github.com/HyperGAN/HyperGAN commits | Ecosyste.ms: OpenCollective

[fix] alignment test didnt quit with error conditions

57666ed8fd49a017780124e4293e5209bbc24820 authored over 7 years ago by Mikkel Garcia <[email protected]>

[refactor] remove hypergan/util/ops

e5474743bf5bcd8d5ab430d8fee3f99e1d32a94e authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] add refactored tensorflow ops

9ac07bb9cffeb62d699aa3047e0fa7e8befaf77f authored over 7 years ago by Martyn Garcia <[email protected]>

[refactor] hypergan/ops/

25a3faa6873cf17135b70dea969ba81c14349c5b authored over 7 years ago by Martyn Garcia <[email protected]>

[refactor] remove legacy ops

fec16db886e9d583e89b45cf7f1a3f187dd63621 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] config.block

c84e0edc8a7beb29f4c389e3d118d1160fba3384 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] part of the graph construction

837a923b7dc6a4713ea2c3cc343490c2b694b3a2 authored over 7 years ago by Martyn Garcia <[email protected]>

[wip] generator tests

2d5b1c89b3ec4e6eaf0d6365fff9aad214bbc991 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] variable width sampling on static batch sampler

8a0721ca1db982c364078a5d5457ee1fd64f41fa authored over 7 years ago by Martyn Garcia <[email protected]>

[refactor] generator with tests

be72d6ca6d1008d2cc1a655b23d0a6e0c56eb9a2 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] alignment test fixes

2bd6b5ad47da7eb7e6ecd1e65e110958f3608ed2 authored over 7 years ago by Mikkel Garcia <[email protected]>

[feature] alignment test

cc8c2ed1a3ede46ec178e0b219e18c8456e34a16 authored over 7 years ago by Mikkel Garcia <[email protected]>

[test] more of TensorflowOps

5896fafb2850cdd0eb450d71d10c43f655eef830 authored over 7 years ago by Martyn Garcia <[email protected]>

[feature] tests on tensorflow_ops

48e40e24393f9fb4d7eb04da62d5f51aa1b6690f authored over 7 years ago by Martyn Garcia <[email protected]>

[wip] cleanup generator and ops

de86b350ab97fffd35a586e0ce37dabd93abfe14 authored over 7 years ago by Martyn Garcia <[email protected]>

[feature] match_discriminator_encoder makes the encoded z the same size as the output from the discriminator. Allows for easier alignment (no math to calculate z)

122e656cb4bf62e1b0941490af6b7247e4f4d5c3 authored over 7 years ago by Mikkel Garcia <[email protected]>

[wip] suggested refactorings

20a20dc4132b9e00b62959a76793662847184def authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] sample only within -1...1

3ae4715b62f1e5e4ef1603c7dbe8f3ffa5afd134 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] static sampler

bf1a9fe739d17f2f846d61c47ecdadade7bef142 authored over 7 years ago by Mikkel Garcia <[email protected]>

[fix] static batch test search

5fd1a901451a8656324a65d1ee9e66b42df16209 authored over 7 years ago by Mikkel Garcia <[email protected]>

[fix] softmax with began

9e6bd7156b0b50dbe470da387bf94a7260d1a84e authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] range

a0b00679301c8df3cf16e035e58ee4ba332251a0 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] integer division

030afd213a4071a16b64f19e0986f3215f660f39 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] syntax

558dbbd8e3324c937cf538d5b951ede70d6d22e8 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] colorizer variable sampler

36eb45ae7a6054d5b3a50bf700ff91951b7f6ab1 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] add layer filter to auteonceder for began

a921f0587701d1b744fdb3df0d4e34a070efe3c6 authored over 7 years ago by Martyn Garcia <[email protected]>

Merge branch 'align-began-exp' into aligned-began

* align-began-exp:
[fix] various bugs and cleanup
[hack] sampler

611a38d509e843a0fea9ba46e849ef21a1720e37 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] various bugs and cleanup

9f7643f3b53f562e7593339775202b73c7174ed8 authored over 7 years ago by Mikkel Garcia <[email protected]>

[hack] sampler

a3794d3614e24087633fc5a0d47421f97c4417ca authored over 7 years ago by Mikkel Garcia <[email protected]>

Merge pull request #82 from 255BITS/aligned-began-exp

Aligned began experiment 2

126cb76a1782a73cdc5243926d192bd270bf5446 authored over 7 years ago by martyn <[email protected]>

[fix] use abs distance instead of square for reconstruction

af85dbf1552d8b8e3f3280be812e920f6266b2f2 authored over 7 years ago by Mikkel Garcia <[email protected]>

[fix] bound k to 0..1

9145d628ada08d85b988467c2766257f3b77509f authored over 7 years ago by Martyn Garcia <[email protected]>

[wip] optimize and generalize xb to mean all other classes

60be170852ff517699ad71f33d96fdecc383aaac authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] add more information to z

7dc1ffa4e853e01cb1479c3fff23f1c2229d73f5 authored over 7 years ago by Martyn Garcia <[email protected]>

[experiment] more than 2 classes

667f109bd385f2431dac9f70c467b9e73ffe9902 authored over 7 years ago by Martyn Garcia <[email protected]>

[optimize] reduce number of networks created

84913019beaaa2509e47dd1bfdb087fa0940bce9 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] crelu filter doubling

7cc289def35a0419c27d7a47e15a5bf82ebbd391 authored over 7 years ago by Martyn Garcia <[email protected]>

[feature] include_xba and include_recdistance for experiments

21340df8028270311cd9c7f08e48b17b6d20d056 authored over 7 years ago by Martyn Garcia <[email protected]>

[experiment] new include flags

9b8810328883f37220bcc523208c537e0d183feb authored over 7 years ago by Mikkel Garcia <[email protected]>

[fix] only show 1x5 samples

86ee8728ebdfe1e7194c1eff6e42baf65997745b authored over 7 years ago by Mikkel Garcia <[email protected]>

[feature] reconstruction loss in losses

46bb83d1a8d074caf5268ce0ae46e0282d42859a authored over 7 years ago by Mikkel Garcia <[email protected]>

[experiment] a bunch of different of includes

9f0afe00dbc8a75a9c4a058bda1c747388570819 authored over 7 years ago by Mikkel Garcia <[email protected]>

[optimize] share weights across the network

71bb0fed2e4fa8bf34bbc8d435aae9116feeab06 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] remove unused loss term

79988630fc431a7066db7ff9d1d85553601d4b6c authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] use same weights in generators

5c8202ad7f28171d364d471d18fe24f90a62d861 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] error signal

3d31c13a44394c0a8d0f16178c863ae65f7a349a authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] xbaab and xabba

8b0f115b1ee40060f9013db8b643a96f730ead2e authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] sampling

f65a7a64f2ecc73a6ab95c39b5edb0a2a3fe468c authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] reuse xba

e0811c0f435710aef42924740d3fbb7de6549dd2 authored over 7 years ago by Martyn Garcia <[email protected]>

[optimize] share d weights

ed1e9d9d72687d42a6d6fe3c0da138727636646e authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] alignment

88089f945a92dccb56995b70823f8514c62a019e authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] include_g* flags

502b5a0346eb1614d31c444526689901fcacc7f7 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] error calculation

4fe35a7d701dd4ef54f8b8e079afc9f8d74f7af2 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] layer_regularizer inherited in generator / discriminator

0a7ec9b6dcd152ca382a0c0a1b64c6a1d6e66730 authored over 7 years ago by Mikkel Garcia <[email protected]>

[fix] custom loss

0847273e9d4b20b66ad5910fc2f64810c4ba3f5f authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] began sampler ordering

439661f6d48843e328abe0168e8f0163abf4f546 authored over 7 years ago by Martyn Garcia <[email protected]>

[feature] alignment through multiple generators

d63abee54ded450bf2fd21921c3031ea83386f84 authored over 7 years ago by Mikkel Garcia <[email protected]>

[experiment] aligned began

95e225220123c4a70931024ff8a58f94561ef9fe authored over 7 years ago by Martyn Garcia <[email protected]>

Merge pull request #76 from 255BITS/boundary-equilibrium-loss

Boundary Equilibrium Loss (BEGAN)

756c150c674ffca22e57393fa9513069ebbf0e98 authored over 7 years ago by martyn <[email protected]>

[fix] dg_sum

82f60885a2d0c3b75dfdaf84c7f3b285f1e1b6b0 authored over 7 years ago by Mikkel Garcia <[email protected]>

[feature] static-batch-test updated with --config_list for testing multiple base configs

6747dd9798e779f6f85c543e3babe56ed9e38174 authored over 7 years ago by Mikkel Garcia <[email protected]>

[fix] rename mispelled file

c20fab9326883426aaf6d79aa156e5b4c318f935 authored over 7 years ago by Martyn Garcia <[email protected]>

[feature] update search criteria in 2d-test

f7e81ed14389a03e206c854c513b73b5376355a2 authored over 7 years ago by Mikkel Garcia <[email protected]>

[fix] s/encoder/decoder

314a1b5cc7140a92d800f938347c8af655927734 authored over 7 years ago by Martyn Garcia <[email protected]>

[feature] generator.decoder_layer_regularizer indepence from generator(began)

18f8ab48df818accc9a8ee8895e4751ed5204f43 authored over 7 years ago by Martyn Garcia <[email protected]>

[feature] support for minibatch

89cc45b4e1a99abb5d88af874cf8a83571f0aa16 authored over 7 years ago by Martyn Garcia <[email protected]>

[feature] strided block

801763372ae8028328a1dd8c36b8f415a21247e7 authored over 7 years ago by Martyn Garcia <[email protected]>

Merge pull request #74 from 255BITS/depth-to-space-generator

Depth to space generator

9919a3d1f58fe214fc4bb535ebf97c49cab0e5fb authored over 7 years ago by martyn <[email protected]>

Merge pull request #77 from 255BITS/improved-lsgan-boundary-equilibrium

began fixes

e1fa328d001d7d84c16c7ccbfd76fbd4cc236902 authored over 7 years ago by martyn <[email protected]>

Merge branch 'boundary-equilibrium-loss' into improved-lsgan-boundary-equilibrium

5ab839fc8bfd1799c27ddbded1c80207486410dd authored over 7 years ago by martyn <[email protected]>

[fix] s/joint_trainer/propotional_control_trainer

7bcce4a3e03aca3dc2582a578eb00a9a5f0348f7 authored over 7 years ago by Martyn Garcia <[email protected]>

[feature] discriminator option `config.block`

1248444e8190c554c37e87f4710f406404633805 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] add prefix to discriminator

b9e0ec509effd21c57217df24c5b3682879bbd37 authored over 7 years ago by Martyn Garcia <[email protected]>

[fix] gradient penalty with began+lsgan

17642e4fd857567075118209e57175c6bc9e0952 authored over 7 years ago by Martyn Garcia <[email protected]>

[feature] expand options

8a6710ed9ddf7476b853822e405434fd94df5714 authored over 7 years ago by Martyn Garcia <[email protected]>

[feature] add batch diversity measurement to find mode collapse

fb2cbb051099fadbdd7e2e4f0d2e1fc83ac49019 authored over 7 years ago by Mikkel Garcia <[email protected]>

[feature] static batch test uses additive g

577775f81213b877f53cdd3d262c65b0ce6bd519 authored over 7 years ago by Mikkel Garcia <[email protected]>

[fix] output error

db6f8e27d3b2c1f99fd18e17b33e926bfcff5167 authored almost 8 years ago by Martyn Garcia <[email protected]>

[fix] runtime errors from refactoring

651b185e51b8c4906ed9556dc057c4847474b669 authored almost 8 years ago by Martyn Garcia <[email protected]>

[fix] return k from began constructor

772c3500998f254b36699c33f04cd4a9e83fe0f0 authored almost 8 years ago by Martyn Garcia <[email protected]>

[refactor] began loss function

870360d3a66bc7c7c14f08b049e060e5980bec8c authored almost 8 years ago by Martyn Garcia <[email protected]>

[feature] discriminator config.block

1427272a07394ecff926e4d059a20486b5336815 authored almost 8 years ago by Martyn Garcia <[email protected]>

[refactor] s/.../gamma_d_real/

86c3101e805e2e0b543aa2ec251186a3964cc3c3 authored almost 8 years ago by Martyn Garcia <[email protected]>

[fix] lsgan k_loss

0b383c94d8e5854c36c29d34b37b3544b6ddae86 authored almost 8 years ago by Martyn Garcia <[email protected]>

[feature] additive pyramid discriminator

3bfb83ba2d55103b4dc3ac7fffb1f516fe8e491f authored almost 8 years ago by Martyn Garcia <[email protected]>

[experiment] --align(not working)

63d850024ca099c5c0d0c01df9b3d1c9bcea942c authored almost 8 years ago by Martyn Garcia <[email protected]>

[fix] config search

d20cd1afdd6ff6c3e41f8473c78a5c0accc19616 authored almost 8 years ago by Martyn Garcia <[email protected]>

[fix] began

e449b98cd04f916b78c35bb70763504d95dea979 authored almost 8 years ago by Martyn Garcia <[email protected]>

[fix] k calculation on lsgan

0e2b2c3e674e8b4dc749a83cba6dbe2548163483 authored almost 8 years ago by Martyn Garcia <[email protected]>

[fix] more fixes for began

2532a3c3c431be1734ae2e91225538cdceac63e8 authored almost 8 years ago by Martyn Garcia <[email protected]>

[fix] gamma calculation

307806c3cea5dc055675c2abb63a1be0184188e3 authored almost 8 years ago by Martyn Garcia <[email protected]>

[feature] be loss search

5892513dde95143c08602d949367c823515d1eec authored almost 8 years ago by Martyn Garcia <[email protected]>

Merge branch 'depth-to-space-generator' into improved_wgan

* depth-to-space-generator:
[fix] initial checkerboard artifacts
[feature] depth to space ge...

1d8f7b1aaf349cad06e64adde44b5505a0b82b75 authored almost 8 years ago by Martyn Garcia <[email protected]>

[fix] gradient policy shape

4561754c01fa1ddbf279a7f43e319c6a74cda200 authored almost 8 years ago by Martyn Garcia <[email protected]>

[fix] lambda

1e23990881330a1d8abac7bc96ec5e3546918f26 authored almost 8 years ago by Martyn Garcia <[email protected]>

[fix] train for longer

94fff965a8a23f8065b1b5ac4e934feeb3f3a1fa authored almost 8 years ago by Martyn Garcia <[email protected]>

[fix] tweak

907d4e39ffac96690ae0ab4d7724f2a150160d8d authored almost 8 years ago by Martyn Garcia <[email protected]>

[fix] disable autoencoder d

a781f26e678c22695bee51f207d2410218ee0978 authored almost 8 years ago by Martyn Garcia <[email protected]>

[fix] search wgan

124d0406cedf52a896c962d286834a34db9af967 authored almost 8 years ago by Martyn Garcia <[email protected]>

[feature] loss functions can have gradient penalty as per https://arxiv.org/abs/1704.00028

0e92208bc995f547f4b1faae18102ad01ee1659f authored almost 8 years ago by Mikkel Garcia <[email protected]>