in the e…
0062555 
…ncoding and to add them with a decay factor
@swagnercarena
updates to training scripts to allow for mixture of proposals
1a4ce05
@swagnercarena
added decay factor to old configs
a5b05cd
@swagnercarena
small print statement
7fc88e4
@swagnercarena
Add a regularizer to avoid exploding gradients
ac439f3
@swagnercarena
new config files with varied decay
27d4254
@swagnercarena
hot fixed an issues with the process dimension and left a TODO for a …
647202a 
…more permanent fix.
@swagnercarena
added configs to compare decay on instant snpe
c81dd8f
@swagnercarena
new configs assume 4 gpus
68bfccc
@swagnercarena
typo in one of the print statements
b09d56b
@swagnercarena
fixed bug in the way the reference image was being used that caused t…
5a87d7c 
…he proposal distribution to be a mix of the training mean and the actual image of interest.
@swagnercarena
allow for average distribution
0c01e52
instant with only 10 epochs of generic
2a3ceb2
@swagnercarena
instant with averaging
6e6b9c6
@swagnercarena
long sequential run with lambda 0.0
5d1698f
@swagnercarena
medium with instant decay
f13c53d
@swagnercarena
updated basic training configuration
69b8f00
clean train configs
9b5cc7e
@swagnercarena
finished cleanup of training configuration files
65a71d1
@swagnercarena
changed number of steps per epoch
aeb50f5
@swagnercarena
added resnet 34 config
998d434
@swagnercarena
added resnet18 config
03c66d4
@swagnercarena
reverting mistaken change to one config file
d1ef688
@swagnercarena
ResNetD model added to library
5046dfe
@swagnercarena
resnetd config
8b90b90
@swagnercarena swagnercarena merged commit d9b3584 into main Mar 12, 2024
@swagnercarena swagnercarena deleted the real_galaxies branch March 12, 2024 19:43
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
1 participant
@swagnercarena