nimlgen/tinygrad
1
0
Fork 0
forked from tinygrad/tinygrad
Code Pull Requests Activity
14,103 Commits 334 Branches 15 Tags
1757067fa65627ce8317f7cce507b259ba409b5b
Commit Graph
14103 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
George HotzandGitHub 1757067fa6 add device range as src[1] to multi (kimi) (#17264) 
* add device range as src[1] to multi (kimi)

* cleanup

* fix tests
 2026-07-28 16:39:34 -07:00
chenyuandGitHub f11f884f49 update dtypes.md (#17266) 2026-07-28 19:30:13 -04:00
chenyuandGitHub 23e9e76e8c DEFAULT_FLOAT/DEFAULT_INT ContextVar [pr] (#17265) 2026-07-28 19:08:18 -04:00
George HotzGitHubteeny-runner
 		755dfb243b rename CPU_COUNT to NUM_CPU_THREADS with cgroup awareness (#17263) 
Rename CPU_COUNT to NUM_CPU_THREADS so it can be overridden via env var.
Default uses _get_cpu_count() which respects cgroup limits:
  - os.process_cpu_count() on Python 3.13+
  - /sys/fs/cgroup/cpu.max on cgroup v2
  - /sys/fs/cgroup/cpu/cpu.cfs_quota_us on cgroup v1
  - os.sched_getaffinity(0) fallback

Use NUM_CPU_THREADS.value in the dataloader instead of cpu_count(),
and update export_model.py and all renderer references.

Co-authored-by: teeny-runner <runner@teeny>
 2026-07-28 15:33:57 -07:00
George HotzandGitHub 0cdddf3495 remove unneeded default args in renderers (#17262) 2026-07-28 14:53:04 -07:00
chenyuandGitHub fde3a8faff CAPTURE_PROCESS_REPLAY=0 default chaging test [PR] (#17261) 
* CAPTURE_PROCESS_REPLAY=0 default chaging test [PR]

quick patch before making default dtype context vars, which is quite annoying

* few more
 2026-07-28 17:32:35 -04:00
George HotzandGitHub 9ce65b71c0 delete dsp_pm_late (#17260) 2026-07-28 14:05:41 -07:00
George HotzandGitHub 7d48926293 switch _device_num to AxisType.DEVICE range (kimi) (#17252) 
* switch _device_num to AxisType.DEVICE range (kimi)

* improve

* don't exclude device for ranges

* current multi ends the range
 2026-07-28 13:52:47 -07:00
drkandGitHub d1c3ae0f91 preserve typed ranges through reshape (#17259) 2026-07-28 16:51:06 -04:00
chenyuandGitHub 749e002286 remove some dtype= when construct UOp [pr] (#17257) 
prep for removing the dtype field
 2026-07-28 16:05:35 -04:00
chenyuandGitHub c9e11544df delete explicit casts [pr] (#17255) 
handled by broadcast
 2026-07-28 13:57:13 -04:00
George HotzandGitHub a9ad08064e make the github actions runners generic for gitea (#17254) 2026-07-28 10:48:49 -07:00
chenyuandGitHub 37cf15942b UOp.const(dtype=None) infers from from_py [pr] (#17253) 
eventually we drop the dtype arg once everything const dtype is inferred
 2026-07-28 12:06:55 -04:00
qazalandGitHub 05e4727feb arange stack regression test (#17250) 
* simple failing test

* commend out stack

* a little smaller

* work
 2026-07-28 20:00:22 +09:00
qazalandGitHub e3b3eea1e2 test for extra copy in allreduce_cast (#16740) 
* test for extra copy in allreduce_cast

* simpler

* cleanup
 2026-07-28 19:02:54 +09:00
chenyuandGitHub 1380d6cc5a pm_long_decomp doesn't depend on operand dtype [pr] (#17249) 
* pm_long_decomp doesn't depend on operand dtype [pr]

the const the rules returned won't have strong dtype

* fix
 2026-07-28 01:15:04 -04:00
chenyuandGitHub 4b7022e8f4 Revert "64-bit UOp.variable support (#17246)" (#17248) 
This reverts commit ab8fb191b2.
 2026-07-28 00:00:26 -04:00
sirhcmandGitHub ab8fb191b2 64-bit UOp.variable support (#17246) 2026-07-27 23:56:09 -04:00
chenyuandGitHub f837ca3587 don't match const dtype in UPat [pr] (#17247) 2026-07-27 22:50:08 -04:00
chenyuandGitHub 654d475d37 update pm_long_decomp [pr] (#17245) 
* update pm_long_decomp [pr]

instead of reading dtype from UOp (which might be const that's going to be weak), just pass output dtype in tag

* cleanup tag
 2026-07-27 22:28:50 -04:00
chenyuandGitHub 550225f603 clean up pm_lower_weak [pr] (#17243) 2026-07-27 19:11:39 -04:00
chenyuandGitHub 0bb36c9989 make Tensor(None) weakfloat (#17241) 
match other float consts
 2026-07-27 16:45:28 -04:00
chenyuandGitHub 896afad9bf don't match strong typed const in UPat [pr] (#17240) 2026-07-27 16:19:01 -04:00
nimlgenandGitHub 8b9ef157d1 run_linear in external_test_gpu_crash (#17239) 2026-07-27 22:00:11 +03:00
b1tgandGitHub bdbb1d702f fix shard axis through symbolic reshape (#17238) 
* fix shard axis through symbolic reshape

* bind
 2026-07-27 11:32:15 -04:00
nimlgenandGitHub 8eaeede96d hcq2: do not cache beam (#17234) 2026-07-27 17:16:24 +03:00
chenyuandGitHub 818a892ebc flip from_py to use weak dtypes [pr] (#17229) 2026-07-27 10:07:00 -04:00
qazalandGitHub a3ca1e55a5 polish the viz readme (#17236) 
* polish the viz readme

* style

* no epilog=
 2026-07-27 22:03:13 +09:00
wozeparrotandGitHub 056974468e gptoss: split no-wd params (#17233) 2026-07-27 02:58:43 -07:00
qazalandGitHub 19c4d736f2 validate json output of viz.cli in CI (#17232) 
* validate viz.cli --json always prints valid JSON

* highest debug level

* jq empty we don't need a print

* gate that import
 2026-07-27 15:44:24 +09:00
chenyuandGitHub f45fc4c566 test updates from weak flip (#17231) 2026-07-27 01:27:42 -04:00
qazalandGitHub 95e3b0066f webgpu failing test for duplicate PARAM in CALL [pr] (#17230) 
* webgpu failing test for duplicate PARAM in CALL [pr]

* typo
 2026-07-27 14:06:06 +09:00
chenyuandGitHub 0f98212e80 skip test_float_to_fp8e4m3_extreme_values (#17228) 
fp8 overflow behavior changed in torch 2.13.0, skipped the test for now
 2026-07-26 21:01:51 -04:00
chenyuandGitHub 456b5b5060 fix python_alu inf (#17227) 2026-07-26 20:04:12 -04:00
chenyuandGitHub afb25a624c simpler minimum and copysign [PR] (#17226) 2026-07-26 19:54:22 -04:00
nimlgenandGitHub 165f0626f8 speedy hcq2 (#17225) 
* faster

* x
 2026-07-27 01:31:56 +03:00
nimlgenandGitHub 97a2265362 hcq2: amd indirect (#17220) 
* ind

* mock
 2026-07-26 21:36:01 +03:00
chenyuandGitHub 4b6760539b fix _prepare_jit_inputs for weak [pr] (#17222) 2026-07-26 14:03:27 -04:00
chenyuandGitHub 94dad3d261 clean up mixin cos and exp [PR] (#17221) 2026-07-26 13:13:05 -04:00
chenyuandGitHub a8d51097dc realize weak is no-op [pr] (#17219) 
None device and weak dtype are both virtual
 2026-07-26 11:52:56 -04:00
Manan GuptaGitHubGeorge Hotz
 		79c07a334c fix ValueError in UOp.axis for shard reshape crossing boundary (#16547) 
Co-authored-by: George Hotz <[email protected]>
 2026-07-26 08:38:10 -07:00
C TandGitHub d70a134845 fix nvrtc_check helper used for jitlink call (#16362) 2026-07-26 08:20:39 -07:00
George HotzandGitHub 960430a5e5 Revert "nv: set lower interleave level to reduce GPU hogging (ai slop) (#15518)" (#17218) 
This reverts commit ac12914506.
 2026-07-26 08:15:16 -07:00
JYCGitHubYasko CClaude Opus 4.6
 		ac12914506 nv: set lower interleave level to reduce GPU hogging (ai slop) (#15518) 
Fixes #10773. The NV backend makes desktop systems unusable (cursor lag,
video drops) because the channel group runs at the default HIGH interleave
level, monopolizing the GPU and starving the display compositor.

Sets tsgInterleaveLevel to LOW (0) by default so the GPU scheduler can
preempt compute work for display refresh. Dedicated compute machines can
restore full priority with NV_INTERLEAVE=2.

Also adds SET_INTERLEAVE_LEVEL to the mock GPU driver's pass-through list.

Co-authored-by: Yasko C <[email protected]>
Co-authored-by: Claude Opus 4.6 (1M context) <[email protected]>
 2026-07-26 08:12:05 -07:00
chenyuandGitHub 5d1aa84901 SHR/SHL are Broadcastable [pr] (#17216) 2026-07-26 02:11:03 -04:00
chenyuandGitHub acc2374b6f minor normalize cleanup [pr] (#17215) 
and add a few no-op clones in test_ops to prep for weak flip
 2026-07-26 00:39:53 -04:00
chenyuandGitHub a96974e70c slight weak behavior tweak and cleanups [pr] (#17214) 
* slight weak behavior tweak and cleanups [pr]

* ruff
 2026-07-25 22:51:24 -04:00
chenyuandGitHub eb889053bf weak frompy prerequisite [PR] (#17213) 2026-07-25 22:35:45 -04:00
chenyuandGitHub a60b5f77ac fix torch backend out= into a view (#17210) 2026-07-25 21:04:42 -04:00
nimlgenandGitHub 3946df787d hcq2, cpu is hcq2-ish (#17197) 
* m

* i

* x

* x

* Df

* x

* x
 2026-07-26 02:38:15 +03:00