Commit Graph

4147 Commits

Author SHA1 Message Date
George Hotz
ef4afdb5d2 tests maybe 2022-07-18 08:24:14 -07:00
George Hotz
a2c4bcf313 disable opencl tests 2022-07-18 08:17:21 -07:00
George Hotz
5093455166 don't shuffle if there's children involved 2022-07-17 21:15:40 -07:00
George Hotz
f76d41812b prune graph 2022-07-17 15:38:43 -07:00
George Hotz
eda6f071b2 default opt level 2 2022-07-17 14:54:40 -07:00
George Hotz
73b0471b25 join expands 2022-07-17 13:42:05 -07:00
George Hotz
cfabbbd6bb more crap to remove without convs 2022-07-17 13:02:27 -07:00
George Hotz
5e96ed523a fix opencl bug, no training on opencl 2022-07-17 12:55:26 -07:00
George Hotz
f93e297804 fix bug caused by rounding 2022-07-17 12:49:58 -07:00
George Hotz
cff297ef9d w/e, that's a later prob 2022-07-17 12:32:50 -07:00
George Hotz
4bc07326d4 we need that opt to make gpu decent speed 2022-07-17 12:26:18 -07:00
George Hotz
6375e7129a opencl not imported 2022-07-17 12:14:39 -07:00
George Hotz
bf299802f8 fixup tests 2022-07-17 12:11:53 -07:00
George Hotz
762e859089 testopencl 2022-07-17 11:56:40 -07:00
George Hotz
608e2431f7 test opencl, commit to removing the crap conv code from GPU 2022-07-17 11:55:37 -07:00
George Hotz
3c4565fa21 SLICE -> PAD,SHRINK 2022-07-17 11:33:59 -07:00
George Hotz
9574dd8559 some permutes are reshapes 2022-07-17 10:34:24 -07:00
George Hotz
77806e0d64 fix permute stacking 2022-07-17 10:24:57 -07:00
George Hotz
c28a99087b add PAD movementop 2022-07-17 10:22:03 -07:00
George Hotz
4527c0453d get_movementroot 2022-07-17 09:37:27 -07:00
George Hotz
b00cc93102 bugfixes 2022-07-17 09:19:22 -07:00
George Hotz
d07f379038 don't merge movement ops 2022-07-17 09:09:50 -07:00
George Hotz
f6ea7c022a Revert "EXPAND -> REPEAT"
This reverts commit 115d2eadf5.
2022-07-17 08:42:10 -07:00
George Hotz
cca089b11d Revert "more expand -> repeat"
This reverts commit 2e7b1630a8.
2022-07-17 08:41:48 -07:00
George Hotz
2e7b1630a8 more expand -> repeat 2022-07-17 08:40:49 -07:00
George Hotz
115d2eadf5 EXPAND -> REPEAT 2022-07-17 08:38:54 -07:00
George Hotz
1eb247f823 contiguous and same length 2022-07-16 08:49:07 -07:00
George Hotz
d04b274cd2 noop removal can replace with reshape 2022-07-16 08:32:42 -07:00
George Hotz
d985217fa4 skip reduce noops 2022-07-16 07:47:43 -07:00
George Hotz
bcf422dfdd Device2 (#358)
* option for matmul

* fixups

* fast like a nascar

* running

* thneed runner

* no buffer id makes no backing buffer

* move constant folding to the top

* runs on mac

* folded biases

* was v slow

* maybe just that

* elu touchup

* speed and float32

Co-authored-by: Comma Device <device@comma.ai>
2022-07-16 07:26:19 -07:00
George Hotz
d651caa864 fixup openpilot/compile.py 2022-07-11 13:59:09 -07:00
George Hotz
5e46561f7e no_grad = NOT backward 2022-07-10 20:54:57 -07:00
George Hotz
b34ae7876f lol chr(10) not chr(13) 2022-07-10 20:03:11 -07:00
George Hotz
817b64f5e5 A conv is a reduce op (#356)
* universal strided conv

* more correct

* hmm, CPU works

* cleaner cl code output

* make noconv a flag

* cleanup __getitem__

* refactor broadcasting

* put that back

* unneeded reshape in getitem

* fix strided for torch
2022-07-10 19:58:50 -07:00
George Hotz
057e4f5aa5 a little faster and cleaner 2022-07-09 08:14:01 -07:00
George Hotz
c77ba7fa3f ops_cpu readibility 2022-07-09 07:48:54 -07:00
George Hotz
0a36475700 no einsum for now 2022-07-09 00:04:40 -07:00
George Hotz
c39a245696 that's not where i thought we'd lose lines... 2022-07-08 23:52:38 -07:00
George Hotz
75e1848b09 always SHUFFLE_RESHAPE_OPS 2022-07-08 23:19:39 -07:00
George Hotz
44848ee5dc prints show we can precompute from the outside 2022-07-08 10:59:20 -07:00
George Hotz
68959be05d precompute weights for opencl 2022-07-08 10:56:48 -07:00
George Hotz
d8e7f1f8bc opencl type ignore 2022-07-08 10:33:55 -07:00
George Hotz
ae335b6d3e opencl works, but tons of kernels 2022-07-08 10:22:04 -07:00
George Hotz
5b66d1bb0b begin fixing up opencl 2022-07-08 10:20:14 -07:00
George Hotz
7e17f2ae8d fix mypy, add TODOs 2022-07-08 09:57:22 -07:00
George Hotz
8557ed88df use ast engine for merged reduceop 2022-07-08 09:37:40 -07:00
George Hotz
3656a5615a MERGE_ELEMENTWISE_INTO_REDUCE 2022-07-08 09:32:28 -07:00
George Hotz
ca9532ce29 less lines, and typing found a bug 2022-07-08 08:57:12 -07:00
George Hotz
2035b89e54 wooo 1k lines 2022-07-08 08:44:57 -07:00
George Hotz
2a8c1071d9 cleanups 2022-07-08 08:36:31 -07:00