github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-14 17:38:06 -05:00
Code Issues Packages Projects Releases Wiki Activity
10,934 Commits 203 Branches 14 Tags
er_prioity
Commit Graph

889 Commits

Author SHA1 Message Date
nimlgen
dafdb4bfb1 test hcq open with pytest (#13124) 
* test hcq open with pytest

* fi
 2025-11-06 20:09:51 +08:00
nimlgen
05e2ff4d87 system: fix flock on pcidevs (#13123) 
* system: fix locking of hcq devices

* rename and fullrun

* force ok

* fix

* fix
 2025-11-06 19:02:13 +08:00
chenyu
18d4ecc1f3 lower nv test_gemm_4096 target (#13107) 2025-11-05 11:05:16 -05:00
chenyu
54141e9cb9 DISABLE_COMPILER_CACHE=1 in speed_v_theoretical (#13096) 2025-11-04 11:28:18 -05:00
chenyu
f6430a0559 add script for one slow openpilot conv (#12953) 
* add script for one slow openpilot conv

* fix ruff
 2025-10-30 18:08:41 -04:00
George Hotz
f5a3b33d33 add fun with nhwc convs 2025-10-28 17:12:22 +08:00
Sieds Lykles
e22c5e7e73 process_replay uses opts argument for KernelInfo.opts_to_apply (#12946) 
* opts_to_apply is opts

* skip beamed kernels

* simpler change

* fix the tensor cores tests for process replay

* use opts
 2025-10-28 09:00:28 +01:00
chenyu
a79832b01f control_flow.py -> linearizer.py [pr] (#12948) 2025-10-27 12:38:13 -04:00
Sieds Lykles
e1f8c82938 Onnx Layer/Group/RMS/Batch-Norm ReduceL2 fp32 intermediates for fp16 (#12109) 
* match onnx spec

* use least_upper_dtype

* promote the square

* just cast before the square
 2025-10-24 12:26:11 +02:00
George Hotz
7762b3558b clean up the spec (#12868) 
* tighten up the spec

* move validate into a different file

* that moved to validate

* after(barr)
 2025-10-22 19:50:42 +08:00
George Hotz
92778c7a8b rename opts to ren, add store ranges back (#12856) 
* rename opts to ren

* fix docs and bring store back
 2025-10-22 09:15:38 +08:00
wozeparrot
62e7b8b870 feat: just use compile3 (#12849) 2025-10-21 07:56:50 -07:00
George Hotz
8960ac54f3 remove RewriteStep premature optimization (#12840) 
* remove RewriteStep premature optimization

* fix ebs

* core line count
 2025-10-21 21:45:20 +08:00
George Hotz
7d9551ce2e move to late/control_flow.py (#12835) 2025-10-21 18:15:06 +08:00
George Hotz
d711a4b933 delete old linearizer (#12834) 
* new linearizer with early endrange

* cleanups

* second stage removal

* not store

* do that later

* end cleanup

* fix globals

* end

* multi end

* fix ends earlier

* work

* do_merge_ends

* mini change

* range_gate

* fix cpu

* test fixups

* ranges on index

* not for ptx

* delete linearizer

* remove more junk

* delete that test

* we insert endif

* all ends
 2025-10-21 17:52:18 +08:00
George Hotz
c780cd9abb new linearizer with early endrange (#12823) 
* new linearizer with early endrange

* cleanups

* second stage removal

* not store

* do that later

* end cleanup

* fix globals

* end

* multi end

* fix ends earlier

* work

* do_merge_ends

* mini change

* range_gate

* fix cpu

* test fixups

* ranges on index

* not for ptx
 2025-10-21 17:37:48 +08:00
George Hotz
d59d4cdbe4 lil less is okay 2025-10-21 17:09:44 +08:00
chenyu
63a23dfe80 test step 0 in TestTrainingOnnxOps (#12790) 
and tighter rtol
 2025-10-19 09:15:49 -04:00
chenyu
e8158afd4b update test_qlinear_add_round_half_to_even (#12789) 
this does not pass locally
 2025-10-19 08:47:27 -04:00
chenyu
285534ce64 delete DONT_REALIZE_EXPAND and DONT_GROUP_REDUCES (#12744) 
does nothing now
 2025-10-16 14:11:33 -04:00
chenyu
98239f1156 few shapetracker cleanups (#12741) 2025-10-16 12:43:27 -04:00
George Hotz
1d1e1d9d88 delete the ShapeTracker (#12720) 
* delete the ShapeTracker

* fix tests

* fix more

* fix gc test
 2025-10-16 15:36:22 +08:00
George Hotz
db4a359374 fix up some slow tests that launch python (#12672) 
* fix up some slow tests that launch python

* svd nonfull in parallel

* split test_advancedindex
 2025-10-14 19:13:55 +08:00
George Hotz
84d4589ed4 remove pylint from pre-commit and CI (#12658) 
* remove pylint from pre-commit and CI

* multidevice test is fast

* faster pre-commit

* 8 is faster than 4

* better name

* how did that typecheck?
 2025-10-14 15:39:59 +08:00
George Hotz
b9eb5b5d49 clean up the LLM tokenizer (#12653) 
* clean up the LLM tokenizer

* simple tokenizer is actually simple

* ugh write good code
 2025-10-14 14:22:01 +08:00
qazal
fd51ecf983 process_replay for get_rangeify_map (#12624) 2025-10-12 15:14:40 +03:00
Sieds Lykles
4300ebc455 cache apply_movement_op (#12609) 
* cache apply_movement_op

* pyling and clear cache

* fix types

* ignore

* cleanup
 2025-10-11 08:53:10 +02:00
qazal
caae46cfba fix process replay progress update (#12587) 2025-10-10 10:20:55 +03:00
chenyu
c8dfd10257 ShapeTracker.real_strides -> is_expanded [pr] (#12579) 
only keep the used part
 2025-10-09 22:52:45 -04:00
George Hotz
a8a9ac0e95 add more uop gc test (#12553) 2025-10-09 14:49:32 +08:00
chenyu
ae51bdd06a remove trivial use of RANGEIFY flag (#12550) 
some tests need update still
 2025-10-09 02:29:38 -04:00
chenyu
be05028419 move ASSERT_MIN_STEP_TIME to compile3 (#12535) 
threshold is current time +20%
 2025-10-08 22:16:59 -04:00
chenyu
28edea5d67 delete FUSE_CONV_BW (#12527) 2025-10-08 10:41:38 -04:00
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
ad49f8148b switch process_replay to rangeify (#12509) 2025-10-08 11:26:43 +03:00
George Hotz
0f25b4b289 move frontend dir to nn [pr] (#12470) 2025-10-07 10:42:22 +08:00
chenyu
940a8d5ba9 default IGNORE_OOB=1 (#12441) 
* default IGNORE_OOB=1

z3 can get very slow with RANGEIFY, also update some kernel numbers to what it is

* add to test
 2025-10-03 04:16:19 -04:00
hooved
1e8945a28c Training loop for Stable Diffusion mlperf (#12315) 
* add diff

* fix edit error

* match master

* point reference to specific commit

* simplify wandb logging

* remove lr test, dehardcode device

* increase stack size limit
 2025-10-03 02:45:38 -04:00
hooved
5d9035f5a6 Eval for Stable Diffusion mlperf (#12316) 
* add diff

* rerun ci

* refactor beam workaround, add test

* fix conflict

* linting
 2025-10-02 02:35:38 -04:00
hooved
0f804c9a83 Stable Diffusion model init for mlperf (#12314) 
* include clip pr diff

* updated unet and sd init

* dehardcode default device

* revert beam hang workaround

---------

Co-authored-by: chenyu <chenyu@fastmail.com>
 2025-10-02 02:28:41 -04:00
qazal
2fcd55583f allow less kernels in external_test_opt (#12412) 
* allow less kernels in external_test_opt

* this was always 2
 2025-10-02 05:05:42 +03:00
hooved
969a1b35ca LR scheduler for Stable Diffusion mlperf training (#12201) 
* add lr scheduler for stable diffusion training

* add lr scheduler test

* rerun ci

* rerun CI

* use np for testing

* move test to CI path

* remove unneeded copy
 2025-09-30 21:21:08 -04:00
George Hotz
a83f219253 fix bad range merges (#12368) 
* fix bad range merges

* fix rng

* fix uop gc
 2025-09-30 19:30:21 +08:00
hooved
c2689c505e Clip model updates for Stable Diffusion mlperf training (#12313) 
* stable diffusion mlperf clip changes

* add clip tests

* set gelu as attribute

* add more tests

* factor out GPUS

* rerun CI

* add imports to if blocks

* remove unneeded axis

* add clip tests to CI

* move clip tests

* add deps, disable max buf size
 2025-09-29 21:50:14 -04:00
chenyu
0dad6cc518 good RANGEIFY kernel counts in external_test_opt (#12242) 
no push permute stuff. the model ones are less clear if it's good, some got slower
 2025-09-18 17:58:54 -04:00
nimlgen
92df52d79a make method_cache account for compiler (#12156) 
* make method_cache account for compiler

* sorry
 2025-09-13 17:00:11 +03:00
ttomsa
a12d0933c1 fix vec dtype in fast idiv (#12080) 
* fix

* add vec dtypes to fuzzer

* add vec=False

---------

Co-authored-by: Sieds Lykles <93992551+S-Lykles@users.noreply.github.com>
 2025-09-12 23:00:43 +02:00
Sieds Lykles
1f3950a484 Invalid idx (#12067) 
* merge index_dtype_3

* new lowering with Invalid idx

* remove that dtype from range

* finish merge

* annotate better

* indentation

* dont need that anymore

* always process replay for openpilot

* more uop_given_valid for idx

* valid past index_child

* fix bug preventing load getting an alt value

* add track_match_stats back in in shapetracker and remove cache

* get_valid_idx -> get_valid and get_idx

* fix heuristics with new idx

* split line

* fix typo

* fix signature

* dont skip idx if stride is 0

the idx may still be invalid

* lower const with new valid

* delete to_indexed_uops

* update shapetracker test

* delete axis_is_masked

* add cache back

* move around comment

* fix get_valid bug

* move invalid fold to symbolic so its earlier

* cleanup

* update applying padto to new idx

* add unit tests

* cleanup

* fold line

* improve spec

* dont try to render Invalid as a float

* more consistent invalid index

* update some tests

* Fold index with true cond

* skip test

* vconst min max if Invalid in arg

* fix signature of UOp.const

* add test for min/max of Invalid CONST/VCONST

* add InvalidType to as_const signature

* is Invalid to isinstance

* Add InvalidType to ConstLike

* index gate is a where gate

* make that a metaclass

* fix heurisics for new idx

* mypy happy
 2025-09-12 01:42:02 +02:00
chenyu
0e266f376c ops_gpu -> ops_cl (#12103) 2025-09-10 15:15:48 -04:00