sirhcm/tinygrad
1
0
Fork 0
forked from tinygrad/tinygrad
Code Pull Requests 2 Actions Activity
11,106 Commits 336 Branches 15 Tags
python_speed
Commit Graph
4667 Commits
Author SHA1 Message Date
Sieds LyklesandGitHub 37d3ca152e Adapt >> for division by power of two to all ints (#10803) 
* Change divison by power of two to always use shift

* Change test to test int instead of uint

* simplify condition

* add old rule back with comment

* remove import

* use sresolve instead of simplify

* use keyword in simplify instead of sresolve

* webgpu cast y to uint

* remove comment

* explicitly set dtype in wgsl

* without simplify

* undo simplify kwarg

* change test to test both int32 and uint32
 2025-06-14 14:55:51 -04:00
chenyuandGitHub 652db5702b move test_conv_shapetracker and some test_search util into unit test (#10812) 2025-06-14 13:29:32 -04:00
leopfandGitHub 118a09ddcf xor self folding (#10806) 
* xor folding

* tests + z3 bitwise xor
 2025-06-14 10:01:17 -04:00
chenyuandGitHub 8c28b5d833 move dtype spec tests into unit test (#10808) 
* move dtype spec tests into unit test

can clean up more after the split

* skip CI test_backward_sum_acc_dtype
 2025-06-13 22:21:22 -04:00
chenyuandGitHub 7a6df0a161 remove .relu() call in several conv tests in test_ops (#10807) 
* remove .relu() call in several conv tests in test_ops

testing negative parts double the effectiveness. keep the relu between two convs and the tests that explicitly test relu

* relax tol
 2025-06-13 17:10:16 -04:00
qazalandGitHub a113c5e3ae viz: update browser test to properly shutdown [pr] (#10793) 
Using `await page.evaluate` can cause non deterministic `TargetCloseError`
exceptions if it cannot find the elements on the page, Puppeteer
doesn't cleanly stop when `browser.close()` is called.
[Failing CI](https://github.com/tinygrad/tinygrad/actions/runs/15596803685/job/43928961323?pr=10763#step:9:61)
 2025-06-12 17:58:42 +03:00
wozeparrotandGitHub eb739bb96a hotfix: lower threshold (#10786) 2025-06-11 19:36:20 -04:00
Sieds LyklesandGitHub 10b61157b9 Support symbolic slice with no start [pr] (#10775) 
* add symbolic slice with no start

* reshape the test

* step must be int

* just add a cast...

* more cast...
 2025-06-11 16:00:38 -04:00
George HotzandGitHub a38947b4bb move symbolic and transcendental to uop [pr] (#10771) 2025-06-10 20:51:22 -07:00
chenyuandGitHub 612cdf5146 move fuzz_shape_ops to run with other fuzzer (#10767) 
* move fuzz_shape_ops to run with other fuzzer

* don't skip CPU
 2025-06-10 17:43:04 -04:00
b1tgGitHubb1tg
 		52c49dd4f3 fix onnx ci (#10762) 
Co-authored-by: b1tg <[email protected]>
 2025-06-10 14:28:40 -04:00
chenyuandGitHub 14fa62c61d move high level tests to unit (#10760) 
either no need a backend, or running on one to check suffice
 2025-06-10 12:55:44 -04:00
Sieds LyklesandGitHub 0daa4c6ed0 Add DType.min and DType.max properties (#10749) 
* add properties

* cleaner test

* remove added newline
 2025-06-10 08:31:34 -07:00
qazalandGitHub 5d9c274924 keep UOp tags if sources are replaced (#10754) 
* keep UOp tags in unified_rewrite

* add failing test, print tag if defined

* remove the repr change
 2025-06-10 08:30:14 -07:00
George HotzandGitHub acf72872b3 move view left to the outer graph prereqs + testing (#10725) 
* move view left to the outer graph

* global view right

* dont need that one

* remove comment

* test kernelize

* simple

* split onnx, test sdxl null

* fix testing

* ugh, wrong one

* Update test.yml
 2025-06-09 20:43:25 -07:00
chenyuandGitHub b7198fdcfd linearizer failure from wino fuse arange cifar (#10739) 2025-06-09 23:10:19 -04:00
George HotzandGitHub 81ef879da3 non recursive top_down_rewrite (#10729) 
* non recursive top_down_rewrite

* nicer algorithm

* rewrite bottom up also

* only top down is broken?

* simpler iterative algo

* no recursion errors

* top down and bottom up

* unified rewrite

* simpler rewrite

* clean up comments

* move that comment
 2025-06-09 16:33:04 -07:00
chenyuandGitHub 53cbd4254b suppress filter_too_much on test_float_cast_to_unsigned (#10733) 
falky, already done in test_float_cast_to_unsigned_overflow and test_float_cast_to_unsigned_underflow
 2025-06-09 18:30:04 -04:00
chenyuandGitHub 55cdbb9a20 fix mask in expand into symbolic size (#10730) 
failed before when old size is 1 and it expands into symbolic size, because `resolve(s != ns, False)` is False and it does not expand the mask
 2025-06-09 17:33:22 -04:00
wozeparrotandGitHub 926b11381c failing test for symbolic expand after pad (#10727) 
* feat: failing test for symbolic expand after pad

* feat: mark test as failing
 2025-06-09 16:55:21 -04:00
chenyuandGitHub 49f999d919 update _reshape_mask for symbolic shape expand (#10726) 
* don't merge shape symbolic reshape symbolic

* proper fix
 2025-06-09 16:35:02 -04:00
wozeparrotandGitHub 27dd97f688 support variable shape none slice in getitem (#10724) 2025-06-09 11:53:02 -07:00
George HotzandGitHub f84c320548 better external_benchmark_schedule [pr] (#10722) 2025-06-09 10:26:11 -07:00
b1tgGitHubb1tgchenyu
 		24d328e313 onnx parser (#10435) 
* onnx parser

* fix compile, lint

* onnx.load -> onnx_load

* compatible with ModelProto

* fix test external_test_onnx_ops.py

* fix tests

* fix signed int

* reduce to 261 lines

* fix TypeProto.Optional

* debug for _parse_message, add TypeProto.Sequence, cleanup

* onnx_load from Tensor

* remove BufferedReader

* 174 lines and reduce tensor copy

* cleanup

* use onnx_load in external_model_benchmark.py

* fix qcom test

* [onnx] parser support external data

---------

Co-authored-by: b1tg <[email protected]>
Co-authored-by: chenyu <[email protected]>
 2025-06-09 12:44:28 -04:00
George HotzandGitHub 81b9c04574 move high level stuff to unit tests [pr] (#10708) 
* move high level stuff to unit tests [pr]

* process replay on unit tests

* fix pr, less compute

* set omp num threads

* set 200MB buffer size limit

* delete junk

* fix tests

* faster

* move test_indexing to unit

* faster
 2025-06-08 14:05:56 -07:00
George HotzandGitHub 4e2c3560b4 smaller tests are faster tests [pr] (#10704) 
* remove del spam from CI

* more

* preconstruct default buffer spec

* ignore those errors

* check exception

* more exception check

* skip stuff

* smaller tests mean faster tests

* a few more
 2025-06-08 10:54:19 -07:00
George HotzandGitHub 32e9949052 rename lazydata to uop (#10698) 2025-06-08 08:42:22 -07:00
uuuvnandGitHub 8e3f337075 Skip flaky test in ci (#10696) 
`test_data_parallel_resnet_train_step` is already skipped on LLVM/CPU:

```python
@unittest.skipIf(CI and REAL_DEV in ("CUDA", "NV", "LLVM", "CPU"), "slow, and flaky on LLVM/CPU")
@unittest.skipIf(REAL_DEV == "WEBGPU" and not OSX, "WEBGPU Vulkan can only run kernels with up to 10 buffers")
def test_data_parallel_resnet_train_step(self):
```

It looks like `test_data_parallel_resnet` (no `_train_step`) is flaky in a similar way:
https://github.com/tinygrad/tinygrad/actions/runs/15472667248/job/43560773882?pr=10642#step:9:64
 2025-06-08 08:24:09 -07:00
George HotzandGitHub 8c76250d31 speed up a few tests (#10692) 2025-06-07 20:39:25 -07:00
iharandGitHub 40c1479267 added unit tests for 'argfix' (#10678) 2025-06-07 22:17:10 -04:00
iharandGitHub 74b849b5e1 remove unnecessary 'argfix' because 'view' is an alias to 'reshape'. all functionality must be inside 'reshape' (#10677) 
* remove unnecessary 'argfix' because 'view' is an alias to 'reshape'. all functionality must be inside 'reshape'

* added the same set of unit tests for 'view' as for 'reshape' since 'view' is just an alias for 'reshape'

* improved tests for 'view' op
 2025-06-07 22:15:31 -04:00
Sieds LyklesandGitHub c29a56dd51 Fix whisper OOB (#10685) 
* fix whisper and test

* remove import
 2025-06-07 20:23:50 -04:00
George HotzandGitHub 53ed64e133 ci speed work 1 (#10676) 
* skip a few slow tests

* use a venv for python packages

* create venv

* no user, it's in venv

* ignore venv

* venv

* new cache key

* try that

* this

* version the python cache
 2025-06-07 16:33:11 -07:00
qazalandGitHub cb61774ab6 move shared viz fields out of serve.py [pr] (#10684) 
* move shared viz fields out [pr]

* update javascript

* update test_viz
 2025-06-07 17:18:18 +03:00
qazalandGitHub b515d796fb inline viz get_name [pr] (#10682) 
* inline viz get_name [pr]

* changing name_fxn makes this simpler

* waitUntil dom
 2025-06-07 11:16:16 +03:00
wozeparrotandGitHub e3805171e2 feat: variable bs bitcast (#10674) 2025-06-06 17:21:53 -07:00
George HotzandGitHub 54db1f8ee8 prevent huge waste of multi ram (#10669) 
* prevent huge waste of multi ram

* fix ram usage

* only define var

* add resolve

* fix tests

* fix cifar training

* remove that logic

* fix test without long
 2025-06-06 17:17:21 -07:00
George HotzGitHubchenyu
 		b68b7dbc2a test winograd is close to normal conv [pr] (#10557) 
Co-authored-by: chenyu <[email protected]>
 2025-06-06 19:11:49 -04:00
leopfGitHubGeorge Hotzgeohotwozeparrot
 		eb7305e6a4 Tensor.keccak("sha3_256") (#7186) 
Co-authored-by: George Hotz <[email protected]>
Co-authored-by: George Hotz <[email protected]>
Co-authored-by: wozeparrot <[email protected]>
 2025-06-06 15:24:05 -07:00
chenyuandGitHub bdede4924e fix odd number in get_test_global_size (#10671) 
factor might not be a integer if input global_size has an odd number in it
 2025-06-06 17:31:35 -04:00
George HotzandGitHub 7f0f97aa76 new test_multitensor tests (#10667) 
* new test_multitensor tests

* cleanup scheduler
 2025-06-06 10:26:28 -07:00
chenyuGitHubgeohot
 		4a6d84c4c3 hotfix llama start_pos vmax is max_context-1 (#10659) 
* hotfix llama start_pos vmax is max_context-1

fixed `IGNORE_OOB=0 python3 examples/llama3.py --size 1B --benchmark --temperature 0`

* hotfix: multitensor transformer test tests kv cache

---------

Co-authored-by: George Hotz <[email protected]>
 2025-06-06 00:41:25 -04:00
geohot 5eb6e1e65a Revert "hotfix: multitensor transformer test tests kv cache" 
This reverts commit ad9f88419a.
 2025-06-05 21:15:34 -07:00
geohot ad9f88419a hotfix: multitensor transformer test tests kv cache 2025-06-05 21:08:57 -07:00
George HotzandGitHub 8325c4f192 tests for multi assign (#10658) 
* tests for multi assign

* transformer tests

* add that assert
 2025-06-05 20:56:40 -07:00
wozeparrotandGitHub 0d86f8d375 fix failed threefry (#10646) 2025-06-05 17:17:42 -07:00
chenyuandGitHub ff1aad7b69 fix const float pow to int tensor (#10655) 
was incorrectly casted into int
 2025-06-05 19:15:12 -04:00
George HotzandGitHub baba274a76 minimal mstack pr to fix allreduce (#10649) 
* minimal mstack pr to fix allreduce

* fix webgpu
 2025-06-05 15:14:53 -07:00
George HotzandGitHub 4c315f8e17 MSTACK little non-functional changes (#10648) 2025-06-05 13:20:22 -07:00
chenyuandGitHub 46811d0d3c minor external_model_benchmark cleanup (#10644) 2025-06-05 14:13:28 -04:00