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

[fix] variable minibatch discriminator

5b522a3b1952bd1b531ceffa07138639b936f030 authored over 7 years ago

[fix] alphagan with segment generator

f7e95dbbbc7519bea2556d6a29d6fe33c1689d7f authored over 7 years ago

[refactor] alphagan

cd3155b6d14375bd11f68146e45358964b1f1148 authored over 7 years ago

[refactor] remove explicit .create()

1e09483cb99fd67fb0831d35c0284d9d556aa5f4 authored over 7 years ago

[fix] alphagan without segment generator

3d780878b08fc68d705f2558c55dda2578fd25ad authored over 7 years ago

[fix] parameterize fully connected discriminator

89fce6b0e5baf9c3b844c068d37956912f9f47e4 authored over 7 years ago

[fix] .reuse in loss instead of .create

3321a4a46d315e0d9bf5888791357c5f29be0884 authored over 7 years ago

[fix] alphagan

9d1439c5b2a115fb749b1b9ea3d8927c4ae4a053 authored over 7 years ago

[fix] baseloss D weight

2f17874144e332138f0d9eba63ec3eb6558e7d9f authored over 7 years ago

[fix] re-enable weight_norm2

f524d33e389d6cb272e8645881af69bb312b4eb2 authored over 7 years ago

[wip] try to revert changes to working state

fdf51c45a643a030d186d8a411f1fd11d7443e82 authored over 7 years ago

[fix] error signal on d was doubled

ebeb437ef78b982e987f949f406c8bf918e5187e authored over 7 years ago

[feature] clamped_unit - binds between -1 and 1 as an alternative to tanh. Name needs work

f2f07b35498f04ce120ec22d4eb7dae501f233d2 authored over 7 years ago

[fix] sampling on multisegment

23028b4a76e13c230e3cd13708c423dfa5920543 authored over 7 years ago

[fix] sampling

2598fba55a35d78feaa7777bbae1126344647b49 authored over 7 years ago

[fix] segment sampling

d098a23efb21725065dac28bd02ee42a2c553a4a authored over 7 years ago

[fix] alphagan segment sampling

e6fcc8a396b84adc998a1981c19b54f0796b938c authored over 7 years ago

[feature] multiple discriminators

21fd03c07956c24bde9dd714dc10334b9ba78999 authored over 7 years ago

[fix] gan.channels

2f8a15298c3b43033b40374f58fd3101cf2b80e4 authored over 7 years ago

[fix] revert back to batch norm for least-squares-gan-optimized

259b2b24c2b6b1aa5edc5ecb0cd11b9f8dc50294 authored over 7 years ago

[fix] base loss generator / d_net

1cf0761803883c6ccf2405048b98bd3d382a2b0d authored over 7 years ago

[fix] deconv after layer filter

dca04ca9e43798b0a23fb28acfdfd897479223a4 authored over 7 years ago

Merge branch 'working' of github.com:255BITS/HyperGAN into working

* 'working' of github.com:255BITS/HyperGAN:
[fix] loading aligned gan
[fix] cli variable con...

7ae2e2ad65f553209fb43c6d8ca45e4c7e9a95e5 authored over 7 years ago

[feature] shake_weights on trainer to help with saddle points

11ba28f5cf141eaaf9090c59acda879435d0f0b0 authored over 7 years ago

[fix] gradient_locally_stable

86ffd7f4e1a2f9d90bb4ecacd1b1369191d7abcb authored over 7 years ago

[feature] multi generator gan using gan to distinguish generators

9a192d9df48c903ad529024c611cc2fab9d3f6cd authored over 7 years ago

[fix] default z_cycle_loss to 0

36bda5192595e83c5f5088babe4b0ac2148daee4 authored over 7 years ago

[fix] loading aligned gan

289961737c6fbd2b355e43489691dbfb61b571f2 authored over 7 years ago

[fix] cli variable config_name was missing

4cbcab6c6787a592c2c04449ece26f55c40b73cf authored over 7 years ago

[fix] sampling for segment_generator with custom mask_generator

83b91baf1f126293e59d9abf6ca92193344bb37b authored over 7 years ago

[feature] MUCH better search criteria

0aca37aee1d30ac0725c698394327e9bcba07abd authored over 7 years ago

[feature] fgan updates

25e8ae04d0a2489be8ec2c200795c2bf30d7754b authored over 7 years ago

[feature] implement many new fgan generators / regularizers

51a2c4c53a1a5930b97660f6c3a8bebbb982dec4 authored over 7 years ago

[feature] config name in output

1166a4754002a1be69e0180575cdc61dd8889666 authored over 7 years ago

[feature] multi loss

8721bff81f4dcd5be88fae4e5facb2274b7f9ab1 authored over 7 years ago

[feature] multi generator

fad62a904f38e77110c72c2f023eda396d014070 authored over 7 years ago

[fix] look for generator in the name for multisteptrainer

efffb5226501f2a2a86659896c8ec453cc3e4b38 authored over 7 years ago

[feature] search for fgan

de075d7161af483f0d76724028166d6a5de75dd5 authored over 7 years ago

[feature] activation \"clamped\" clamps between 0 and 1

30829de6405c8015c61ddf8717d6d3316b22510c authored over 7 years ago

[feature] fgan updates

11dc6fbfb2edb01b91fad70fce5907ae7174d5e7 authored over 7 years ago

[refactor] reorganize loss regularizers

8cf42cb9725e4d9d3e8080d2205ddbc59ad64446 authored over 7 years ago

[fix] allow for loss reuse

04c41956e539e2c4c71ac9db6ffe25f4a93437b7 authored over 7 years ago

[feature] allow for mask_generator to return a sigmoid final activation

46ad176ea5655925fc880d93a945e387547136d2 authored over 7 years ago

[feature] abstract out project_from_prior and remove convs in skip_linear, breaks graph

e9b8ff3f91247015f1090faa07db3c3f991f96f9 authored over 7 years ago

[feature] alphagan introduces z_cycloss, breaks graph

add828a63e7ac68bda6284249bf22310db476b24 authored over 7 years ago

[refactor] remove duplicate step() definitions

d7ad56c20e18a09ae7237094e1fbe6d5ed6797bb authored over 7 years ago

[feature] allow z to be passed into the encoder objects

77130e8d80e872729b1faf00efb071df5c970596 authored over 7 years ago

[fix] discriminator reuse

cd0022742d246b69d1c7557e32b93c1dc812f26d authored over 7 years ago

[fix] 2d distribution test

779ced0fabaaad57feedf544f452a69cf1c9baf6 authored over 7 years ago

[feature] tensorflow debug mode

7e93fa5c037077ce82d8ef8ae8a9ff99c0b438c0 authored over 7 years ago

Merge branch 'alphagan-updates' of github.com:255BITS/HyperGAN into weight_and_cosine_norm

* 'alphagan-updates' of github.com:255BITS/HyperGAN:
[feature] gradient penalty working with f...

01e82dd4c30baceac888d9fb4017f469fae461a5 authored over 7 years ago

[feature] gradient penalty working with fgan part 2

c129d4b998705d63919f47eb2052871830943dac authored over 7 years ago

[feature] gradient penalty working with fgan

528e9560ef5910852e9d9a362580c9e0cb92989b authored over 7 years ago

Merge branch '0.10' of github.com:255BITS/HyperGAN into alphagan-updates

* '0.10' of github.com:255BITS/HyperGAN:
[feature] g_loss_type = standard | rkl | kl | alpha
...

107d97d17ceb8c497cf45318648083861a9f068c authored over 7 years ago

Merge branch 'segment_generator' of github.com:255BITS/HyperGAN into weight_and_cosine_norm

* 'segment_generator' of github.com:255BITS/HyperGAN:
[feature] new weight norm implementation...

3ffe7dc059ffb9ef00363aa58f657ef1374f84f7 authored over 7 years ago

[feature] limit range of f divergences to avoid explosion

9433a92d3464e1805555818e8ff14862199bc372 authored over 7 years ago

[feature] normalized softplus activation

391532596595dbb0340a7274fe4e94a3ea4e165e authored over 7 years ago

Merge branch '0.10' of github.com:255BITS/HyperGAN into weight_and_cosine_norm

* '0.10' of github.com:255BITS/HyperGAN:
[feature] g_loss_type = standard | rkl | kl | alpha
...

a0394f2e459757ce69bec04074cd05ee728df013 authored over 7 years ago

[feature] g_loss_type = standard | rkl | kl | alpha

56c9a6be57100cd7c32972db5b097e8ac33a8654 authored over 7 years ago

Merge pull request #156 from 255BITS/0.9.x

Update 0.10 from 0.9.x

29df8a31ceaa18f687c4d23bca20bd26d57d1671 authored over 7 years ago

Merge branch '0.10' into 0.9.x

d314a06f52e4b3db448dfbdf9d385ad88633abc7 authored over 7 years ago

[feature] new weight norm implementations. 2 works best right now

97afa30fedc4e266f161f431b0817ad89364e3cf authored over 7 years ago

[feature] new weight norm implementations. 2 works best right now

ecd04f01408e69da3976f6dacbc24f977e4ac276 authored over 7 years ago

Merge pull request #152 from 255BITS/fgan

f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization

71127f932098de4109cffa725f37c62b98f2a259 authored over 7 years ago

[feature] another generator that shares output generator

cca64bf03d5a3c8828a8a53c5ede27725041b932 authored over 7 years ago

Readme review (#112)

* [doc] clarify about section

* [doc] initial features section

* [doc] json feature

* [...

ccedf1ac8298b942bcc3b42e16c7779709428e62 authored over 7 years ago

Getting started (#143)

* [doc] getting started

* [fix] table

* [doc] copy

* [doc] remove unco...

9622d733bc0f2be27a4cdcdace2f9a9afe6b4ede authored over 7 years ago

Dragan (#135)

* [fix] chargan

* [feature] dragan gradient penalty regularizer

* [fix] improved-gan and i...

2c492adcec2019dac14f96a7186e2bf145a3a823 authored over 7 years ago

[fix] batch_size minimum

0c9ca46fd84c407a11c11cf5de478968b3351e91 authored over 7 years ago

[feature] mask and g1x/g2x outputs

f099975a438481c8381c8cd57818c2313ffc88fb authored over 7 years ago

[fix] undo experimental commit

31e316e9323ef3fcb5818d344b94740fa72f5f12 authored over 7 years ago

[fix] add segment sampler to debug sampler

580a6ebaa43e6776abcde477be443595a432dd15 authored over 7 years ago

[fix] use base_sampler#sample

b4ee663ae97155e69c014201c99c783df9496148 authored over 7 years ago

[feature] multi-segment-optimized config

3acfa8791c1d71b2d292be16e5fbf2bf0300d23f authored over 7 years ago

[feature] multisegment generator

a381bfc1bdebd99b28696833abc84083bcec7d5b authored over 7 years ago

[feature] sampling adjustments to rely on mask

60f3292ff27f129d57542ba0e52b88b97a5be7e4 authored over 7 years ago

[feature] generator.config.channels overides gan.channels

75d4387fc5eaa94a369b3ea7f87d1e8b11343b79 authored over 7 years ago

[feature] update least squares configs

da014ad56a5d85712d68c2b8164e44c2d8cc1167 authored over 7 years ago

[feature] segment generator. applies a generated mask to two generators

(background/foreground)

7878576f690bcd6751bea15a9b122bd596a35ccd authored over 7 years ago

[feature] f-gan implementation

ee4d8f4848e50998139c294fe4f08e595a518563 authored over 7 years ago

Merge pull request #110 from 255BITS/source-comments-1

Source code comments 1

c74c965cc7523fcc3986760fdcd85358de19cc6b authored over 7 years ago

Merge branch 'master' into source-comments-1

0ea0ca9692767b5758840cc0b331d89f52785929 authored over 7 years ago

Merge pull request #132 from 255BITS/samples

Generate samples for each example config

1bc5a412a036c806d37b0495d49c9149c8939c33 authored over 7 years ago

[feature] update least squares configs

beb1787f6c53ad93a49e919861d8b589b9433290 authored over 7 years ago

[feature] separate out encoder loss config (eloss) and better descriptions in alphagan

a2060e7045b2cacf6c42ecac7801b2658a585e61 authored over 7 years ago

Merge pull request #148 from 255BITS/code-coverage

Code coverage

dc64ee7d05deb71042c70d247c51840aca09432c authored over 7 years ago

[fix] branch for deploying docs

7350466362514ca441e1022ad3e56aed86473c40 authored over 7 years ago

[fix] link to coverage

46506efe50ba27e628d60e54046828d0f971f40d authored over 7 years ago

[fix] link to api by versionw

a337b3eb559ab112e2b0e5a99ba1bdf303cc5845 authored over 7 years ago

[fix] use subdirectory instead of different bucket

241c34e65cb210930eefc33a17ea16231c9e6d74 authored over 7 years ago

[fix] install coverage plugin

70a6def118dcc2e13c0e55a4dc4d4bf32b990302 authored over 7 years ago

[fix] coverage location

731ebcbc122cfd987231d6572d76008592552baa authored over 7 years ago

[feature] coverage as artifact

aa1c1d297b6dcfea0b2af78c2766242f26a29464 authored over 7 years ago

[feature] upload code coverage to s3

b7a4d3590e26db09c3be6b30c46d7af64a96eec4 authored over 7 years ago

[feature] turn on code coverage

8252fa81232df22febdc61d7b752c8b79b4c2193 authored over 7 years ago

[fix] tests

f2e1d04efc6bf4e1a03b17826d21eb239cfa29ff authored over 7 years ago

[fix] syntax error

3d276558fad3eb236e3d2c68844f4ad56dba1445 authored over 7 years ago

Merge pull request #133 from 255BITS/gradient_locally_stable

Gradient descent GAN optimization is locally stable

45445ac9d84c2d8848ce9efd1baf775f8e09d2ce authored over 7 years ago

Merge pull request #147 from 255BITS/vral

VRAL loss

7a4060f8fcfa7a6e0bbea97055a25fa778763696 authored over 7 years ago

[feature] accomodate vral loss in pyramid discriminator / fully connected discriminator

a950bb7b2bb426f13d363a87a295734264c89495 authored over 7 years ago