HuwCampbell/grenade
1
1
Fork 0
mirror of https://github.com/HuwCampbell/grenade.git synced 2024-11-26 10:11:37 +03:00
Code Issues Projects Releases Wiki Activity
112 Commits 5 Branches 1 Tag 13 MiB
f9b53792fd
Commit Graph

112 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Huw Campbell
f9b53792fd Network scale autodiff test 2017-04-11 10:08:08 +10:00
Huw Campbell
ffbbbf222f Move more into the forAll for nicer diffs 2017-04-10 14:45:20 +10:00
Huw Campbell
56686db343 Slightly overkill close to operator 2017-04-10 12:55:05 +10:00
Huw Campbell
6c6e706e66 Use a blinded forAll for proxy like generators. 
This is a bit ironic, as the Typable constraint only exists for
7.10 printing in hedgehog, but the 7.10 proxies don't have typeable.
 2017-04-10 12:37:45 +10:00
Huw Campbell
adf731218c Docs 2017-04-10 12:07:02 +10:00
Huw Campbell
f4e1f2899b Add missing file 2017-04-10 12:07:02 +10:00
Huw Campbell
a40d3c7e20 Remove submodule handling from travis script 2017-04-10 11:10:20 +10:00
Huw Campbell
da810e6f4e Move tests to hedgehog. 2017-04-10 09:43:52 +10:00
Huw Campbell
8ebbea6a1c Change up submodules 2017-04-07 22:53:39 +10:00
Huw Campbell
f9c282bc35 Docs 2017-04-07 22:50:28 +10:00
Huw Campbell
235b76763a Add serialisation for Concat and Merge 2017-04-06 16:29:33 +10:00
Huw Campbell
10a6003e73 Add Deconvolution layer and update the GAN example to use it 
This cuts the runtime by about 70% which is nice, and it's a
better algorithm for it anyway.

I've also refactored the Convolution layer such that there's
only one actual implementation instead of two, and with that
provided a few more instances for 2D and 3D shapes in and out.

Update to the README and mnist show higher levels of composition.
 2017-04-06 13:34:43 +10:00
Huw Campbell
e6293b8461 Cleanup imports and move examples to new project 2017-03-20 17:20:33 +11:00
Huw Campbell
b855dd140d Update gan-mnist.hs 2017-02-08 22:06:21 +11:00
Huw Campbell
3e9549e186 Set theme jekyll-theme-minimal 2017-02-08 19:29:31 +11:00
Huw Campbell
c3f0373fbe Merge pull request #25 from HuwCampbell/topic/generative-adversarial 
Topic/generative adversarial
 2017-02-08 19:20:18 +11:00
Huw Campbell
6db2c40646 Add a generative adversarial network 2017-02-08 18:14:13 +11:00
Huw Campbell
60e28aad8e Add more reshape instances 2017-02-08 18:10:35 +11:00
Huw Campbell
ac9b43eaac Merge pull request #24 from HuwCampbell/topic/travis 
Sketch out travis
 2017-02-07 21:52:04 +11:00
Huw Campbell
1b2ebb4303 Everybody loves badges 2017-02-07 21:29:08 +11:00
Huw Campbell
d91d2cbf21 Sketch out travis 2017-02-07 20:58:24 +11:00
Huw Campbell
7632c5eb99 Merge pull request #23 from HuwCampbell/topic/recurrent-niceties 
Topic/recurrent niceties
 2017-02-07 18:04:13 +11:00
Huw Campbell
145aba4592 Make tests work in 7.10 and 8.0 2017-02-07 16:20:09 +11:00
Huw Campbell
bad6d0effc Fix input backprop 2017-02-07 13:24:12 +11:00
Huw Campbell
a153aff744 Make the recurrent api like core 2017-02-07 11:15:47 +11:00
Huw Campbell
7e47a6ad7e Merge pull request #22 from HuwCampbell/topic/tape 
Topic/tape
 2017-02-04 00:04:13 +11:00
Huw Campbell
c2d48a368c Add tests for Pad and Crop layers 2017-02-03 23:56:47 +11:00
Huw Campbell
df9f80a9cf Update Readme with new network type 2017-02-03 22:47:28 +11:00
Huw Campbell
c8cfa399e7 Split back propagation function into two. 
We now have one function which gives a Tape and
the output, and another which takes the Tape and
a loss gradient.

This has a great benefit, in that people can build
their own losses, which makes things like GANs
really easy.
 2017-02-03 22:44:31 +11:00
Huw Campbell
7d476bf0f6 Semigroup dep for optparse 2017-02-03 21:18:53 +11:00
Huw Campbell
1e461cb07a Make Pad 3D faster. 
Using dependent types in the deeper functions and
requiring a Proxy to reach them meant we required
dictionary passing to get the Nats. This made the
pad and crop layers almost 1000 times slower than
they should have been.
 2017-02-03 21:05:14 +11:00
Huw Campbell
dbeb962ae6 Fix many warnings 2017-02-03 14:43:07 +11:00
Huw Campbell
7af621cbb6 Add c 3d pad and crop functions 2017-02-02 20:27:02 +11:00
Huw Campbell
a23103ee7b Add extra composition 2017-02-01 22:13:20 +11:00
Huw Campbell
32f3e317e8 Change fusing layers so we can have two paths of sub networks 2017-02-01 14:17:18 +11:00
Huw Campbell
e6179ce5a4 Add recurrent tapes 2017-01-28 14:13:00 +11:00
Huw Campbell
d82a121797 Remove Show constraint on UpdateLayer 2017-01-24 13:47:47 +11:00
Huw Campbell
148c7778ab Simplify Network Type 2017-01-23 12:47:10 +11:00
Huw Campbell
6ad9803286 Start working on Tape 2017-01-21 12:46:40 +11:00
Huw Campbell
51ae147306 Add benchmark file 2017-01-20 18:53:18 +11:00
Huw Campbell
16573158d8 Nicer type 2017-01-19 23:08:01 +11:00
Huw Campbell
9989092216 Update Readme 2017-01-19 20:52:22 +11:00
Huw Campbell
df75beee8d Missing test file 2017-01-19 19:49:13 +11:00
Huw Campbell
2369ec3407 Merge pull request #15 from HuwCampbell/topic/cereal 
Add very basic model saving and loading
 2017-01-19 19:06:36 +11:00
Huw Campbell
7c4916409d Add command line parameters for shakespeare example 2017-01-19 19:04:15 +11:00
Huw Campbell
bcd1856988 Update LSTM 2017-01-19 18:56:22 +11:00
Huw Campbell
48263cd9e2 Update README.md 
Add recurrent
 2017-01-17 19:47:39 +11:00
Huw Campbell
4dc408f39d Add very basic model saving and loading 2017-01-16 10:16:40 +11:00
Huw Campbell
20e7e483d7 Fix 7.10 builds 2017-01-16 00:29:25 +11:00
Huw Campbell
cc70f1fc0c Merge pull request #18 from HuwCampbell/topic/recurrent 
Recurrent nets and other changes
 2017-01-14 21:16:48 +11:00