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

• 32ed128598 fixing transformer training bug (#9877) Nishant RajadhyakshaandGitHub 2025-04-13 16:34:20 -07:00
• bd5939514d clean up a few patterns [pr] (#9873) George HotzandGitHub 2025-04-14 03:19:37 +08:00
• 78a6af3da7 Use $CUDA_PATH/include for CUDA headers (#9858) Alexey ZaytsevandGitHub 2025-04-13 12:20:19 -03:00
• e2a40fb523 update bert mi300x script (#9872) chenyuandGitHub 2025-04-13 10:07:36 -04:00
• e201bc3e93 process replay kernel asts in toposort order [pr] (#9869) qazalandGitHub 2025-04-13 17:20:34 +08:00
• 7191f88551 add asserts for KERNEL op ast [pr] (#9868) qazalandGitHub 2025-04-13 16:50:18 +08:00
• 5ee9c343e6 add device to NullRenderer [pr] (#9867) qazalandGitHub 2025-04-13 13:17:16 +08:00
• 2793cca9a6 RetinaNet MLPerf (#8385) Francis LataandGitHub 2025-04-12 22:11:51 -04:00
• 23b67f532c amd: minor comments and readme updates (#9865) nimlgenandGitHub 2025-04-12 23:24:05 +03:00
• 7c466c24f7 am_smi: refactor to support arches (#9864) nimlgenandGitHub 2025-04-12 20:37:01 +03:00
• a9430b4118 am: fix metrics table for smu14_0_2 (#9863) nimlgenandGitHub 2025-04-12 19:07:22 +03:00
• 3bce5ad2b4 clang should not emit the .comment section (#9859) Alexey ZaytsevandGitHub 2025-04-11 23:59:11 -03:00
• 7dda6aae7d Skip CLOUD in external_test_example (#9857) Alexey ZaytsevandGitHub 2025-04-11 23:17:44 -03:00
• 7919bb4f8a amd: do not use log2 (#9852) nimlgenandGitHub 2025-04-11 19:53:06 +03:00
• ada0f67d3d am: fix speed of ring copies (#9854) nimlgenandGitHub 2025-04-11 17:28:06 +03:00
• 4aab16ca6a bert script cleanup and assert nan loss (#9851) chenyuandGitHub 2025-04-11 05:41:49 -04:00
• ad677f8e55 create_ast cleanups from kernelize [pr] (#9849) qazalandGitHub 2025-04-11 16:10:21 +08:00
• cbc5e7ed45 unbind variables when creating ScheduleItems [pr] (#9846) qazalandGitHub 2025-04-11 15:23:53 +08:00
• 6896197978 relax ATOL for TC half tests more (#9847) chenyuandGitHub 2025-04-11 03:20:22 -04:00
• dd52951dd0 fix single kernel softmax with cast (#9842) George HotzandGitHub 2025-04-11 12:12:02 +08:00
• 8c6299bced move hand_coded_optimizations to heuristic.py [pr] (#9844) chenyuandGitHub 2025-04-10 23:40:16 -04:00
• e0ec8be37d use CPU for test_schedule_ring (#9843) chenyuandGitHub 2025-04-10 23:20:53 -04:00
• 7045920786 give _apply_map_to_tensors substitutes name [pr] (#9840) qazalandGitHub 2025-04-11 10:38:57 +08:00
• 40ef2f2857 add ast fixup stage to tensor_map [pr] (#9839) qazalandGitHub 2025-04-11 09:24:01 +08:00
• fbc6aa53d4 script for local process_replay + fix viz name [pr] (#9837) qazalandGitHub 2025-04-11 00:39:18 +08:00
• a35b475d18 fix am driver for gfx1201 (#9836) b1tgandGitHub 2025-04-11 00:33:02 +08:00
• 16956b79de canonicalize Device.DEFAULT (#9835) qazalandGitHub 2025-04-10 23:02:11 +08:00
• f666dd14eb fix get reduce contraction with test (#9834) George HotzandGitHub 2025-04-10 22:24:21 +08:00
• c3fa470852 hotfix: remove tracebacklimit, it persists if you catch the exception and made webgpu flaky geohot 2025-04-10 20:29:25 +08:00
• 7fa5f29582 add test_embedding to test_softmax_fusion (#9832) chenyuandGitHub 2025-04-10 08:25:34 -04:00
• 995d20673a increase bert TRAIN_STEPS for mi300x (#9833) chenyuandGitHub 2025-04-10 08:25:09 -04:00
• 25e2a3cf5d hotfix: fix get_contraction_with_reduce geohot 2025-04-10 20:18:19 +08:00
• 53f0b2aad7 fix infinite loop in flash attention (#9827) George HotzandGitHub 2025-04-10 20:06:44 +08:00
• 16afe04f45 move process replay to grouper (#9830) qazalandGitHub 2025-04-10 18:27:42 +08:00
• c8f47c1d07 not_support_multi_device helper (#9831) chenyuandGitHub 2025-04-10 05:25:29 -04:00
• 817746b30e add contiguous to EmbeddingBert output (#9829) chenyuandGitHub 2025-04-10 04:31:19 -04:00
• fd4f06e623 kernelize prereqs [pr] (#9811) qazalandGitHub 2025-04-10 15:22:20 +08:00
• c462162db8 update benchmark bert scripts with BS and ACC_DTYPE (#9826) chenyuandGitHub 2025-04-10 02:06:02 -04:00
• 498a2bf738 add err handling tests to viz + cleanups (#9825) qazalandGitHub 2025-04-10 14:05:05 +08:00
• a0b72f066a don't free intermediate for bert mi300x (#9824) chenyuandGitHub 2025-04-10 01:48:34 -04:00
• 566e389585 more relaxed ATOL for HALF=1 simple_matmul test (#9823) chenyuandGitHub 2025-04-10 00:46:16 -04:00
• eb2e59db42 RetinaNet model type annotations and loss functions (#9822) Francis LataandGitHub 2025-04-10 00:31:37 -04:00
• 06a928b341 higher ATOL for half input TC test (#9821) chenyuandGitHub 2025-04-09 23:57:25 -04:00
• 7bb36d71b2 remove openimages iterate (#9820) Francis LataandGitHub 2025-04-09 22:54:12 -04:00
• 2e1002e179 EVAL_BS=96 and BEAM=3 for bert green (#9819) chenyuandGitHub 2025-04-09 22:37:27 -04:00
• 3ee317ffed Fix kfd autogen and verify it in ci (#9818) uuuvnandGitHub 2025-04-10 06:53:42 +05:00
• d7330ea6ad amd: refactor sqtt into sep functions (#9816) nimlgenandGitHub 2025-04-10 00:39:45 +03:00
• 0ca98b9f20 amd: gfx9 use cache ctrls in acquire_mem (#9815) nimlgenandGitHub 2025-04-09 20:17:02 +03:00
• fce432d2e3 Ops.FUSE makes softmax a single kernel (#9808) George HotzandGitHub 2025-04-09 22:56:28 +08:00
• 1798ce7e52 amd: faster xcc sync (#9783) nimlgenandGitHub 2025-04-09 15:56:50 +03:00
• 3bd992dc95 multi stage graph_rewrite_map (#9803) qazalandGitHub 2025-04-09 15:59:45 +08:00
• 57f4bc3fbb add numpy to setup linting (#9806) chenyuandGitHub 2025-04-09 03:47:03 -04:00
• bf769fa5c5 label ranges with their number (#9805) George HotzandGitHub 2025-04-09 14:31:18 +08:00
• c5db5b83b9 add SHOULD_USE_TC=1 check to simple_matmul (#9802) chenyuandGitHub 2025-04-09 02:24:42 -04:00
• f27dbc8c35 becomes_map cleanups [pr] (#9790) qazalandGitHub 2025-04-09 14:11:53 +08:00
• 7d2349c827 track_rewrites in scheduler [pr] (#9801) qazalandGitHub 2025-04-09 12:48:14 +08:00
• bb18adb0d5 reduce with a mul chain (#9799) George HotzandGitHub 2025-04-09 12:42:32 +08:00
• 78caf55154 Revert "FP8 support on NVIDIA (#8631)" geohot 2025-04-09 12:27:41 +08:00
• d1505137ad Revert "move TestOpsFp8s skipTest (#9797)" geohot 2025-04-09 12:27:40 +08:00
• 14928fecff Revert "fix TF32 tensor core dropped in tc_sm89 (#9798)" geohot 2025-04-09 12:27:39 +08:00
• 1ed4eae510 hotfix: don't add shape to SINK viz node (#9800) qazalandGitHub 2025-04-09 12:04:33 +08:00
• 7c9a96824f fix TF32 tensor core dropped in tc_sm89 (#9798) chenyuandGitHub 2025-04-08 23:20:50 -04:00
• a3aaf92b21 move TestOpsFp8s skipTest (#9797) chenyuandGitHub 2025-04-08 22:44:07 -04:00
• 2c8e4ea865 FP8 support on NVIDIA (#8631) 5 people 2025-04-09 03:54:04 +02:00
• 5d85765327 types for WebGPU runtime (#9791) hoovedandGitHub 2025-04-08 15:52:17 -04:00
• 4c8582a7ce pipe allow_test_size in _time_program (#9789) chenyuandGitHub 2025-04-08 09:07:40 -04:00
• 8fe83385ec add system json for mi300x mlperf (#9786) chenyuandGitHub 2025-04-08 06:36:44 -04:00
• 4a807ee952 remove duplicated z3-solver in setup.py (#9787) chenyuandGitHub 2025-04-08 06:12:58 -04:00
• 21e872df44 remove consts from sched_sink [pr] (#9782) qazalandGitHub 2025-04-08 16:08:24 +08:00
• f13e9cf2d9 move view_left to grouper.py + tiny reorders [pr] (#9780) qazalandGitHub 2025-04-08 15:39:28 +08:00
• 7a28133b37 failed test for single softmax backward (#9778) chenyuandGitHub 2025-04-08 02:36:32 -04:00
• fefee5d3ab single kernel softmax (#9776) George HotzandGitHub 2025-04-08 12:35:48 +08:00
• 9963bb51e0 grouper tests cleanups [pr] (#9777) qazalandGitHub 2025-04-08 12:33:11 +08:00
• 4cc7422769 use AM driver in bert mlperf (#9775) chenyuandGitHub 2025-04-07 23:40:27 -04:00
• db22094d35 hotfix: update softmax fusion test geohot 2025-04-08 11:22:12 +08:00
• f8fe15e64e move BoxCoder to mlperf helpers (#9773) Francis LataandGitHub 2025-04-07 20:27:06 -04:00
• bb7922b95f Vectorize Transcendental Regression Tests (#9753) Eitan TurokandGitHub 2025-04-07 20:27:39 +03:00
• 7c4a739fe4 full script for bert mi300x (#9772) chenyuandGitHub 2025-04-07 11:41:31 -04:00
• 07d1aefaf4 fast idiv (#9755) 3 people 2025-04-07 14:32:24 +02:00
• fa888ee077 minor test cleanups (#9770) nimlgenandGitHub 2025-04-07 15:29:12 +03:00
• 3069ebfad1 use BERT_LAYERS=2 in bert init (#9769) chenyuandGitHub 2025-04-07 07:46:37 -04:00
• 891322fd51 split into grouper.py (#9768) qazalandGitHub 2025-04-07 18:40:59 +08:00
• 219b8c9e8b return becomes_map in scheduler [pr] (#9766) qazalandGitHub 2025-04-07 17:36:23 +08:00
• 6306dea6e2 add a graph_rewrite pass for creating asts [pr] (#9765) qazalandGitHub 2025-04-07 16:32:11 +08:00
• 07eea567d4 reorder tensor_map and grouper parts [pr] (#9764) qazalandGitHub 2025-04-07 15:36:13 +08:00
• 8ddb1357c0 fix UPat.location after pickle (#9763) qazalandGitHub 2025-04-07 15:16:42 +08:00
• 4cd27aa0e6 hotfix: viz recenter and unlimited zoom (#9760) qazalandGitHub 2025-04-07 14:38:03 +08:00
• d0dace4306 update doc for permute to 3d tensor (#9758) chenyuandGitHub 2025-04-07 00:38:05 -04:00
• b190d85ad7 benchmark script bert softmax (#9759) chenyuandGitHub 2025-04-07 00:31:18 -04:00
• 58785181a8 AMD bf16xf32 TC (#9717) Ignacio SicaandGitHub 2025-04-07 11:41:04 +08:00
• 43e4565148 weighted linear in external_benchmark_bert_matmuls (#9757) chenyuandGitHub 2025-04-06 23:35:42 -04:00
• 28e06d2d44 minor cleanups from patternmatcher [pr] (#9756) George HotzandGitHub 2025-04-07 11:28:14 +08:00
• 1ce4912770 viz profiler ui (#9664) qazalandGitHub 2025-04-07 00:30:17 +08:00
• 8a585dc5c1 benchmark script for matmuls in bert (#9752) chenyuandGitHub 2025-04-06 07:34:25 -04:00
• 139999c6d7 map viz files + query params cleanup [pr] (#9754) qazalandGitHub 2025-04-06 16:20:00 +08:00
• 71b8890dd6 use validation dataloader inside retinanet eval (#9747) Francis LataandGitHub 2025-04-05 16:46:55 -04:00
• 5f7c79676f jit: prune independent copies (#9749) nimlgenandGitHub 2025-04-05 20:50:28 +03:00
• c2573b247c jit: rename optimize_weights -> replan_buffers_memory_layout (#9751) nimlgenandGitHub 2025-04-05 20:35:15 +03:00
• 493fb315b1 fix RDNA2 support (#9700) uuuvnandGitHub 2025-04-05 20:42:47 +05:00
• 5a04f4d4ba revert bert hparams for green and red (#9744) chenyuandGitHub 2025-04-05 07:38:01 -04:00