geohot
cfabbbd6bb
more crap to remove without convs
2022-07-17 13:02:27 -07:00
geohot
5e96ed523a
fix opencl bug, no training on opencl
2022-07-17 12:55:26 -07:00
geohot
f93e297804
fix bug caused by rounding
2022-07-17 12:49:58 -07:00
geohot
cff297ef9d
w/e, that's a later prob
2022-07-17 12:32:50 -07:00
geohot
4bc07326d4
we need that opt to make gpu decent speed
2022-07-17 12:26:18 -07:00
geohot
6375e7129a
opencl not imported
2022-07-17 12:14:39 -07:00
geohot
bf299802f8
fixup tests
2022-07-17 12:11:53 -07:00
geohot
762e859089
testopencl
2022-07-17 11:56:40 -07:00
geohot
608e2431f7
test opencl, commit to removing the crap conv code from GPU
2022-07-17 11:55:37 -07:00
geohot
3c4565fa21
SLICE -> PAD,SHRINK
2022-07-17 11:33:59 -07:00
geohot
9574dd8559
some permutes are reshapes
2022-07-17 10:34:24 -07:00
geohot
77806e0d64
fix permute stacking
2022-07-17 10:24:57 -07:00
geohot
c28a99087b
add PAD movementop
2022-07-17 10:22:03 -07:00
geohot
4527c0453d
get_movementroot
2022-07-17 09:37:27 -07:00
geohot
b00cc93102
bugfixes
2022-07-17 09:19:22 -07:00
geohot
d07f379038
don't merge movement ops
2022-07-17 09:09:50 -07:00
geohot
f6ea7c022a
Revert "EXPAND -> REPEAT"
...
This reverts commit 115d2eadf5
2022-07-17 08:42:10 -07:00
geohot
cca089b11d
Revert "more expand -> repeat"
...
This reverts commit 2e7b1630a8
2022-07-17 08:41:48 -07:00
geohot
2e7b1630a8
more expand -> repeat
2022-07-17 08:40:49 -07:00
geohot
115d2eadf5
EXPAND -> REPEAT
2022-07-17 08:38:54 -07:00
geohot
1eb247f823
contiguous and same length
2022-07-16 08:49:07 -07:00
geohot
d04b274cd2
noop removal can replace with reshape
2022-07-16 08:32:42 -07:00
geohot
d985217fa4
skip reduce noops
2022-07-16 07:47:43 -07:00
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 <[email protected]
2022-07-16 07:26:19 -07:00
George Hotz
d651caa864
fixup openpilot/compile.py
2022-07-11 13:59:09 -07:00
geohot
5e46561f7e
no_grad = NOT backward
2022-07-10 20:54:57 -07:00
geohot
b34ae7876f
lol chr(10) not chr(13)
2022-07-10 20:03:11 -07:00
George Hotz and GitHub
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
geohot
057e4f5aa5
a little faster and cleaner
2022-07-09 08:14:01 -07:00
geohot
c77ba7fa3f
ops_cpu readibility
2022-07-09 07:48:54 -07:00
geohot
0a36475700
no einsum for now
2022-07-09 00:04:40 -07:00
geohot
c39a245696
that's not where i thought we'd lose lines...
2022-07-08 23:52:38 -07:00
geohot
75e1848b09
always SHUFFLE_RESHAPE_OPS
2022-07-08 23:19:39 -07:00
geohot
44848ee5dc
prints show we can precompute from the outside
2022-07-08 10:59:20 -07:00
geohot
68959be05d
precompute weights for opencl
2022-07-08 10:56:48 -07:00
geohot
d8e7f1f8bc
opencl type ignore
2022-07-08 10:33:55 -07:00
geohot
ae335b6d3e
opencl works, but tons of kernels
2022-07-08 10:22:04 -07:00
geohot
5b66d1bb0b
begin fixing up opencl
2022-07-08 10:20:14 -07:00
geohot
7e17f2ae8d
fix mypy, add TODOs
2022-07-08 09:57:22 -07:00
geohot
8557ed88df
use ast engine for merged reduceop
2022-07-08 09:37:40 -07:00
geohot
3656a5615a
MERGE_ELEMENTWISE_INTO_REDUCE
2022-07-08 09:32:28 -07:00
geohot
ca9532ce29
less lines, and typing found a bug
2022-07-08 08:57:12 -07:00
geohot
2035b89e54
wooo 1k lines
2022-07-08 08:44:57 -07:00
geohot
2a8c1071d9
cleanups
2022-07-08 08:36:31 -07:00
geohot
e6733286df
unify conv and reduce
2022-07-08 08:27:30 -07:00
geohot
9c34b3eef3
tighten up gpu kernels
2022-07-08 07:59:04 -07:00
geohot
563bf2d8e8
force input/weight to be contiguous (uncached)
2022-07-08 07:40:30 -07:00
geohot
1cf805a56a
fix no MERGE_MOVEMENT_OPS bug
2022-07-08 07:27:53 -07:00
geohot
c0ef998b48
remove finished todo
2022-07-07 11:39:25 -07:00
geohot
715e335c60
fix types
2022-07-07 11:36:09 -07:00
geohot
