Commit Graph

Select branches

Hide Pull Requests

geohot/updates

gh-pages

master

remove_anchors

#1

#2

#3

v0.10.0

v0.10.1

v0.10.2

v0.10.3

v0.11.0

v0.12.0

v0.13.0

v0.4.0

v0.5.0

v0.6.0

v0.7.0

v0.8.0

v0.9.0

v0.9.1

v0.9.2

Mono Color

• 53d6f1e1bb Add bitonic cat sort (#9422) 3 people 2025-03-18 00:01:23 +08:00
• f53be010d7 lower bert learning rate (#9481) chenyuandGitHub 2025-03-17 10:49:56 -04:00
• e03c0aacf2 more explicit DONT_PUSH_VIEWS [pr] (#9479) qazalandGitHub 2025-03-17 20:43:21 +08:00
• 3b00a778ba fix view_left for unsafe pad ops [pr] (#9478) qazalandGitHub 2025-03-17 19:02:02 +08:00
• 813f713edc merge_views for buffer ops + create valids last (#9472) qazalandGitHub 2025-03-17 17:15:44 +08:00
• bd1f71c1e2 simple failing test for extra ops in VALID [pr] (#9474) qazalandGitHub 2025-03-17 17:02:40 +08:00
• e26caf4c3a hotfix: skip test_mean_half_precision_underflow on amd ci (#9476) qazalandGitHub 2025-03-17 16:47:48 +08:00
• 824c5f41ac dsp work try 3 (#9475) George HotzandGitHub 2025-03-17 16:42:12 +08:00
• 242daa4f9a ptrcat (#9473) George HotzandGitHub 2025-03-17 16:06:37 +08:00
• 52ae9af4dd Fast DSP for MobileNetV2 (try 2) (#9467) George HotzandGitHub 2025-03-17 15:10:36 +08:00
• 15ee742afa add get_children_map to uop (#9470) George HotzandGitHub 2025-03-17 14:36:13 +08:00
• d2cfbd8a4d bert lower learning rate and total steps (#9466) chenyuandGitHub 2025-03-16 17:21:20 -04:00
• 09e7708b49 minimum change for rdna4 [pr] (#9455) George HotzandGitHub 2025-03-16 13:39:24 +08:00
• be2161652b reorder into swizzler + ast_fixup [pr] (#9456) qazalandGitHub 2025-03-15 09:00:14 +01:00
• cb7a7f69c7 quantization preprocessor from DSP, should be universal (#9437) George HotzandGitHub 2025-03-15 07:49:37 +08:00
• ca5064a5b6 remove Kernel.float4_axis [pr] (#9448) chenyuandGitHub 2025-03-14 17:54:32 -04:00
• 0e591baf43 redo simple_matmul change (#9450) chenyuandGitHub 2025-03-14 17:53:52 -04:00
• b0f63d3c04 Revert "simple_matmul.py uses np to generate random (#9438)" (#9449) chenyuandGitHub 2025-03-14 17:14:22 -04:00
• 14018050c1 simple_matmul.py uses np to generate random (#9438) Ignacio SicaandGitHub 2025-03-14 17:36:50 -03:00
• 2a50e6440d filter sink by DONT_PUSH_VIEWS + remove extra base [pr] (#9446) qazalandGitHub 2025-03-14 21:27:46 +01:00
• 3af7a08a06 ast_fixup in one graph_rewrite pass [pr] (#9444) qazalandGitHub 2025-03-14 20:14:31 +01:00
• bd4ae5ac53 am: hotfix: import modules (#9443) nimlgenandGitHub 2025-03-15 03:10:18 +08:00
• 77a8430616 am: use smu based on discovery (#9441) nimlgenandGitHub 2025-03-15 02:10:45 +08:00
• 5ff90cb261 am: less magic values (#9440) uuuvnandGitHub 2025-03-14 23:10:35 +05:00
• 459d0cd14f add arch to AMDRenderer and HIPRenderer (#9431) Ignacio SicaandGitHub 2025-03-13 13:06:27 -03:00
• 357e364ab8 am: turn off unord dispatch (#9433) nimlgenandGitHub 2025-03-13 23:59:28 +08:00
• 99b0287e4e add GROUP and GROUPTOP to test_arange (#9432) chenyuandGitHub 2025-03-13 11:28:38 -04:00
• 90ffa9bd45 swizzle without buffer ops try 2 [pr] (#9427) qazalandGitHub 2025-03-13 11:00:40 +02:00
• 4df2b6347d hotfix: bump tinybox red training CI timeout to 30 minutes (#9426) qazalandGitHub 2025-03-13 10:31:44 +02:00
• 931436204c hotfix: 12000 lines, for AMD stuff geohot 2025-03-13 10:48:14 +08:00
• bfc68d1953 add gep rules to simplify (#9419) George HotzandGitHub 2025-03-13 09:46:25 +08:00
• 0bed9b6cd2 benchmark huggingface onnx models (#8493) 3 people 2025-03-13 08:13:12 +08:00
• 4992958dae update bert beam params (#9423) chenyuandGitHub 2025-03-12 13:00:41 -04:00
• 12978f0d05 reorder contiguous/assign ast rules [pr] (#9420) qazalandGitHub 2025-03-12 13:13:27 +02:00
• 5f6d5b057d expand index isn't grouping by access size (#9418) George HotzandGitHub 2025-03-12 17:24:10 +08:00
• 815ad0b7a8 support load/store grouping in DEVECTORIZE=0 (#9409) George HotzandGitHub 2025-03-12 11:34:37 +08:00
• 4714c4f9ad torch backend multigpu - add devices and tests (#9414) Priyank PatelandGitHub 2025-03-11 20:33:11 -07:00
• 22fc0a2e36 bert sum acc in half (#9412) chenyuandGitHub 2025-03-11 23:03:15 -04:00
• f995b465b8 am: set doorbell offsets to nb (#9413) nimlgenandGitHub 2025-03-12 10:35:47 +08:00
• 95e0f069be hotfix: gitignore *.log [pr] (#9410) qazalandGitHub 2025-03-11 22:39:19 +02:00
• 78ebade125 Merge pull request #9408 from nimlgen/hcq_progress_during_wait nimlgenandGitHub 2025-03-11 19:40:23 +08:00
• e174c6c3bc new devectorizer (#9331) George HotzandGitHub 2025-03-11 18:47:56 +08:00
• 69fac5fe89 Merge pull request #9407 from tinygrad/no_const_after_sym qazalandGitHub 2025-03-11 12:24:09 +02:00
• 4d09ea4c06 hcq: reset timer on progress in singal.wait nimlgen 2025-03-11 10:02:07 +00:00
• fa69fd3afc no const/view in schedule sink after sym [pr] qazal 2025-03-11 10:58:23 +01:00
• 68f062c8be cast_vec on transcendental (#9406) George HotzandGitHub 2025-03-11 17:30:46 +08:00
• e85001b6ee SQTT profiling (#9278) 3 people 2025-03-11 10:19:56 +05:00
• 2780e2027e devectorize prereqs [pr] (#9404) George HotzandGitHub 2025-03-11 12:33:29 +08:00
• beed00eabe fix torch backend memory leak (#9395) 3 people 2025-03-10 19:48:20 -07:00
• 01e8b60911 acc_dtype -> dtype (#9402) chenyuandGitHub 2025-03-10 16:05:30 -04:00
• 59dfb234eb replace hardcoded ast with tensors in TestSwizzle [pr] (#9401) qazalandGitHub 2025-03-10 20:33:57 +02:00
• 796c3bbb23 torch: support in-place operations on views (#9371) 3 people 2025-03-10 08:29:00 -07:00
• 2afc7759a7 sink in kernel op [pr] (#9397) qazalandGitHub 2025-03-10 14:13:42 +02:00
• 25847080f0 olmoe (from stream, wip) (#9390) George HotzandGitHub 2025-03-10 13:46:33 +08:00
• 1d64c12f2b add Topk to tensor (#9343) geohotstanandGitHub 2025-03-10 08:01:42 +08:00
• a1f41fadf6 test_schedule cleanups + add DONT_GROUP_REDUCES [pr] (#9392) qazalandGitHub 2025-03-09 16:01:08 +02:00
• b6fe5ab4dd fix: correct gfx10 ctl stack size (#9384) wozeparrotandGitHub 2025-03-09 00:03:20 -05:00
• 456697d0be always create kernels for assign/contiguous/copy [pr] (#9388) qazalandGitHub 2025-03-08 16:32:06 +02:00
• 286b480f82 do not replace assign with the offset buffer [pr] (#9387) qazalandGitHub 2025-03-08 12:57:44 +02:00
• ecfccdea8e remove views from the kernel graph minimum diff (#9385) qazalandGitHub 2025-03-08 11:14:42 +02:00
• 0d2762c010 prep refactor for adding buffer ops last [pr] (#9383) qazalandGitHub 2025-03-08 09:00:14 +02:00
• bde0347618 amd: support relocatable elf (#9380) 3 people 2025-03-08 02:21:49 +08:00
• 243078dda9 am: optimize tlb usage (#9049) nimlgenandGitHub 2025-03-07 19:37:29 +03:00
• 46720294d6 reorder ScheduleItem creation [pr] (#9379) qazalandGitHub 2025-03-07 18:20:53 +02:00
• dc89dae994 remove unmasked valid after swizzles (#9377) qazalandGitHub 2025-03-07 17:43:16 +02:00
• 088d86691b fix onnx gather and onnx auto_pad VALID mode (#9375) geohotstanandGitHub 2025-03-07 23:27:23 +08:00
• 3565c08df5 refactor to kernel ast fixup [pr] (#9376) qazalandGitHub 2025-03-07 16:47:38 +02:00
• 304afe0d55 tinychat in browser, Part 3: browser app (#9276) 3 people 2025-03-07 02:07:33 -05:00
• 136cf7b8b1 hotfix: load >2 GiB from disk on macOS (#9361) hoovedandGitHub 2025-03-07 01:51:58 -05:00
• dbdefbbe54 Typed methods in tensor.py (#9356) Friedrich Carl EichenrothandGitHub 2025-03-06 01:34:18 +00:00
• 77f7ddf62a gfx10 correct ctl stack size (#9365) nimlgenandGitHub 2025-03-05 23:04:16 +03:00
• c8a74b11ed am: resize bar error msg (#9366) nimlgenandGitHub 2025-03-05 23:02:04 +03:00
• 9bd13de44c lower test_gemv_4096_16384 to 750 for red (#9367) nimlgenandGitHub 2025-03-05 22:44:48 +03:00
• b75f307234 amd: autogen ip bases (#9360) uuuvnandGitHub 2025-03-06 00:30:38 +05:00
• 2cb2fce8d9 lower test_gemm_8192 amd_tflops to 65 (#9364) chenyuandGitHub 2025-03-05 14:06:11 -05:00
• b3ac60ce53 Fix test_ops for tiny backend part 2 (#9358) 3 people 2025-03-06 00:08:40 +05:30
• c6d76770e4 Increase timeout on macos tests (#9362) uuuvnandGitHub 2025-03-05 23:04:16 +05:00
• cd9d74f7ea use am in training benchmarks (#9357) nimlgenandGitHub 2025-03-05 19:13:47 +03:00
• 2af129c078 bert corealize multiple outputs (#9359) chenyuandGitHub 2025-03-05 10:58:37 -05:00
• d550583657 am: flush hdp (#9354) nimlgenandGitHub 2025-03-05 16:31:59 +03:00
• f048256341 fix TORCH_DEBUG=1 sigsegv (#9352) Priyank PatelandGitHub 2025-03-05 01:24:53 -08:00
• ad72269f08 bert put eval copy and getting lr in jit (#9350) chenyuandGitHub 2025-03-04 20:57:03 -05:00
• 7576a1da23 hotfix: line count to 11500, lines for SQTT and AMDLLVM geohot 2025-03-05 09:21:18 +08:00
• 9eb45eb629 add a flag to skip bert train (#9349) chenyuandGitHub 2025-03-04 17:13:00 -05:00
• 14c88abf27 add some options to allreduce bench (#9348) nimlgenandGitHub 2025-03-04 23:46:36 +03:00
• 993ef42bd5 am: hdp cg (#9346) nimlgenandGitHub 2025-03-04 20:44:09 +03:00
• e301f21f63 CI ubuntu-20.04 -> ubuntu-22.04 (#9345) chenyuandGitHub 2025-03-04 11:39:12 -05:00
• 01f7a4fadc tinychat in browser, Part 2: model export (#9274) 3 people 2025-03-04 02:53:30 -05:00
• 94db8426cb Add consistent typing for python 3.10 in tensor.py (#9326) 5 people 2025-03-03 21:00:17 +00:00
• 019417743c ruff torch backend (#9341) chenyuandGitHub 2025-03-03 15:15:23 -05:00
• f9e4c638f1 torch_hook fixes (#9334) nimlgenandGitHub 2025-03-03 23:07:30 +03:00
• 40619a4bbc separate workflow for TINY_BACKEND=1 mnist (#9339) chenyuandGitHub 2025-03-03 13:05:24 -05:00
• bafa40fe12 Tiny backend test_ops fix part1 (#9338) Anish UmaleandGitHub 2025-03-03 23:06:51 +05:30
• 0d4ba7dd87 import tinygrad.frontend.torch (#9337) George HotzandGitHub 2025-03-04 00:15:29 +08:00
• b4028e48ae Torch Backend Refinement (#9327) Friedrich Carl EichenrothandGitHub 2025-03-03 15:24:02 +00:00
• 23084fd850 merge merge_views and remove_movement_ops [pr] (#9333) qazalandGitHub 2025-03-03 13:38:59 +02:00
• ece0a0f305 use empty for test instead of rand (#9332) George HotzandGitHub 2025-03-03 16:19:06 +08:00
• 27e899aea5 Make const moving not depend on the order (#9245) Sieds LyklesandGitHub 2025-03-03 09:09:27 +01:00
• 2cc4cb74f0 reorder binops (#9328) George HotzandGitHub 2025-03-03 14:58:18 +08:00
• 146eb73790 fix Tensor.view with a tuple arg (#9330) chenyuandGitHub 2025-03-02 23:35:23 -05:00