Commit Graph

  • 203a93363c Revert "after clean up of locals (#12813)" (#12814) George Hotz 2025-10-20 19:33:35 +08:00
  • 5e155d0947 Revert "after clean up of locals (#12813)" revert-12813-after_cleanups George Hotz 2025-10-20 19:33:22 +08:00
  • 5d0d3d7aac after clean up of locals (#12813) George Hotz 2025-10-20 19:24:24 +08:00
  • d1e2c393f8 after in sym, axis_letters in range (#12811) George Hotz 2025-10-20 18:54:37 +08:00
  • a8e4614436 remove REAL_SUBSTITUTE=0 and make it fast (#12809) Sieds Lykles 2025-10-20 12:44:20 +02:00
  • 1e93d19ee3 stable diffusion --fakeweights (#12810) Sieds Lykles 2025-10-20 12:41:06 +02:00
  • 890897553d this work? after_qol George Hotz 2025-10-20 18:35:45 +08:00
  • b5e36e3c6c nv: check if jitlink is avail (#12808) nimlgen 2025-10-20 18:13:16 +08:00
  • ec97cec952 this is better George Hotz 2025-10-20 18:07:34 +08:00
  • 154b6d5901 after in sym, axis_letters in range George Hotz 2025-10-20 17:37:52 +08:00
  • 45032b5c96 fixups two_stage_remove George Hotz 2025-10-20 17:30:07 +08:00
  • 6efc381825 two stage removal George Hotz 2025-10-20 16:21:39 +08:00
  • b8a9cce783 replace NOOP with AFTER in reg init (#12804) George Hotz 2025-10-20 15:34:32 +08:00
  • 12fd2c9c7b explicitly set ignore_indexing for schedule only (#12803) qazal 2025-10-20 13:11:57 +08:00
  • 734c99f722 viz: show indexing rewrites during run_rangeify (#12802) qazal 2025-10-20 12:37:03 +08:00
  • 2e9082e0bc after op (#12801) George Hotz 2025-10-20 12:27:56 +08:00
  • 339e6edb7d viz: ui prereqs for hierarchical rewrites (#12799) qazal 2025-10-20 12:15:15 +08:00
  • aecd51f54a start multioutput support George Hotz 2025-10-20 11:17:00 +08:00
  • 357dac8425 feat: allow tuple indexing on uops (#12797) wozeparrot 2025-10-19 19:11:05 -07:00
  • ba593f7b98 don't render index (#12796) George Hotz 2025-10-20 09:48:36 +08:00
  • cad3ada909 tinygpu: build with SIP off works George Hotz 2025-10-20 09:11:09 +08:00
  • 9cd35deae7 amd: fix alignment + pointers for aql over usb (#12793) nimlgen 2025-10-19 23:55:57 +08:00
  • 59784a5972 amd: ensure ts is written (#12794) nimlgen 2025-10-19 23:55:49 +08:00
  • 63a23dfe80 test step 0 in TestTrainingOnnxOps (#12790) chenyu 2025-10-19 09:15:49 -04:00
  • e8158afd4b update test_qlinear_add_round_half_to_even (#12789) chenyu 2025-10-19 08:47:27 -04:00
  • 1df9c7d7e7 reduce_collapse uses symbolic_flat (#12766) Sieds Lykles 2025-10-19 12:27:47 +02:00
  • fd6ef4801c rangeify uses symbolic_flat (#12786) Sieds Lykles 2025-10-19 12:27:14 +02:00
  • 89e7f2fa00 mmapeak: gfx1103 support George Hotz 2025-10-19 16:57:28 +08:00
  • 617614beb7 add mi350x support to mmapeak (#12784) George Hotz 2025-10-19 16:11:07 +08:00
  • c8ef4b60f6 viz: share match tracing and TINY device profiler (#12783) qazal 2025-10-19 14:30:07 +08:00
  • 350a4754a9 Update openpilot models (#12780) chenyu 2025-10-18 20:32:35 -04:00
  • 30ff84d050 update test_conv2d_ceildiv_edge_case (#12779) chenyu 2025-10-18 16:43:32 -04:00
  • 442218266d qcom: fix profiler (#12778) nimlgen 2025-10-19 01:27:59 +08:00
  • addc54b96c Simplify openpilot compile3.py (#12748) Harald Schäfer 2025-10-18 07:12:22 -07:00
  • 037f6e8fa0 qcom: ioctl for 7xx (#12777) nimlgen 2025-10-18 20:33:14 +08:00
  • 82f10cfe2e feat: assert on bufferview math (#12772) wozeparrot 2025-10-17 14:20:08 -07:00
  • fcdf4ab37e remove a contiguous in LARS (#12770) chenyu 2025-10-17 17:07:30 -04:00
  • 910d698b78 system: cleanup page sizes (#12771) nimlgen 2025-10-18 02:06:42 +08:00
  • 062a6d68d7 test flash attention backward (#12762) George Hotz 2025-10-17 23:15:59 +08:00
  • 70a1126830 ugh test_fa George Hotz 2025-10-17 23:14:36 +08:00
  • b22790eb0f fix tests George Hotz 2025-10-17 23:05:01 +08:00
  • dad778564c reset ending ranges George Hotz 2025-10-17 22:52:17 +08:00
  • c5617ed8cf Merge branch 'master' into test_fa George Hotz 2025-10-17 22:41:31 +08:00
  • 33025b99f6 small changes from fa backward (#12769) George Hotz 2025-10-17 22:41:18 +08:00
  • e0d0d4372d fix shape of m and v in onnx Adam with FUSE_OPTIM (#12768) chenyu 2025-10-17 10:32:41 -04:00
  • bd662bea67 viz: light up program runs (#12764) qazal 2025-10-17 19:33:18 +08:00
  • 28efb4395c multiout at every level George Hotz 2025-10-17 19:32:38 +08:00
  • bc9048ccca very big George Hotz 2025-10-17 19:11:19 +08:00
  • 7c80285fa8 render colors George Hotz 2025-10-17 18:58:44 +08:00
  • 05f69b48e9 end ranges George Hotz 2025-10-17 18:25:31 +08:00
  • 5fa053a5ee TODO: fix pcontig George Hotz 2025-10-17 17:51:36 +08:00
  • eb5070786a test flash attention backward George Hotz 2025-10-17 17:28:59 +08:00
  • c9a3464f76 those decimals never mattered (#12760) George Hotz 2025-10-17 17:16:24 +08:00
  • 206b46687b locals are different buffers no_decimals George Hotz 2025-10-17 17:04:29 +08:00
  • 78b2d76e3b real substitute fixes pcontig George Hotz 2025-10-17 17:01:26 +08:00
  • 0160f034d6 viz: show display name for copy runners (#12761) qazal 2025-10-17 16:59:51 +08:00
  • 4f7005f72a improve debug George Hotz 2025-10-17 16:45:54 +08:00
  • c2af5c806b this George Hotz 2025-10-17 16:35:25 +08:00
  • 253d32b065 viz: add metadata to buffer user list (#12758) qazal 2025-10-17 16:28:54 +08:00
  • 8d35780e1a those decimals never mattered George Hotz 2025-10-17 16:28:36 +08:00
  • 935a60db72 bring back partial contig and flash attention (#12756) George Hotz 2025-10-17 16:19:05 +08:00
  • f6bc620169 UOp.prod and UOp.sum methods (#12755) Sieds Lykles 2025-10-17 10:02:01 +02:00
  • d1bb5c0426 slightly flatter symbolic (#12757) Sieds Lykles 2025-10-17 09:58:45 +02:00
  • 5417e4b099 viz helper cleanups (#12754) qazal 2025-10-17 15:20:24 +08:00
  • 3196a7aae3 viz: pre reqs for lighting up programs (#12753) qazal 2025-10-17 15:03:21 +08:00
  • 978502be46 experiments with multi being range multi_range George Hotz 2025-10-17 14:11:55 +08:00
  • dfb8f9fc9e viz: annotate buffer mutability in the memory graph (#12750) qazal 2025-10-17 11:53:02 +08:00
  • 79c2f1ae26 remove reduce_rangless and replace with reduce_unparented (#12749) Sieds Lykles 2025-10-17 04:46:05 +02:00
  • 9561803cb0 fix assert in test_schedule (#12745) chenyu 2025-10-16 15:39:50 -04:00
  • 285534ce64 delete DONT_REALIZE_EXPAND and DONT_GROUP_REDUCES (#12744) chenyu 2025-10-16 14:11:33 -04:00
  • 98239f1156 few shapetracker cleanups (#12741) chenyu 2025-10-16 12:43:27 -04:00
  • 53478c741d relax ASSERT_MIN_STEP_TIME for space lab policy (#12742) chenyu 2025-10-16 11:40:36 -04:00
  • 5d209ee7ec onnx helper intermediate node output validation (#12740) geohotstan 2025-10-16 23:17:47 +08:00
  • bce2bc0465 Revert "use RTLD_GLOBAL on macos" (#12738) Christopher Milan 2025-10-16 10:07:21 -04:00
  • f34f26bca0 fix gpt2 with benchmark (#12736) chenyu 2025-10-16 09:55:20 -04:00
  • 55db1b0e0e reduce where that is cut from two sides (#12733) Sieds Lykles 2025-10-16 15:25:15 +02:00
  • cf9baeea61 Revert "nv: check if jitlink is avail (#12731)" (#12735) nimlgen 2025-10-16 20:41:49 +08:00
  • fabe7c9849 Revert "nv: check if jitlink is avail (#12731)" revert-12731-jitlink_init_check nimlgen 2025-10-16 20:41:14 +08:00
  • 8be7844b2e use apply uop for assign to fix assign metadata (#12732) George Hotz 2025-10-16 20:34:12 +08:00
  • 3aa2277b8f nv: usb4 (#12696) nimlgen 2025-10-16 20:11:19 +08:00
  • a069a45d14 nv: check if jitlink is avail (#12731) nimlgen 2025-10-16 19:58:50 +08:00
  • a498ec9c18 cleanup names of postrange + fast FUSE_OPTIM (#12730) George Hotz 2025-10-16 19:38:31 +08:00
  • 8f740e07ff no broadcasting/vectors in reduce collapse (#12729) Sieds Lykles 2025-10-16 13:22:57 +02:00
  • 533f18b22c viz: add trace data for inflight buffers (#12728) qazal 2025-10-16 19:15:03 +08:00
  • af4479c169 faster stable diffusion load (#12725) George Hotz 2025-10-16 18:31:59 +08:00
  • aef4a496b1 failing tests sd_load_simple George Hotz 2025-10-16 18:26:42 +08:00
  • 7383ab9b80 faster stable diffusion load George Hotz 2025-10-16 18:09:08 +08:00
  • e7c057d5dc system: alloc_sysmem return view (#12724) nimlgen 2025-10-16 17:55:01 +08:00
  • b86a33a312 ptx: support bw (#12722) nimlgen 2025-10-16 15:38:08 +08:00
  • b8cd66c7a2 nv: support all gb20x and small bar (#12721) nimlgen 2025-10-16 15:37:54 +08:00
  • 1d1e1d9d88 delete the ShapeTracker (#12720) George Hotz 2025-10-16 15:36:22 +08:00
  • 592e86f6f5 remove UOp.st (#12716) George Hotz 2025-10-16 14:44:09 +08:00
  • cc2dfe22f5 tinyfs: fetch file utility (#12719) wozeparrot 2025-10-15 23:38:56 -07:00
  • 3ed543f956 system: reorder funcs + barrier on macos (#12714) nimlgen 2025-10-16 14:38:01 +08:00
  • b77bdbbc62 viz: count unpickle in server startup time (#12715) qazal 2025-10-16 13:07:46 +08:00
  • 7c19db00f1 remove st from jit/split_reduceop (#12713) George Hotz 2025-10-16 12:50:58 +08:00
  • 069177c1be trace buffer producer and consumers (#12639) qazal 2025-10-16 11:11:31 +08:00
  • 4a151e7533 make xcode signing happy, waiting for entitlement (#12712) George Hotz 2025-10-16 10:20:34 +08:00
  • c3278e5622 clean up old tests (#12708) chenyu 2025-10-15 17:53:17 -04:00
  • b8cf35fb77 print macOS version in CI (#12705) chenyu 2025-10-15 15:05:33 -04:00