Ecosyste.ms: OpenCollective

An open API service for software projects hosted on Open Collective.

github.com/HyperGAN/HyperGAN

Composable GAN framework with api and user interface
https://github.com/HyperGAN/HyperGAN

[fix] mark broken gang loss

b8adf4f0dadad4be9f86c3e517307b0f6a79b0cc authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] grid sampler without style

9a0f4ca9354a9bba9709cf4bf68021e99be78dc1 authored over 6 years ago by Mikkel Garcia <[email protected]>
[experiment] new gang objectives

85ec99a6dde50da9576344974bb28be8b901c7cc authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] detect stuck state

2273f4bcec3c0163a8373153d30393bae3113e7d authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] experimental messy gang loss terms

ccb184a6cce22d02f50826bc1fa8a807a29e5a84 authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] typo

696261a3dc517f2f0fb6ea660b84da995fb50fd1 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] experimental gang loss for least squares game

d51d18e895eb861f60de20a397258cfdc8f38521 authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] sampling gang next frame

684bd358e86178ad014691fa859df69e675b1f60 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] speed up gang tests by not using fitness and not locking z

a1acf506862f847a244b9cc07fb1f32505c81003 authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] gang with next frame

4e2ab861f1f60f97c837b2e0965694cc410b87ac authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] incorrect variables for d optimizer

ee4732b17c17b43121bb367acab11366b8179e80 authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] catch error on nash training

86437cba8555a49688589cb678798f7866f6ffce authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] experimental gang losses and recognize fitness failure state

1ae716dc9f1bf3102869bc79fb4e75f876bb5485 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] recognize failure states

50d28fb71b2af88aa78b1d2710a08636a8467011 authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] decay of min_fitness auction

c3e42b85a6b8a340e0ea5763b10504adf4213a06 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] rgan least squares formulation

e823774b9cb4532dd0f5e9e06612c8f0824fa8cd authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] train_d_on_sds and [fix] detect degenerate games

2676a5a6155ecdfa4669d5bd258f2d3a93af8dac authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] single step update rule

56e03b595efa7da44fd2c6d6bf51e1b865229289 authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] soft failure on degenerate games / fitness

da72c666afc3c7c7d5f4de5513f1b8f78dd5814e authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] gang criteria and fitness end

e093cb95e030349bd187033c7d647ad07068f773 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] ragan

0260ab27252e1d6a81409e5eb8c00c84ebd3c98c authored over 6 years ago by Martyn Garcia <[email protected]>
[fix] gang trainer subgame

2ac7d9b126f27deb84d58f10682a2d394d25beea authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] swap nash results

0e2a72a4bbb0a28d91cb47750f771541d204e280 authored over 6 years ago by Martyn Garcia <[email protected]>
[refactor] assign_gd split into g/d

812f417992e62a044b3fb539eb760f91317e2eb3 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] gang sampler

6cc40481a3177d29134ca18352c4b9c940ec8153 authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] remove worst candidate, not best

9cdfa1e331e9eb0bbf94ce77ed2ba1079a0fa77e authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] various examples with new trainer

5d27031f85ff0bb1da67e23fc265adee52c54c0c authored over 6 years ago by Martyn Garcia <[email protected]>
[fix] use gan.fitness_inputs to be explicit

9fe97bcac4312b05ef2d50eef4b17404a2408dab authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] variable leak

c4006e80f3447650836e29dfff118a22af24ddc2 authored over 6 years ago by Martyn Garcia <[email protected]>
[fix] resist nan failures

cc3b682177be465328eda63ea86ef76449a6c967 authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] fitness_inputs defined for standard gan

2074bde046b4cc785e7844c7a149d9b3a09902f4 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] lemke_howson

8cc6c4c7380e941b0663d4655e8cca4c44f04af5 authored over 6 years ago by Martyn Garcia <[email protected]>
[fix] disable mixture of slot vars for now

b7b202bc64a8c8b4f88e4e1518dbd49f7d9d9fe5 authored over 6 years ago by Martyn Garcia <[email protected]>
[fix] fitness inputs on standard gan

f5c84335863ead72f216b08bddda761f8ce63587 authored over 6 years ago by Martyn Garcia <[email protected]>
[fix] lock fitness variables

3d50b8c2a7363c141ddca0e89a4f55e0cc47473f authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] gang trainer resource usage on larger datasets

7c58fd726cac0261298999519a8a64339f5dd249 authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] next-frame example on gang

6e1cc564aa845b536e2cb9c534f38eb1a2af5c2b authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] next_frame with gang part 1

516b25988f9970ddb08b1034e37961753c681d4f authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] include slot vars from optimizer

34c44402eef7d1dd2f1affbc5d01d8005cd8f985 authored over 6 years ago by Mikkel Garcia <[email protected]>
[experiment] different nash solver

4f48cefdd6e16cd0e864c1de9b8692cdec66f435 authored over 6 years ago by Martyn Garcia <[email protected]>
[optimize] one game per nash

206c271f38dcbbc88b63be4abcd32887811e1619 authored over 6 years ago by Martyn Garcia <[email protected]>
[fix] object namespace

1dc98f8666485503dbd442d746a3d3c684a9baf9 authored over 6 years ago by Martyn Garcia <[email protected]>
[experiment] nash solver

6e4764d2ab98ed21a84f0c698be462183ea2ceb3 authored over 6 years ago by Martyn Garcia <[email protected]>
[experiment] gang variant

2df6af182a8fcc7770056b7e77aad6402cbc60d1 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] different evolution strategy(mixup)

f471c019b2a72d4503c4bc4216bd90501accfccf authored over 6 years ago by Martyn Garcia <[email protected]>
[fix] gang with next frame

0ae55121a5ca57493e0b8268ac57bf3b4eac905a authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] gang trainer

9ad57f08ab29dcded46167d12eca85fe5d9b0ee8 authored over 6 years ago by Mikkel Garcia <[email protected]>
[experiment] separate next frame inputs

d5047a5445cfa781e1eb2752bc85e4e06ff45a3b authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] next-frame-slot / fitness interaction

7a896d7ae1ba6d0c492bc0804087bc6482820485 authored over 6 years ago by Mikkel Garcia <[email protected]>
[experiment] next frame through fill in the blank

e7e9d766de8f9b28d21288eff4ad9ee9c96f05e6 authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] set d_fake and d_real on the loss object

21b5c19ab9ba7d2e10edfbb53b161e687e7a2e6a authored over 6 years ago by Martyn Garcia <[email protected]>
[fix] fitness trainer name

42511b682394e61fadfd7f9b7c1a4e62749ff7fa authored over 6 years ago by Mikkel Garcia <[email protected]>
[experiment] fitness trainer trains only if g passes fitness test

c5b491f180afbe130863df6aa66b762437403039 authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] evolution trainer multiple parents

1c51cecde6aba5e6aa61158dd441326da5bb79a4 authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] evolution gan

895af1d64c9cd13d5ddfea120ba642e187de9674 authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] another attempt at exponential moving average

7e6824457efbd86add6701a055dffbf77b5e65b8 authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] moving average

57c07ec7b0ed64ee9562b1b3eb74c181dbf15ceb authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] exponential moving average on g

b83c44153b60eea180e503c2e98ddef06fa9311d authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] discriminator selection

0afdbc4a6d350b3145f6bcc3dfe22e38f224a40f authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] training methods for kbeam trainer

f633f14e56c45a5027201de9a31d995b44b7c147 authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] some configuration options

4500ba96c31ac7ca4268a589348c1ac4e439be0b authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] stop gradient from training g through unselected gradients

5c945b726427222d91deea843f63120b70c81d61 authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] kbeam output on which d is used

a04a4494fd68484dcefb98513661e0ef22cb2fa7 authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] multiheaded attention

93faa927492415e2dc8ddf706d3898540fbd7807 authored over 6 years ago by Martyn Garcia <[email protected]>
Update setup.py

0f6ba812bd6e23531e58a0da7131afec2306cdc9 authored over 6 years ago by lukas2005 <[email protected]>
[fix] variable case

a176332adee710b5c4f4bb5813e2a880f7e6134f authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] experiments with two samples and attention

9abed17a96166eff69812cbaeab77a2452410b6a authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] use same data in kbeam trainer

a3829206e77c8caf73d4a1e202fa469db5830693 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] enable_at_step option for attention layer

5073316a0ef6b3d238b0e5ada7bf7b31f3ee888c authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] use g-loss as per the paper

1c990f899377d5f92fd2bdc2f9a2882aa5275a87 authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] various next frame changes

94b9cfb7f11568f2f999736efee4499ee5b89b82 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] kbeam search

37b3d5091734a36a64b6536e0eb58f31319cafa9 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] some more next frame experiments

4065b4d27a7ff9d0eac989a83829467461e99d49 authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] new experiment with scheduled training steps(not working)

d96cc56a35574d9858011d887dc15fb886f1f3ea authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] spectral norm

ce13fe712ff9599529b18e0a9ea25557ccc4d6d7 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] experimental config search options

67cadcccf927c93dec42cc9d0c6885d47bbc9eff authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] dot product

30d4d07a8eb3189817426b18c8e06f29b4dbfe19 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] attention layer

373960a31000bef3d32b91538080efe43408f3db authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] relational layer

d2bc8ad068f4a9a696ad413a2f4ec7e1996712f0 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] various non-working experiments

6e306273f72f92cca7b6e4da4215a99cc8c01cd2 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] experimental feature g_random

500282af8cbeb49d00acc9f5825953940457103f authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] --shuffle

6472f56cd2e113fb27f05c7b31b567159910f90f authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] same_g without style

dc1c2003e12228ee945e21cf1dec6bcb970dd36b authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] sampling (style not explicitly set)

7a170ed030b599d21b9bfdc9655bf4fa71faab25 authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] support for shared generator

8372d9ccba9e1786a95435d08a5c2e69226bb940 authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] --zoom, sampling and experiment

6b047e6be891e74e9828f395abb0ef1fd22d0326 authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] config.style = false

0d4d0eb377a35f21f876c46de4124aaac0a7682f authored over 6 years ago by Martyn Garcia <[email protected]>
[fix] alice loss term

885535b6909214d28944464eecec9c398a76b62f authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] new next-frame experiment

680343e99285bd1f6d0d73ff32ef8e30c4b17cfb authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] experimental loss terms. very messy

13726a56078df896a2b156e80f004115a635e5cd authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] some experimental flags

274e49b56b4ac335681b86bfd02b88c28743dfcd authored over 6 years ago by Mikkel Garcia <[email protected]>
[fix] some example options

ba7e00f6054594dc9965fe7fabd30a76a927aa02 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] messy experimental flags(wip)

9d09bfb733d212302a9715222bb2ed86b95c4b54 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] sampler for experiment

5a6945452a5b4f04e259174087d5bfe343ab4737 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] experiment

b8059ecc7e8557cf1173241db287cc9d7666808d authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] initial conditional gan from https://research.fb.com/wp-content/uploads/2017/04/neural_tests.pdf

6c5d2b98a56e4ed78d2a6e3aa0f59d9fec60dddd authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] more aligan options and some refactoring

54b04d8f639ba9fe5b1b74e50d50d0e2447814ef authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] ali style gan experiment

cbde6413e00bab545b698971bac2bbcceac82c60 authored over 6 years ago by Martyn Garcia <[email protected]>
[feature] move adaptive instance norm feature(kinda hacked for now)

3785cd1debdb4b2befd3185f3c56b7e4967943a8 authored over 6 years ago by Mikkel Garcia <[email protected]>
[feature] style walk sampler

969501390214bd9af875a5f197c84577d6f7da12 authored over 6 years ago by Mikkel Garcia <[email protected]>