Commit Graph

11106 Commits

Author SHA1 Message Date
George Hotz
ecc0903451 Revert "faster deepwalk"
This reverts commit 931500a098.
2022-01-15 21:02:18 -08:00
George Hotz
931500a098 faster deepwalk 2022-01-15 20:57:57 -08:00
cjg91
7025c9bbeb Transfer learning for ResNet (#295)
* Transfer learning for ResNet

* moved ResNet depth specifics into the class
2022-01-15 23:22:10 -05:00
George Hotz
55d792b065 Revert "fixup resnet"
This reverts commit 4eabe677ed.
2022-01-15 20:22:01 -08:00
George Hotz
4eabe677ed fixup resnet 2022-01-15 20:21:02 -08:00
George Hotz
e0bef0bd01 training is False by default 2022-01-15 19:57:41 -08:00
George Hotz
8ba3d1f803 fix bn test, affine is True 2022-01-15 19:52:15 -08:00
George Hotz
8ec2341cca fix bn training 2022-01-15 19:47:01 -08:00
George Hotz
0e6832a8ea support torch GPU, only autoinit cuda in the buffer 2022-01-15 19:15:12 -08:00
George Hotz
d8d19ed468 wikimedia wasn't returning 200 2022-01-15 19:09:29 -08:00
George Hotz
52918fbf78 cuda stub 2022-01-15 19:02:17 -08:00
Jacky Lee
81664baf64 Fix OpenCL installation (#301) 2022-01-06 10:35:48 -05:00
George Hotz
b0511f9392 Revert "does pybind fix CI?"
This reverts commit d128e4fcae.
2021-12-30 14:13:58 -05:00
George Hotz
d128e4fcae does pybind fix CI? 2021-12-30 14:11:39 -05:00
George Hotz
5efb6653c4 a bit of work on metal 2021-12-30 13:53:08 -05:00
George Hotz
c0c2c0b041 support larger ViT models 2021-12-12 10:45:10 -08:00
George Hotz
785fe8ead7 unify torch Slice 2021-11-30 16:25:37 -05:00
George Hotz
e59381d0da cleanups, remove np 2021-11-30 16:22:00 -05:00
George Hotz
e28cdfb0cf clean up resnet 2021-11-30 16:14:54 -05:00
George Hotz
8f5779eeaa very minor change 2021-11-30 15:54:03 -05:00
George Hotz
d31ef0ae48 make vit names match pytorch 2021-11-30 11:34:14 -05:00
George Hotz
4b7c31b5b7 break vit into it's own file 2021-11-30 11:19:22 -05:00
George Hotz
46bbbcf7f0 model touchups 2021-11-30 11:13:34 -05:00
George Hotz
7d7e2b690d clean up Conv2d 2021-11-30 11:02:55 -05:00
George Hotz
835869974c clean up vit code 2021-11-30 10:58:03 -05:00
George Hotz
6884add850 save two lines 2021-11-30 01:16:48 -05:00
George Hotz
9b538629bb cosmetic 2021-11-30 01:01:39 -05:00
George Hotz
38dccb3a2e same simpler sum and max for gpu 2021-11-30 00:59:05 -05:00
George Hotz
5d60df2b10 simpler sum and max 2021-11-30 00:53:27 -05:00
George Hotz
c39824bc62 oops, forgot some stars 2021-11-30 00:46:14 -05:00
George Hotz
908db3bdea support bias in conv like linear 2021-11-30 00:44:59 -05:00
George Hotz
bd21304e3c linear takes in weight and bias 2021-11-30 00:38:47 -05:00
George Hotz
535f02cc64 use sequential 2021-11-30 00:25:39 -05:00
George Hotz
de938c2d9d vit is now tested 2021-11-30 00:23:06 -05:00
George Hotz
aff810e722 unify transformer block 2021-11-29 18:58:15 -05:00
George Hotz
58ed46963e fix broadcastdot 2021-11-29 18:54:57 -05:00
George Hotz
033b04494a resnet pretrained is broken 2021-11-29 18:13:52 -05:00
George Hotz
125e74293f promote layernorm to tensor op 2021-11-29 18:08:21 -05:00
George Hotz
dca076dbf1 remove dumb nn ops 2021-11-29 18:05:31 -05:00
George Hotz
33720e733f support keepdim 2021-11-29 17:47:43 -05:00
George Hotz
8a02bd56a1 refactor: canonicalize axis 2021-11-29 17:29:18 -05:00
George Hotz
70544e7e9f sum hook override 2021-11-29 17:14:24 -05:00
George Hotz
8097b8f7d6 vit works 2021-11-29 16:28:14 -05:00
George Hotz
7c07c5efdd plz fix vit 2021-11-29 15:45:19 -05:00
George Hotz
ca160504e1 affine is always the last dim 2021-11-29 15:22:49 -05:00
George Hotz
e86f7a4aa3 deterministic 2021-11-29 15:10:15 -05:00
George Hotz
f909ab194f gelu with broken test 2021-11-29 15:00:50 -05:00
George Hotz
9ce881f88c fix bug in getitem, drop int axis 2021-11-29 14:01:24 -05:00
George Hotz
c752033283 fix GPU OOM in test 2021-11-29 13:05:59 -05:00
George Hotz
1eafa5580e layernorm with learnable parameters 2021-11-29 13:03:57 -05:00