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

Merge branch 'master' of github.com:255BITS/hyperchamber-gan

* 'master' of github.com:255BITS/hyperchamber-gan:
[feature] better cli
[feature] better cli...

aaf25a72bef5f4ca9208827ace8b0869d639ab7a authored about 8 years ago by Mikkel Garcia <[email protected]>

[doc] setup.py

3c8c3448b41d923075cb9930d05e1332041f87f0 authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] better cli

6ec154328fef5d8d6a2d001c989608da8fbe98b3 authored about 8 years ago by Martyn Garcia <[email protected]>

[feature] better cli docs

a8fdb9d5e94579d18e05dde881b0f451437bfb54 authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] hypergan executable (pip package)

c5833711612544e75b4e17f4227b87b9ed00813d authored about 8 years ago by Martyn Garcia <[email protected]>

[feature] --use_hc_io defaults to false, support local save/load

954516e9bfd4ee1cbe0f4103d5e24a90722667f1 authored about 8 years ago by Martyn Garcia <[email protected]>

[feature] better defaults

c47974b25dfc347af722b9c20d694797ddfa5f9f authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] ability to remove batch normalization

e5bb4656c8397c14ef1845e7596ea2527471d769 authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] new type of generator that has an extra layer before output

44ffd1af9ac753112736f61d8e1028df9108f13e authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] use layer norm instead of batch norm to prevent manifold from

breaking

0e188733b3fc7eeace3f85ef18fc40ba1e10745a authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] content size on gif

36df1fd877890b9e4450c9b949b4e9102ae1d4d8 authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] youtube link

d8008d6b705a6191d79e1385b7ae111402341a10 authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] hypergan visuals

cbfddbb4bf5b96066503e2a1141a81e40dd22e5e authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] frame sampler, grid sampler

725d3851f3a50b89d3467e12b3b26ffbc2897016 authored about 8 years ago by Mikkel Garcia <[email protected]>

Merge pull request #2 from 255BITS/prelu-2

A lot of changes that lead to a better overall result. Documentation update coming in a differe...

3afda23cdf01ac336830e04e01f2878596db2703 authored about 8 years ago by martyn <[email protected]>

Merge branch 'master' into prelu-2

594ac04648a892a6d9b598a63a81165960e40f5a authored about 8 years ago by martyn <[email protected]>

[fix] remove unneccessary reshape

4588de6c9e20c81fd0054cdcd66adafa53771117 authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] unsupervised learning mode

25fd7bdbd9de94c15a182835031046baf21c62f2 authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] cheap discriminator. doesn't work very well but useful for debugging maybe?

2414384d16708fd06219b336a04dfaff5d7fe336 authored about 8 years ago by Mikkel Garcia <[email protected]>

[refactor] support both supervised or unsupervised(sigmoid/softmax) output

4c60119647c7c5347a6d31265c223137c7ddd896 authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] disable categories

02913c38f8e4ca5393dfe8ca6872f598b036c789 authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] readd dropout and set random combo encoder to old encoder

2a3b31417b4c43cffcaf5d387ce07b353c53982b authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] batch size in prelu

e77c39b7fa27259dca180743fa826bf2069d6294 authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] more defaults and switch g to prelu

ed16f5415f1097f0577211dec5d01fca2e57b769 authored about 8 years ago by Martyn Garcia <[email protected]>

[feature] random gaus encoder and better defaults

19835c41666aa3049c73f3e2a9f2d410d7094b66 authored about 8 years ago by Martyn Garcia <[email protected]>

[feature] requirements.txt

c2fa21e65d8f24b2a1e22f2bf07d68d67b46b714 authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] remove linear layers at the end of pyramid_nostride. If we want

that lets use a different D

a342681e9c44ffe4ab507e4cf9e678ee7cfa71d0 authored about 8 years ago by Mikkel Garcia <[email protected]>

[tweak] 24 batch size and lower stddev on last linear layer(more inputs)

3c915d110392bcfd3ff2a996ef77dcd056b2c245 authored about 8 years ago by Mikkel Garcia <[email protected]>

[experiment] adam trainer

80bb36e42a79418fd3ca888e293841812f0312b4 authored about 8 years ago by Mikkel Garcia <[email protected]>

[tweak] adam trainer

5bd71655599f4365e68ebd0f2c27dbebde1adfa3 authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] slowdown trainer

090c8ddfb09c644208cbbfe27a4c43aed58b1c83 authored about 8 years ago by Mikkel Garcia <[email protected]>

[experiment] a bunch of changes for visualization

339fc87d3cf10b349345e598c84d6fee45232852 authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] moments on d

b0c4f30b4e71ba3f5c9a21cdc822a31c8dd86e38 authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] random combo encoder

b2a13ed55b7bc80e234eab5d479074e9b18f420b authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] video creation

91157b86cb666002daf979a5603fdb2ed57df80e authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] disable encoder with prelu

b95203f6fb435efe9996c5074dcd339fd1e025de authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] dropout after activation

0cac8a127bff58650802bd507f90c55b653587a6 authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] activations = prelu

3fb2a0866c30c44bc291706d29f3fb88df72a682 authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] savefile

72328a7fe6ff9f9986378cd795f1d3a58498e582 authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] 2d on z

0302791b269078754cd9c8c7b5c49f3ef3ef36de authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] prelu

e3c5fd72282dd39bb41b848a811f60bd4f979659 authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] build / load

58fffafabd98d8313d601063cc70f5436db1398d authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] make g use layer norm

55cf8b23195afc99972dc6326b102ae01ab04c4f authored about 8 years ago by Mikkel Garcia <[email protected]>

[experiment] some batch norm some layer norm

504b4fe5117516d6d1d211e881f430cd44f59c63 authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] layer norm

dd6d0b996a4346a24cb2c9ba3c9b9d417214e6cc authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] tensorflow upgrade to 0.11

69f04e7a845d4058c7e3d17eea32b9979dd08e7b authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] abstract out layer normalization technique

1f579d56c89972b1fef00e0691e00cc23450cf7e authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] pyramid w/ stride

1aa6f71331656a09a2cc5a6f85a584ee362cedae authored about 8 years ago by Martyn Garcia <[email protected]>

[feature] z -> linear -> batchnorm -> activation on g becomes z -> linear(l2 loss) -> activation

2360718e02409ffc20f21f975a04b107c8d4032a authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] variational

352f03f4e2f7b3aa71d991de00cde549a89a4c60 authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] new default rmsprop trainer

a5eaecd2af629bf6363f446b97cbc56229c664bc authored about 8 years ago by Mikkel Garcia <[email protected]>

Merge branch 'master' of github.com:255BITS/hyperchamber-gan

* 'master' of github.com:255BITS/hyperchamber-gan:
[fix] add missing config variable
[fix] u...

e96788a8a4a620fa9dc58b63bd77b15e9d1b3aab authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] rmsprop trainer

fda529612cb27fd1d45e8fe5fd2a7fd2b1e84426 authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] add missing config variable

019bbe0c150bc94daca91dcb8e53adc7ff88051f authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] use encoder on generator part of graph

7d3699613588698eefd5994687f05ef7059477cd authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] build and serve commands working again

6bb70ce00fbaf2cb76e0ca136cc1b75a11b7cf31 authored about 8 years ago by Mikkel Garcia <[email protected]>

[refactor] cli arguments

cfaf04ac1e484320891420078ebdbbb349fc976b authored about 8 years ago by Mikkel Garcia <[email protected]>

Merge branch 'develop'

* develop:
[feature] dropout in the discriminator, best samples yet https://hyperchamber.255bi...

c35cec64757e6a549938eeafde63e7b835a6e7d1 authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] dropout in the discriminator, best samples yet https://hyperchamber.255bits.com/f19a3ad8bda75af772296edadb12f87d/faces:1.0/samples/b60116ef6f108f210225f1f57d46053d

63920b769fe73bf41185419db470b7839dd9c8be authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] docs

d6be31782e897b52c058e5c17545c97dca099154 authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] add -c

9d1e21408083a48b7a8b416598746e02322bb536 authored about 8 years ago by Martyn Garcia <[email protected]>

[feature] better cli

0a39260d3424d1ebb165ff9800cf43054fd3e4f8 authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] encoder

90ee38a6b45d0f0a5f79a27713c5085b92520f6b authored about 8 years ago by Martyn Garcia <[email protected]>

[refactor] cleanup more options

7a6abbdfc033d90952deb7dbd1e925e8e4c255df authored about 8 years ago by Martyn Garcia <[email protected]>

[refactor] s/z_dim/generator.z

899b9f9b97fc911b92a9c9cd87b4ae387dbb86fe authored about 8 years ago by Martyn Garcia <[email protected]>

[experiment] new type of discriminator termination, more noise regularization on the pyramid layers, and remove categories(for now)

322f71f0f8c06a738c47cf825f8726ee5f1f496a authored about 8 years ago by Martyn Garcia <[email protected]>

[wip] try linear layer in the pyramid_nostride_discriminator

63b0b620b884bf7e59faece58e9b1453c6e6c7f7 authored about 8 years ago by Mikkel Garcia <[email protected]>

[tweak] better defaults for d

f5ad8c8b92d94730739460ec0596aa20ad5674cf authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] no stride discriminator

de8494a59edc4f67325194d2b4ac1f957dbe3ce6 authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] bigger defaults and disable D dropout

7754b34dae35e737476202a01637db6fa4a44d00 authored about 8 years ago by Mikkel Garcia <[email protected]>

[feature] added painters loosely based on https://github.com/inejc/painters

30e509f3c35ee2ab761b5a9f72800388e1b72e0b authored about 8 years ago by Mikkel Garcia <[email protected]>

[wip] try a basic discriminator for sanity test

c6871dcf5be89d9ca7d0ff4720091cd573c4139e authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] initial g now is linear -> conv -> resize instead of linear ->

resize

f82d1b711a6437fa776c9300dc713f87ce2a653c authored about 8 years ago by Mikkel Garcia <[email protected]>

[wip] start z at 4,4

9c2d07f6bed26d3457e0d9a9b527f6746d6cb953 authored about 8 years ago by Martyn Garcia <[email protected]>

[wip] stabilize g

94f6cb166b7167c68e7096303150918c2eea2460 authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] remove erronous batch norm

cb94a9598faaa821c8f83a226c790e3fd179dc7c authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] use larger g

839d2c678274dd85505e88462525ccc916d51fe0 authored about 8 years ago by Mikkel Garcia <[email protected]>

[fix] oversized noise addition

8548bb59a89b8bf3d4c0e8153cb059e9524b5bf1 authored about 8 years ago by Martyn Garcia <[email protected]>

[feature] faster densenet

69f7ab7ca17a574ffc20df94bdb4060188f9666a authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] linear on g

e4ee8e989f0ee8138798c290233c1c3ec7d3fb61 authored about 8 years ago by Martyn Garcia <[email protected]>

[feature] densenet

a02527e58d09b7f86575f5b790db1fa9dbc8e4e6 authored about 8 years ago by Martyn Garcia <[email protected]>

[feature] larger linear layer

7febb610e1ddeb8517e633c6d256871ee4253baf authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] add missing file

de8dfe0db0a6f1f0708761864c62563af14ef5e7 authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] server mode

becb7d6e14df4d2dcc162383cf41d4bb9dfca921 authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] bugfix for primes == 1

748e0265e8ddfab9a602bb5ed65aec89caf53b93 authored about 8 years ago by Martyn Garcia <[email protected]>

[experiment] attempt at a conv based g/d combo

8bc19b0795e1bbc8cad009b5edb169abc3eb4689 authored about 8 years ago by Martyn Garcia <[email protected]>

[optimize] add avg pool layer back in to try to get rid of lines

7741717bf0df57774878d7b891ea5a49ccf23d3f authored about 8 years ago by Martyn Garcia <[email protected]>

[experiment] attempt to remove g linear layer

5258284d84ede36eb6d0639ecb8517ee6927dda4 authored about 8 years ago by Martyn Garcia <[email protected]>

[wip] moment snippet

aed91916e39567719105995faabe89e574778464 authored about 8 years ago by Mikkel Garcia <[email protected]>

[tweak] stronger default g

841c984d4daccc73c509a6131ef74fcb0319d685 authored about 8 years ago by martyn <[email protected]>

[fix] initial values and conv before avg_pool

26b58c84c76052eb5bbc5863147c2724f1908876 authored about 8 years ago by martyn <[email protected]>

[fix] add missing file

a9d966c4d65b9e97ba9bd3b1966a6256b95626c4 authored about 8 years ago by Martyn Garcia <[email protected]>

[hack] crazy experiment

ae19b696bcea0049c2502d2572ee58f9f2be2c5f authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] stddev on last linear layer

ed2bf5a3cad8a4668694c62f2dc5b21606233f05 authored about 8 years ago by Martyn Garcia <[email protected]>

[feature] use smaller conv in the graph

1172110a815dc46239272a4a79f71ca2536de40b authored about 8 years ago by Martyn Garcia <[email protected]>

Merge branch 'develop' of github.com:255BITS/hyperchamber-gan into develop

f98e76b95ef2c3fe5824586f5313777cb165365d authored about 8 years ago by martyn <[email protected]>

[feature] progressive_enhancement_minibatch_regularizer

8050560cf1054845736a8508f07fc5d968560134 authored about 8 years ago by martyn <[email protected]>

[feature] remove discriminator.fc_layer

61901b80f609ec0d833db2bd6f0b235ce6318722 authored about 8 years ago by Martyn Garcia <[email protected]>

[fix] empty array issue when doing function lookup

1012cd94bdb434d3c2297ac0d092b169aeeffedf authored about 8 years ago by Martyn Garcia <[email protected]>

[documentation] improve README

e0673a92e87a659b6f11d84efdb5f1327e908b49 authored about 8 years ago by Martyn Garcia <[email protected]>