Commit Graph

  • 32d69d07d7 rangeify: enable multitensor TestBatchNorm (#12342) qazal 2025-09-30 06:05:00 +03:00
  • d55d829635 Lower index dtype spec fix (#12337) Sieds Lykles 2025-09-30 04:26:50 +02:00
  • c38f6ce140 unified_rewrite: use deque and dont add nodes to the stack multiple times (#12320) Sieds Lykles 2025-09-30 04:02:28 +02:00
  • c2689c505e Clip model updates for Stable Diffusion mlperf training (#12313) hooved 2025-09-29 21:50:14 -04:00
  • cdfa0f29fd add rendering to index (#12338) George Hotz 2025-09-30 09:18:05 +08:00
  • baf3b60cfb fix gpt2 on rangeify (#12335) George Hotz 2025-09-29 21:16:44 +10:00
  • 9513f025c5 apply multi before rangeify (#12298) qazal 2025-09-29 14:16:31 +03:00
  • b899392f30 fix llm app with rangeify (#12334) George Hotz 2025-09-29 20:42:44 +10:00
  • 7ae6898e31 better late bufferview (#12333) wozeparrot 2025-09-29 03:08:34 -07:00
  • 3291e00df7 fix efficientnet slowness on rangeify (#12332) George Hotz 2025-09-29 20:01:01 +10:00
  • 9d2f2b8e34 skip test_mean_half_precision_overflow (#12331) chenyu 2025-09-29 18:15:04 +09:00
  • 9915bcf2b4 remove no-op contiguous from rand (#12329) qazal 2025-09-29 11:53:16 +03:00
  • 76c87d81b3 delete test_backward_sum_acc_dtype (#12330) chenyu 2025-09-29 17:46:17 +09:00
  • fd2e4f2353 failing rng test (#12328) George Hotz 2025-09-29 18:06:45 +10:00
  • 29469577e8 tighten spec: fixup devectorizer types / rangeify (#12327) George Hotz 2025-09-29 17:41:11 +10:00
  • a982480512 feat: late to_bufferview (#12271) wozeparrot 2025-09-29 00:29:43 -07:00
  • e01a3eb59a rangeify whitespace cleanups [pr] (#12326) qazal 2025-09-29 10:04:51 +03:00
  • cf925d1ac5 remove metadata for rangeify codegen (#12325) George Hotz 2025-09-29 16:29:28 +10:00
  • b252f890da add support for SPEC=1 (#12322) George Hotz 2025-09-29 14:55:01 +10:00
  • 292cb6ae26 viz: 404 if the requested rewrite doesn't exist (#12323) qazal 2025-09-29 07:51:10 +03:00
  • 250cb10e8f rangeify permuted assign (#12299) qazal 2025-09-29 07:27:57 +03:00
  • ed90de6583 Revert "Bufferize early, fix "children not making progress" on big graphs (#1…" (#12318) Sieds Lykles 2025-09-28 19:10:21 +02:00
  • 8b4a963789 Revert "Bufferize early, fix "children not making progress" on big graphs (#1…" revert-12308-bufferize_early Sieds Lykles 2025-09-28 18:56:38 +02:00
  • 29f0886395 skip test_softmax_fusion tests if RANGEIFY==1 (#12310) Sieds Lykles 2025-09-27 05:57:40 +02:00
  • b98f1881ef dsp opt test has different axis number on rangeify (#12309) Sieds Lykles 2025-09-27 05:06:11 +02:00
  • 6f1cf717de Bufferize early, fix "children not making progress" on big graphs (#12308) Sieds Lykles 2025-09-27 04:17:15 +02:00
  • 0104b16b9b rangeify: fix empty tags in reshapes (#12307) qazal 2025-09-26 16:32:48 +03:00
  • f5eb46a3d9 fix limit buf metal on non rangeify (#12303) nimlgen 2025-09-26 11:06:28 +03:00
  • 8b2e0930d7 rangeify: enable passing multi test (#12301) qazal 2025-09-26 08:31:13 +03:00
  • 74411984fc Rangeify IMAGE (#12304) Sieds Lykles 2025-09-26 07:21:02 +02:00
  • d2cd269e28 fix: try close mmap (#12306) wozeparrot 2025-09-25 20:54:27 -07:00
  • 17cec8d645 RANGEIFY winograd test (#12297) chenyu 2025-09-24 23:42:32 -04:00
  • 476a2a0a96 test_qcom: update (#12293) nimlgen 2025-09-24 21:45:58 +03:00
  • 38ecefaacb RANGEIFY=1 allreduce (#12260) qazal 2025-09-24 18:13:08 +03:00
  • 0e778296be rangeify: refactor const folding (#12291) qazal 2025-09-24 17:58:39 +03:00
  • 6c9d8c7e41 rangeify: simplify noop copy (#12289) qazal 2025-09-24 17:01:23 +03:00
  • 1400ce105f rangeify: fix sharding (#12288) qazal 2025-09-24 14:33:56 +03:00
  • 154c865966 rangeify: fix ram usage in multi (#12286) qazal 2025-09-24 13:48:58 +03:00
  • e8945c74de fix infinite symbolic loop with VCONST (#12285) Sieds Lykles 2025-09-24 07:06:22 +02:00
  • 45c7252aed Better div nesting 2 (#11812) Sieds Lykles 2025-09-24 04:50:26 +02:00
  • 6146c64d81 lower the invalid gate last (#12164) Sieds Lykles 2025-09-24 04:27:35 +02:00
  • ad7c8c21ea rangeify: INDEX doesn't passthrough MSELECT (#12279) qazal 2025-09-23 21:36:50 +03:00
  • 02a7b7fe48 rangeify: fix test_setitem (#12269) nimlgen 2025-09-23 20:42:36 +03:00
  • 2f145a98e0 rangeify: fix contiguous multi (#12278) qazal 2025-09-23 20:05:29 +03:00
  • 5f4eeb054c rangeify: passes now (#12277) nimlgen 2025-09-23 18:46:49 +03:00
  • 680ce54dd4 add types to replace_dnum (#12276) qazal 2025-09-23 14:43:04 +03:00
  • fffce0a6b4 use more no_range in simplify [pr] (#12275) chenyu 2025-09-23 02:33:56 -04:00
  • 51b88b2265 process replay tests in rangeify (#12274) chenyu 2025-09-23 01:30:06 -04:00
  • b54cb272d0 move test_qcom to test/device (#12272) chenyu 2025-09-22 21:07:10 -04:00
  • d21e34e617 enable test_sum_twice (#12270) Sieds Lykles 2025-09-23 00:57:29 +02:00
  • 5a4b244e6b Check for group inside another reduce (#12268) Sieds Lykles 2025-09-23 00:32:41 +02:00
  • a6fd96f620 rangeify: don't tag movement ops (#12267) qazal 2025-09-22 16:40:17 +03:00
  • b03ceb806e move test_sample to test_randomness (#12266) chenyu 2025-09-21 21:11:32 -04:00
  • 25e0b725d1 cleanup section 0 rangeify (#12264) qazal 2025-09-22 00:30:44 +03:00
  • 1aba668a37 cleanup buffer_view matcher (#12263) qazal 2025-09-21 23:45:48 +03:00
  • b53a266254 rangeify: fix test_optim (#12262) nimlgen 2025-09-21 18:08:35 +03:00
  • 461e9becec srender UOp in movement op arg (#12261) qazal 2025-09-21 13:55:45 +03:00
  • 9569fdfa36 use str for AxisType and AddrSpace __repr__ (#12252) Sieds Lykles 2025-09-21 05:24:41 +02:00
  • 8365c28cd5 viz: put a limit of brightness scale (#12259) qazal 2025-09-20 18:52:55 +03:00
  • 4762a24022 test_free_intermediates force buffers (#12255) nimlgen 2025-09-20 18:14:39 +03:00
  • 57c7e0a8f8 RANGEIFY=1 test_jit (#12254) qazal 2025-09-20 17:34:32 +03:00
  • 393c6b236c test case to sum twice in different order (#12253) chenyu 2025-09-20 10:11:57 -04:00
  • 4756971c88 skip test_bf16_disk_write_read on CL=1 (#12256) qazal 2025-09-20 17:11:06 +03:00
  • 5e794be8af tighter spec for RANGE (#12250) chenyu 2025-09-20 07:59:50 -04:00
  • 73c8dae60d add missing remove_blockend case (#12251) Sieds Lykles 2025-09-20 06:29:19 +02:00
  • dc4dd898b7 fix: close mmap (#12249) wozeparrot 2025-09-19 14:09:12 -07:00
  • bb1f376ae6 profile z3 (#12248) Sieds Lykles 2025-09-19 22:52:06 +02:00
  • 7e06d3ebba enable test_symbolic_jit (#12245) Sieds Lykles 2025-09-19 20:23:42 +02:00
  • bb59eed82f rangeify: don't tag consts, they are global (#12247) qazal 2025-09-19 15:25:03 +03:00
  • cc038b31b6 Shrink instead of reshape to unregister symbolic (#12241) Sieds Lykles 2025-09-19 06:04:35 +02:00
  • a531a649fb test_resize_upsample_scales_cubic_align_corners_cpu is fixed (#12244) chenyu 2025-09-18 20:55:26 -04:00
  • 8d703a6369 z3 xor doesnt use bitcast (#12243) Sieds Lykles 2025-09-19 00:31:44 +02:00
  • 0dad6cc518 good RANGEIFY kernel counts in external_test_opt (#12242) chenyu 2025-09-18 17:58:54 -04:00
  • cff1065f5e test CL=1 RANGEIFY=1 onnx (#12240) chenyu 2025-09-18 16:49:46 -04:00
  • ef05178855 fix 0//0 infinite rewrite in rangeify onnx (#12239) Sieds Lykles 2025-09-18 21:59:50 +02:00
  • 87707ef0b8 unify range_start [pr] (#12236) chenyu 2025-09-18 13:52:54 -04:00
  • 825f148469 rangeify: fix copy size mismatch errs (#12232) qazal 2025-09-18 18:23:32 +03:00
  • f82b16a0e9 RANGEIFY test_tensor (#12235) chenyu 2025-09-18 10:35:43 -04:00
  • 7487c13b61 truncate_fp16 -> float_to_fp16 (#12234) chenyu 2025-09-18 09:48:27 -04:00
  • 54c15d74a4 python float8 support (#11960) b1tg 2025-09-18 21:17:09 +08:00
  • dbbc261075 rangeify: fix COPY simplifier (#12233) qazal 2025-09-18 14:35:33 +03:00
  • f1108f1cbe Enable test_symbolic_ops on rangeify (#12230) Sieds Lykles 2025-09-18 02:12:36 +02:00
  • 812f485cd7 Enable threefry_doesnt_use_long test on rangeify (#12229) Sieds Lykles 2025-09-18 01:58:34 +02:00
  • 3c5b8bf50c am: bump fw to rocm7 (#12226) nimlgen 2025-09-17 21:20:22 +03:00
  • 525f80e0d2 rangeify: enable putting consts back in the tensor graph (#12225) qazal 2025-09-17 19:45:04 +03:00
  • edffc246ed MUL in reduce_unparented (#12223) chenyu 2025-09-17 11:56:39 -04:00
  • 7733c217c5 remove spam comments in test_schedule (#12224) qazal 2025-09-17 18:24:55 +03:00
  • d917895569 map out rangeify errors in test_schedule (#12211) qazal 2025-09-17 09:10:28 +03:00
  • 158506b91e Upgrade some divmod folding for symbolic divs (#12216) Sieds Lykles 2025-09-17 03:00:50 +02:00
  • 328bfe6b9b fix map_expand for symbolic shapes (#12218) Sieds Lykles 2025-09-17 01:20:18 +02:00
  • 5b12764b83 add arange cat arange test (#12217) chenyu 2025-09-16 17:12:32 -04:00
  • 53655a4ee5 cuda: cleanup old comment (#12215) nimlgen 2025-09-16 23:11:32 +03:00
  • 6b808c5fe6 update TestSymbolicJit.test_plus1_pad (#12214) chenyu 2025-09-16 15:57:50 -04:00
  • 2a72b00679 Add test for 2D tensor indexing in setitem (#12193) Shun Usami 2025-09-16 11:57:25 -07:00
  • c7b03457d7 Revert "Revert "more llvm intrinsics (#11961)" (#12194)" (#12195) chenyu 2025-09-16 14:55:31 -04:00
  • 494bb12500 skip slow cifar bf16 on red benchmark (#12213) chenyu 2025-09-16 14:55:01 -04:00
  • 419e997187 increase benchmark timeout (#12212) chenyu 2025-09-16 14:09:02 -04:00
  • 84d2d047ea Tensor.pad_to and Tensor.shrink_to (#12210) chenyu 2025-09-16 12:24:55 -04:00
  • 122a50fe8c assert kernel count (#12205) qazal 2025-09-16 14:24:39 +03:00
  • a8140a5c7f start support for adreno 830 qcom_830 George Hotz 2025-09-16 14:05:31 +08:00