mirrors/tinygrad
2
0
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-06-24 02:14:17 +00:00
Code Issues Projects Releases Packages Wiki Activity
11,106 commits 280 branches 15 tags 325 MiB
Commit graph

11,106 commits

This branch
This branch
All branches
Author SHA1 Message Date
qazal
2eab8021fb
 		remove inputs+outputs attributes from ScheduleItem [pr] (#9192) 
* remove inputs/outputs from ScheduleItem

* fix test_linearizer

* fix test_conv_shapetracker

* fix test_schedule + lint

* test_image_dtype + multitensor + search
 2025-02-21 13:48:11 +01:00
George Hotz
e87be0131e
 		torch backend start (#9191) 
* start torch backend

* progress

* ugh, you need cpp crap

* 1+1 works

* 1+1 works

* becoming a real backend

* ready to merge?
 2025-02-21 16:57:28 +08:00
George Hotz
d3a21cced2 hotfix: bump version to 0.10.2 v0.10.2 2025-02-21 10:43:49 +08:00
chenyu
2e7c2780a9
 		CLANG -> CPU (#9189) 2025-02-20 18:03:09 -05:00
nimlgen
f986e12f91
 		metal: choose compile spec based on macos (#9188) 
* metal: choose compile spec based on macos

* correction
 2025-02-21 00:43:39 +03:00
chenyu
3e22747799
 		run unit test on windows ci (#9187) 
* factor out testing_minimal in setup.py [pr]

* testing_unit + windows
 2025-02-20 14:40:41 -05:00
chenyu
287de4ecc6
 		use torch in test_gradient (#9186) 
used torch.autograd.grad, but not sure if it can be a template like jax
 2025-02-20 12:26:11 -05:00
qazal
574a905291
 		Fix running VIZ=1 after package installation + test (#9183) 
* test running viz from pip install

* add pkg

* do 10 connection attempts

* include assets in package_data

* quiet curl

* better print
 2025-02-20 15:02:00 +01:00
chenyu
1692087db5
 		_one_hot_along_dim input needs to be int (#9179) 
* _one_hot_along_dim input needs to be int

indexing and onehot compare with arange, and non-int dtype is likely a bug
 2025-02-20 09:00:43 -05:00
George Hotz
bf36967883
 		cuda hooking (#9180) 
* cuda hooking

* progress

* more hook cuda

* fix params

* compile + cuMemHostAlloc hook

* work

* revert that
 2025-02-20 19:20:01 +08:00
chenyu
3b37cc898b
 		add bert tiny config (#9177) 
set with BERT_SIZE=tiny. easier to study embedding and fusion
 2025-02-19 14:57:03 -05:00
qazal
5662c898f1
 		correctly step through bottom_up_rewrites in viz [pr] (#9176) 2025-02-19 19:20:57 +01:00
peppingdore
b1ddb2a1a6
 		fix win32 CPUProgram missing cache flush (#9171) 
* win32: fix missing inst cache flush, rename ptr->self.mem for consistency with posix code

* fix types, remove assert

* fix memory leak

* rm whitespace
 2025-02-19 21:38:51 +08:00
qazal
1bb9d78c7a
 		hotfix: add output buffer back to kernel parents + comment [pr] (#9174) 2025-02-19 14:22:01 +01:00
chenyu
975c318dbc
 		bert use int32 for input ids (#9173) 
original data was int32 for these. float might have caused precision issues
 2025-02-19 08:17:27 -05:00
qazal
e4a8bf28ea
 		scheduler cleanups + better cycle assert [pr] (#9172) 
* scheduler cleanups + better cycle assert [pr]

* type_verify after assign fixup

* don't need base

* always realize sink parents
 2025-02-19 13:30:58 +01:00
qazal
cf315d544b
 		rename can_pad arg to cache [pr] (#9170) 2025-02-19 12:24:59 +01:00
qazal
2fc8bf115d
 		remove support for VIEW with two sources in ops [pr] (#9168) 
* only 1 src views can exist [pr]

* views can still exist without a base, this is a separate project
 2025-02-19 11:10:18 +01:00
Ahmed Harmouche
a2afa523a0
 		Only add enable f16 directive if ShaderF16 is supported (#9163) 
* F16 in check in wgsl renderer

* Default in renderer to fix pickle

* Refactor f16 check
 2025-02-19 17:20:03 +08:00
Ahmed Harmouche
0f94b98646
 		Force WebGPU backend type [pr] (#9164) 
* Force webgpu backend type

* Mypy fix

* Rename to WEBGPU_BACKEND

* Add it to env_vars docs

* Remove link
 2025-02-19 17:19:39 +08:00
qazal
4bc708a9b0
 		do not create buffers we never realize in scheduler (#9165) 
* work

* delete

* fix

* works

* FUSE_CONV_BW

* FUSE_ARANGE

* becomes_map

* fix assign p1

* fix assign (diamond) - 2

* fix test_assign_double_diamond_reduce

* fix subbuffer

* faster rewrite

* fix simple_pads

* start metadata work

* do some diff cleanups

* make things that can't be images not images

* openpilot fix

* fix linter

* diff

* minimal diff

* more work on the diff

* metadata
 2025-02-19 10:11:47 +01:00
George Hotz
1c4e9bc363
 		image fixup tensor map [pr] (#8611) 
Co-authored-by: qazal <qazal.software@gmail.com>
 2025-02-19 10:11:06 +02:00
qazal
2a5fe3e700
 		whitespace changes from the map_tensors branch [pr] (#9167) 2025-02-19 09:52:59 +02:00
qazal
a773ff73e3
 		match image cast folding on the cast itself [pr] (#9166) 2025-02-19 09:31:34 +02:00
qazal
9a20063837
 		create subbuffer immediately before constructing ScheduleItem [pr] (#9162) 2025-02-18 21:07:52 +01:00
qazal
1c92534bff
 		hotfix: viz should show if there's a rewrite [pr] (#9161) 2025-02-18 19:11:03 +01:00
George Hotz
a330f3338c
 		save applied opts in ProgramSpec [pr] (#9150) 2025-02-19 00:40:03 +08:00
chenyu
ff05bff221
 		put bert data shard inside jit (#9160) 
python time 45ms -> 9ms, it was spending time to schedule the shard

also init bert data on CLANG since it's from numpy, so we don't create the tensor on default device then shard into GPUS
 2025-02-18 10:36:54 -05:00
qazal
679291e26a
 		assert only base maps to buffer [pr] (#9159) 2025-02-18 15:46:47 +01:00
qazal
4f592eeea6
 		hotfix: remove extra matcher for copy/buffer_view [pr] (#9157) 2025-02-18 13:21:24 +01:00
George Hotz
ff9b985d9f hotfix: View Base AST 2025-02-18 18:48:34 +08:00
George Hotz
30f470eaa3
 		UNIQUE UOp for buffer instead of arg (#9156) 
* UNIQUE UOp for buffer instead of arg

* factor out buffer spec
 2025-02-18 16:59:59 +08:00
qazal
38f5ea2132
 		increment writable buffers refcount from the kernel graph [pr] (#9153) 2025-02-18 10:20:02 +02:00
George Hotz
ddddcc165b
 		colors back in DEBUG=2 [pr] (#9155) 2025-02-18 16:17:57 +08:00
George Hotz
6d62966bf7
 		add support for named rewrites [pr] (#9152) 2025-02-18 16:07:04 +08:00
George Hotz
caee42e8a6
 		Revert "name from uops [pr] (#9151)" (#9154) 
This reverts commit 28897be9a2.
 2025-02-18 16:06:44 +08:00
George Hotz
28897be9a2
 		name from uops [pr] (#9151) 2025-02-18 15:52:03 +08:00
George Hotz
a4dab3ec3f
 		add name uop (#9149) 
* add name uop, TODO: refactor renderer to use

* renderer uses name uop

* fix tests

* render

* ptx
 2025-02-18 15:26:58 +08:00
George Hotz
2db8b4046a
 		minor linearizer refactor to finalize in rewrite [pr] (#9148) 2025-02-18 12:42:22 +08:00
George Hotz
df3b320f46
 		rewriter -> devectorizer [pr] (#9147) 2025-02-18 12:42:08 +08:00
chenyu
5dc1257ce0
 		clean up bert fake data iterator [pr] (#9145) 
reuse the same get_data_bert path in setup and real run
 2025-02-17 20:03:38 -05:00
qazal
751c517b6c
 		cancel viz request after the kernel clicked away [pr] (#9144) 2025-02-17 20:19:09 +01:00
chenyu
465421b525
 		fix Tensor.isclose (#9143) 
many corner cases around inf and nan
 2025-02-17 12:03:12 -05:00
qazal
36741cbbc1
 		enable real_size assert for test_conv_2x2_backward_one_view [pr] (#9142) 2025-02-17 17:53:44 +01:00
qazal
e9ff4ef4f7
 		s/ScheduleContext/GrouperContext [pr] (#9141) 
* refactor to kernel context [pr]

* s/ScheduleContext/GrouperContext [pr]
 2025-02-17 17:14:17 +01:00
qazal
96cc9f59e0
 		refactor to kernel context [pr] (#9140) 2025-02-17 16:57:14 +01:00
qazal
df6781332e
 		remove var_vals from the scheduler context [pr] (#9139) 
* remove var_vals from the scheduler context [pr]

* maps to int
 2025-02-17 16:43:50 +01:00
Ali Ladjevardi
35e9c4657b
 		Use proper units when printing beam time (#9103) 
* use proper units when printing beam time

* refactor DEBUG=2
 2025-02-17 23:41:38 +08:00
Clément Verrier
a7f91224eb
 		add Tensor.isclose() (#8844) 
* add `Tensor.isclose()`

* support `equal_nan`

so as to match PyTorch's behavior

* update unit tests

* remove some tests temporarily

* re-enable one test

* re-enable other test

* try to fix failing tests during CI

* save one line of code

---------

Co-authored-by: George Hotz <72895+geohot@users.noreply.github.com>
 2025-02-17 10:11:40 -05:00
qazal
2b787c3b17
 		hotfix: lower ul.disabled opacity for viz [pr] (#9138) 2025-02-17 15:16:48 +01:00