github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-16 18:37:57 -05:00
Code Issues Packages Projects Releases Wiki Activity
10,490 Commits 203 Branches 14 Tags
remove_on_stack
Commit Graph

10490 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
George Hotz
6fb74beb79 remove on_stack from graph_rewrite 2025-10-09 14:14:25 +08:00
George Hotz
1dc500426e remove restrictions on range ending in indexing (#12543) 
* remove restrictions on range ending in indexing

* early simplify

* Revert "early simplify"

This reverts commit 657d9972c2.

* disable const folding tests
 2025-10-09 13:53:08 +08:00
chenyu
585bd95b50 fix ruff 0.14.0 [pr] (#12547) 2025-10-09 01:52:30 -04:00
qazal
6af29b913b viz: format rewrite time as a comment (#12545) 
* viz: format rewrite time as a comment

* put above
 2025-10-09 07:14:27 +03:00
qazal
baab7e334d put match times in viz (#12544) 
* put match times in viz

* float
 2025-10-09 06:56:10 +03:00
George Hotz
51420d1f99 rangeify profiling (#12540) 
* clean up stable diffusion weight loading

* add profiling to run_rangeify

* fix tests
 2025-10-09 11:32:34 +08:00
chenyu
43bce1f39f delete View minify [pr] (#12538) 2025-10-08 23:25:53 -04:00
qazal
9f9a8b0b5b viz: fix tiny device linking (#12541) 2025-10-09 06:25:33 +03:00
George Hotz
6e6059dde0 clean up stable diffusion weight loading (#12452) 2025-10-09 11:13:11 +08:00
chenyu
20d98b19c3 delete more unused ShapeTracker stuff (#12536) 2025-10-08 23:09:44 -04:00
qazal
bb5671a837 some more ops.py cleanups (#12525) 
* remove GroupOp.Meta and st_arg

* inline axis_arg

* only allow .buffer on reshapes (or the buffer)

* gate is the other way

* still want can_pad?

* use op_in_backward_slice_with_self

* .buffer is recursive

* lint

* pathlib there
 2025-10-09 06:06:44 +03:00
chenyu
be05028419 move ASSERT_MIN_STEP_TIME to compile3 (#12535) 
threshold is current time +20%
 2025-10-08 22:16:59 -04:00
George Hotz
615ec6acf0 refactor to apply_movement_op (#12533) 
* refactor to apply_movement_op

* new pm_mops is fine

* make mypy happy

* cleanup apply_movement_op function
 2025-10-09 10:16:09 +08:00
chenyu
c4732a18bd update tests that depend on SPLIT_REDUCEOP (#12534) 2025-10-08 21:53:30 -04:00
chenyu
5986d656a2 tighter ASSERT_MIN_STEP_TIME (#12531) 
set to about 1.2x of actual time now
 2025-10-08 21:22:54 -04:00
George Hotz
fc2bd53700 chatgpt nits (#12529) 
* tsink_base wasn't needed

* nits from chatgpt
 2025-10-09 07:34:44 +08:00
nimlgen
89ec2b3a74 memory: move bump allocator (#12505) 2025-10-08 23:12:04 +08:00
George Hotz
84fc34b274 tsink_base wasn't needed (#12528) 2025-10-08 22:46:06 +08:00
chenyu
28edea5d67 delete FUSE_CONV_BW (#12527) 2025-10-08 10:41:38 -04:00
George Hotz
2653147cb7 delete the lowerer (#12526) 2025-10-08 21:58:18 +08:00
George Hotz
0774575442 delete the old rangeify path and all the children stuff (#12524) 
* delete the old rangeify path and all the children stuff

* remove the on_stack stuff and any retries

* don't use the p word

* Revert "remove the on_stack stuff and any retries"

This reverts commit 49a2b328b9.
 2025-10-08 21:24:04 +08:00
Rudeus
a65ec5c693 fix fromarray depreceation (#12512) 2025-10-08 09:13:26 -04:00
qazal
b6835f4134 remove Ops.VIEW and related UOp methods (#12522) 
* remove Ops.VIEW and related UOp methods

* update abstractions2.py

* no ShapeTrackers in abstractions2.py

* it's a size 1
 2025-10-08 14:47:02 +03:00
George Hotz
3b0b3a2e64 fast RANGEIFY (#12504) 
* rtoposort is fast, can replace rangeify with this

* fast rangeify

* work

* fast rangeify works for mnist

* should work

* progress

* pad fix

* FAST

* tests passing

* don't delete those shape ops

* put in rangeify map

* ending ranges fix

* tests

* mstack/mselect no hacks

* move to indexing.py

* touch up tests + add comments

* disable failing test

* actually make the file readable

* failing

* error
 2025-10-08 19:38:06 +08:00
qazal
9448924d9e update gpt2 kernel count tests in CI=0 (#12523) 2025-10-08 14:29:11 +03:00
qazal
c5a1f9f5f9 no ShapeTrackers in multi.py (#12521) 
* switch multi to all movement ops

* inline dvars
 2025-10-08 14:04:05 +03:00
chenyu
ee0382ad99 remove ShapeTracker.invert (#12520) 2025-10-08 18:37:34 +08:00
chenyu
d5058427ea remove ShapeTracker.real_size (#12519) 2025-10-08 06:15:29 -04:00
qazal
6f26603f06 delete swizzler.py (#12518) 
* delete swizzler

* remove merge_views tests

* don't need rewrites_for_views

* apply_rewrites
 2025-10-08 13:02:34 +03:00
qazal
7e0b14243e delete grouper and kernelize (#12517) 
* delete grouper and kernelize

* +sys.setrecursionlimit
 2025-10-08 12:27:26 +03:00
chenyu
942022c309 smaller LLAMA_LAYER in Test llama 3 training (#12516) 
very slow now
 2025-10-08 05:10:51 -04:00
chenyu
e701106a64 remove FUSE_ARANGE (#12511) 
it was the default already
 2025-10-08 04:54:07 -04:00
qazal
291a19650b move Kernel dataclass to rangeify (#12510) 2025-10-08 11:30:06 +03:00
qazal
ad49f8148b switch process_replay to rangeify (#12509) 2025-10-08 11:26:43 +03:00
chenyu
da1f46ff3f remove RANGEIFY specific test jobs (#12507) 2025-10-08 04:12:04 -04:00
George Hotz
1e567a5cf8 make RANGEIFY=1 the default (#12161) 
Co-authored-by: chenyu <chenyu@fastmail.com>
Co-authored-by: Sieds Lykles <93992551+S-Lykles@users.noreply.github.com>
Co-authored-by: qazal <77887910+Qazalin@users.noreply.github.com>
 2025-10-08 03:46:09 -04:00
nimlgen
9e7103647d amd: rename cmd_id to sqtt_next_cmd_id (#12503) 
* amd: rename cmd_id to sqtt_next_cmd_id

* and typo
 2025-10-08 15:16:19 +08:00
nimlgen
4a756a37d8 amd: support rocm7 (#12502) 
* amd: support rocm7

* mock
 2025-10-08 14:30:39 +08:00
qazal
60b6dca5ba update some tests instead of expect_rangeify_fails (#12500) 
* update test_clone_doesnt_dedup to use base

* new_flat_buffer passes

* fix test_reorder_expand

* remove the view stuff

* remove that test, we don't want this view const behavior

* test_setitem_becomes_subbuffer is good
 2025-10-08 07:42:31 +03:00
qazal
84597ed53c early assert for device mistmatched asts in rangeify (#12499) 
* early assert for device mistmatched asts in rangeify

* alt also passes
 2025-10-08 07:19:36 +03:00
qazal
2e19354c1c viz: reorder timeline graphs (#12498) 
* viz: reorder timeline graphs

* update test_viz with the new order
 2025-10-08 07:10:23 +03:00
George Hotz
d06226b575 fix SPEC and all_tensors iterator (#12496) 2025-10-07 23:18:17 -04:00
qazal
a7cb80bfab use recursive_property in UOp device (#12477) 
* simple failing test with RecursionError

* switch to @recursive_property

* merge 2

* diff
 2025-10-08 06:15:05 +03:00
George Hotz
a6d59a0b45 backward_slice to get srcs recursively (#12494) 
* change name to backward_slice

* faster check

* clean up comments and names

* comment
 2025-10-08 10:31:42 +08:00
chenyu
eb3bc277b3 remove ASSERT_MIN_STEP_TIME in external_benchmark_openpilot (#12495) 
should add for compile3 and compile 3 only
 2025-10-07 22:13:42 -04:00
qazal
239f9a3029 update viz to not use children [pr] (#12493) 2025-10-08 04:35:01 +03:00
Sieds Lykles
b465c17b56 Revert "UOp.factor and add chain sorting (#12413)" (#12492) 
This reverts commit e74be4a140.
 2025-10-08 03:20:23 +02:00
George Hotz
945cc46475 delete children tracking from uop (#12491) 
* delete children tracking from uop

* uop children no longer exists

* no tracked children

* that test is flaky too
 2025-10-08 09:04:14 +08:00
nimlgen
648e5bb223 hcq: do not raise when fini (#12487) 
* hcq: do not raise when fini

* Revert "hcq: do not raise when fini"

This reverts commit 44af5f7d05.

* this way

* runtime is fine

* nn
 2025-10-07 23:27:03 +08:00
George Hotz
a2345787b9 parents is faster than sparents (#12490) 2025-10-07 21:31:50 +08:00