Skip to content

Activity

Showing most recent first

Deleted branch

tung-nddeleted iterative_refiner • 
on Nov 28, 2023

Deleted branch

tung-nddeleted iterative • 
on Nov 28, 2023

Deleted branch

tung-nddeleted randomized_within_batch • 
on Nov 28, 2023

Deleted branch

tung-nddeleted iterative_diff • 
on Nov 28, 2023

update notebook

tung-ndpushed 6 commits to iterative_diff • fa67487…1ee293b • 
on Nov 10, 2023

adapt to new pytorch-lightning version

tung-ndpushed 1 commit to iterative_diff • 96464c5…fa67487 • 
on Nov 4, 2023

linear lead time embedding works better

tung-ndpushed 2 commits to iterative_diff • 84d2594…96464c5 • 
on Nov 3, 2023

minor fix

tung-ndpushed 1 commit to iterative_diff • d4a5c6e…84d2594 • 
on Nov 1, 2023

option to use homogeneous lead times only

tung-ndpushed 1 commit to iterative_diff • 7db49e4…d4a5c6e • 
on Nov 1, 2023

different data points in a batch can have different lead times

tung-ndpushed 1 commit to iterative_diff • fd1403e…7db49e4 • 
on Nov 1, 2023

add configs for vit + adaln model

tung-ndpushed 1 commit to iterative_diff • 362adca…fd1403e • 
on Oct 31, 2023

adaptive layernorm for lead time conditioning model

tung-ndpushed 1 commit to iterative_diff • f3ed93a…362adca • 
on Oct 31, 2023

test script for randomized models

tung-ndpushed 1 commit to iterative_diff • 9ba8a18…f3ed93a • 
on Oct 30, 2023

randomize lead times within a batch

tung-ndcreated randomized_within_batch • 872a845 • 
on Oct 29, 2023

multi-step loss with randomized lead time

tung-ndpushed 1 commit to iterative_diff • 78eb2ee…9ba8a18 • 
on Oct 29, 2023

add randomized lead time training

tung-ndpushed 1 commit to iterative_diff • b0e1415…78eb2ee • 
on Oct 28, 2023

implement replay buffer by optimizing last prediction step only

tung-ndpushed 1 commit to iterative_diff • be4b196…b0e1415 • 
on Oct 25, 2023

attempt to use SphericalHarmonics positional embedding but there are …

tung-ndpushed 2 commits to iterative_diff • 5d10375…be4b196 • 
on Oct 23, 2023

update configs to use weigted loss by default

tung-ndpushed 1 commit to iterative_diff • 66b4130…5d10375 • 
on Oct 20, 2023

add an option to use weighted loss for training

tung-ndpushed 1 commit to iterative_diff • e29853f…66b4130 • 
on Oct 17, 2023

allow scaling the attention bias

tung-ndpushed 1 commit to iterative_diff • a320973…e29853f • 
on Oct 17, 2023

fix stupid bug

tung-ndpushed 1 commit to iterative_diff • 7474ad2…a320973 • 
on Oct 16, 2023

add more implementations of position bias

tung-ndpushed 1 commit to iterative_diff • 6ccbbec…7474ad2 • 
on Oct 16, 2023

update default configs

tung-ndpushed 1 commit to iterative_diff • 6132424…6ccbbec • 
on Oct 16, 2023

option to choose attn pbaas scale

tung-ndpushed 1 commit to iterative_diff • 26ad3bb…6132424 • 
on Oct 15, 2023

add a new vit model with absolute positional embedding

tung-ndpushed 1 commit to iterative_diff • bfbc94b…26ad3bb • 
on Oct 15, 2023

update default setting

tung-ndpushed 1 commit to iterative_diff • 0c2293d…bfbc94b • 
on Oct 15, 2023

allow using different base lead times

tung-ndpushed 1 commit to iterative_diff • 830bce9…0c2293d • 
on Oct 14, 2023

update notebook

tung-ndpushed 1 commit to iterative_diff • 24fbb79…830bce9 • 
on Oct 13, 2023

update vit scratch config

tung-ndpushed 1 commit to iterative_diff • 4148321…24fbb79 • 
on Oct 13, 2023