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

• 41a179f542 fix test_xlm_roberta_large (#14564) chenyuandGitHub 2026-02-05 14:56:06 -05:00
• aa9dc50577 dtype decomps don't require bitshifts (#14542) sirhcmandGitHub 2026-02-05 11:42:30 -08:00
• b47397ab17 list ml_dtypes as dependency for DSP (#14562) sirhcmandGitHub 2026-02-05 11:27:50 -08:00
• 2b47a9a1b5 skip test_xlm_roberta_large (#14563) chenyuandGitHub 2026-02-05 14:00:24 -05:00
• 42c18da88a add Ops asserts in toposort sched_sink [pr] (#14561) chenyuandGitHub 2026-02-05 12:40:02 -05:00
• 483bba4f05 nv: use prof_exec_counter (#14559) nimlgenandGitHub 2026-02-05 19:00:14 +03:00
• 190042358f llama: faster bf16 matmul / rope backward (#14558) qazalandGitHub 2026-02-05 09:57:25 -05:00
• b398335f62 assembly/amd: fix saturation in python remu (#14557) 3 people 2026-02-05 18:35:57 +08:00
• c1ea6687e5 fa: simpler is faster (#14548) wozeparrotandGitHub 2026-02-05 01:13:17 -08:00
• 43e7eda4e7 grad_b uses custom gemm (#14550) 4 people 2026-02-05 14:22:27 +08:00
• f9cfb64cd9 test asm_gemm in CI (#14551) qazalandGitHub 2026-02-04 23:32:22 -05:00
• c0ca7f9c51 use more UOp.sum and UOp.prod [pr] (#14549) chenyuandGitHub 2026-02-04 22:05:20 -05:00
• e8dace41b6 clean up UOp.vars [pr] (#14547) chenyuandGitHub 2026-02-04 20:52:25 -05:00
• 232848d086 PYTHONREMU: VOP3P integer operations with constants don't cast to fp16 (#14546) sirhcmandGitHub 2026-02-04 17:10:59 -08:00
• 2966619834 feat: llama uses enable_gqa during training (#14545) wozeparrotandGitHub 2026-02-04 16:22:31 -08:00
• 664f1bf76d minor ops/jit cleanups [pr] (#14543) chenyuandGitHub 2026-02-04 17:21:34 -05:00
• 03d0fa9c3f merge as_buf into buf_uop [pr] (#14541) chenyuandGitHub 2026-02-04 16:32:23 -05:00
• 43ef24a8af remove buf_target [pr] (#14540) chenyuandGitHub 2026-02-04 15:03:47 -05:00
• 8b7343b950 clean up is_realized [pr] (#14538) chenyuandGitHub 2026-02-04 14:24:10 -05:00
• 5338ce6b74 test S_PACK in extra/assembly/amd/test/hw (#14537) sirhcmandGitHub 2026-02-04 11:17:16 -08:00
• 9052db678f remove allow_shape_mismatch in Tensor.replace (#14536) chenyuandGitHub 2026-02-04 12:38:18 -05:00
• ec2b6bbda8 hcq: update signal logic (#14531) nimlgenandGitHub 2026-02-04 19:32:56 +03:00
• 62786d488a am: mi3xx perf (#14529) nimlgenandGitHub 2026-02-04 19:32:43 +03:00
• d57d24c7d4 Buffer.as_buffer -> Buffer.as_memoryview [pr] (#14535) chenyuandGitHub 2026-02-04 11:31:11 -05:00
• 024f57ecf5 jit input_buffers cleanup [pr] (#14532) chenyuandGitHub 2026-02-04 10:14:38 -05:00
• 67f91e897b UOp.is_contiguous -> UOp.has_buffer_identity [pr] (#14530) chenyuandGitHub 2026-02-04 09:21:26 -05:00
• fb9df1e031 pretty print binary (#14520) George HotzandGitHub 2026-02-04 18:04:35 +08:00
• 8c3c026d86 decomp float16 to float32 (#14417) sirhcmandGitHub 2026-02-03 22:37:47 -08:00
• ecbce5269e PYTHONREMU properly supports S_PACK_LL_B32_B16 (#14527) sirhcmandGitHub 2026-02-03 20:45:33 -08:00
• 720c9597a9 feat: llama uses is_causal on sdpa during training (#14528) wozeparrotandGitHub 2026-02-03 20:24:30 -08:00
• 9c2fc118ef relax setitem target check (#14526) chenyuandGitHub 2026-02-03 22:32:49 -05:00
• d1bfbe9ce3 isolate slow llama gemm (#14525) qazalandGitHub 2026-02-03 22:20:10 -05:00
• 2f55005ad9 qcom: sync cpu cache when from_blob (#14518) nimlgenandGitHub 2026-02-03 21:51:03 +03:00
• ee9d6a1f36 remove DEFINE_VAR in to_define_global [pr] (#14522) chenyuandGitHub 2026-02-03 10:12:33 -05:00
• af4c74bb41 delete extra cast (#14517) Nino RisteskiandGitHub 2026-02-03 14:29:04 +01:00
• 9d1e9e643e removed a duplicated remove_bufferize rule [pr] (#14519) chenyuandGitHub 2026-02-03 08:28:07 -05:00
• d59e6e7a37 move more tests to test/null, split some existing ones (#14512) George HotzandGitHub 2026-02-03 20:20:20 +08:00
• a98c53769a ASM_GEMM=1 runs the UOp gemm on non cdna (#14516) qazalandGitHub 2026-02-03 06:42:02 -05:00
• 5c1d21349e viz: profiler command line tool (#14515) qazalandGitHub 2026-02-03 05:51:25 -05:00
• dd2de4f838 rename all DEFINE_GLOBAL to PARAM (#14511) George HotzandGitHub 2026-02-03 15:09:38 +08:00
• dc77b3318b move files that pass with NULL=1 to test/null (#14508) George HotzandGitHub 2026-02-03 13:52:36 +08:00
• 888819ee09 call autodiff gradient (#14510) George HotzandGitHub 2026-02-03 13:51:02 +08:00
• bbcd3d67a3 fa: faster (#14453) wozeparrotandGitHub 2026-02-02 21:34:17 -08:00
• e579613b90 IR3 has aux (#14509) sirhcmandGitHub 2026-02-02 20:46:41 -08:00
• 85c7b23160 add pytest -nauto to benchmark for mac (#14458) George HotzandGitHub 2026-02-03 12:26:09 +08:00
• a5d7eb37db IR3 works on versions earlier than 3.14 (#14507) sirhcmandGitHub 2026-02-02 20:10:19 -08:00
• 33c886cafa disable copyout on NULL backend by default (#14506) George HotzandGitHub 2026-02-03 11:57:47 +08:00
• 3c5845e8a5 remove cut_store_range (#14505) chenyuandGitHub 2026-02-02 21:58:36 -05:00
• 4f2e7aed24 fix multiple REDUCE on same RANGE (#14504) chenyuandGitHub 2026-02-02 20:42:09 -05:00
• 93c41a78fa clean up NOOP [pr] (#14503) chenyuandGitHub 2026-02-02 19:46:45 -05:00
• 66d2b02f11 delete files that depends on extra.optimization.helpers (#14499) chenyuandGitHub 2026-02-02 13:33:33 -05:00
• ec0398fceb test amd gpu crashes (#14459) George HotzandGitHub 2026-02-02 23:57:47 +08:00
• 6e4238c016 amd: recovery (#14461) nimlgenandGitHub 2026-02-02 18:57:35 +03:00
• 61ca19ff24 after with empty src is self [pr] (#14496) chenyuandGitHub 2026-02-02 10:19:05 -05:00
• 6e958dbfd4 assembly/amd: add RDNA4 support to emulator (#14341) George HotzandGitHub 2026-02-02 21:35:59 +08:00
• a908f447d5 remove disk special case in mstack_early_shrink [pr] (#14494) chenyuandGitHub 2026-02-02 08:34:45 -05:00
• 965940dd00 sqtt: update examples after event field change (#14493) qazalandGitHub 2026-02-02 07:39:48 -05:00
• 965149a46d assembly/amd: add ds perm instructions (#14486) George HotzandGitHub 2026-02-02 16:02:00 +08:00
• 1746d1f997 remove SPEC=0 context in custom_kernel tests, pyrender always skips it (#14489) qazalandGitHub 2026-02-02 02:32:01 -05:00
• d4007f36e0 remove DEFINE_GLOBAL (it is PARAM now) (#14488) George HotzandGitHub 2026-02-02 14:56:37 +08:00
• 6c487656f9 viz: kernel metadata from rodata entry (#14487) qazalandGitHub 2026-02-02 01:41:42 -05:00
• d75a1b0d5a usbgpu: use BOT interface for patch.py (#13644) Robbe DerksandGitHub 2026-02-02 04:54:46 +01:00
• 2931b52875 skip autogen if MTLCompiler is loaded (#14466) sirhcmandGitHub 2026-02-01 19:12:27 -08:00
• 9a32d6e090 add depth limit for SPEC=2 (#14485) George HotzandGitHub 2026-02-02 10:43:28 +08:00
• 368a692e1a make SPEC=2 work for everything (#14476) George HotzandGitHub 2026-02-02 10:30:56 +08:00
• ea1f1d2b9d test_assign_to_bitcast_view (#14483) chenyuandGitHub 2026-02-01 16:46:04 -05:00
• 6deeccc192 fix RING with single dest (#14482) chenyuandGitHub 2026-02-01 12:14:46 -05:00
• 3ff390159b don't implicitly change dtype in assign (#14481) chenyuandGitHub 2026-02-01 11:48:54 -05:00
• 2111762a48 failed test case for RING output device (#14191) 3 people 2026-02-01 11:48:43 -05:00
• 02afae04f4 atol in test_call_gemm (#14480) chenyuandGitHub 2026-02-01 11:24:58 -05:00
• 5705398a1f assign cleanup [pr] (#14479) chenyuandGitHub 2026-02-01 09:10:22 -05:00
• da500dbe06 simplify late_buffer_view [pr] (#14478) chenyuandGitHub 2026-01-31 22:38:40 -05:00
• b4f96301e0 remove unused rules [pr] (#14477) chenyuandGitHub 2026-01-31 21:29:30 -05:00
• 54e78dbec8 viz: remove hardcoded strings in cfg tests (#14462) qazalandGitHub 2026-01-31 19:30:43 -05:00
• 5d38db9da6 generic bitcast assign (#14474) chenyuandGitHub 2026-01-31 17:29:20 -05:00
• b38fc43b07 assert assign dtype mismatch for disk [pr] (#14473) chenyuandGitHub 2026-01-31 17:08:54 -05:00
• ced886f26c failed test case for assign into bitcast (#14469) chenyuandGitHub 2026-01-31 14:26:47 -05:00
• 81eee5b30a unused spec [pr] (#14468) chenyuandGitHub 2026-01-31 13:53:16 -05:00
• f873c7b6c5 amd: fetch_name is file_name (#14465) nimlgenandGitHub 2026-01-31 20:11:07 +03:00
• c765641215 remove unused allow_any_len [pr] (#14464) chenyuandGitHub 2026-01-31 11:05:42 -05:00
• b4f5a51ebb move tests to unit (#14463) chenyuandGitHub 2026-01-31 10:49:31 -05:00
• 616e9c1483 CDNA assembly gemm in tensor.py with flag (#14310) qazalandGitHub 2026-01-31 08:34:14 -05:00
• 55f806b713 tighter late_buffer_view match [pr] (#14456) chenyuandGitHub 2026-01-31 07:28:26 -05:00
• d69bc5aa1a make DEV=NULL EMULATE=AMD amd_asm_matmul run (#14460) qazalandGitHub 2026-01-31 06:45:24 -05:00
• 4976544bf9 multi ram usage tests on the NULL device (#14457) qazalandGitHub 2026-01-31 00:14:53 -05:00
• 99b44121bc failed test case for non-consecutive disk read (#14455) chenyuandGitHub 2026-01-30 23:44:04 -05:00
• b705c9143c assembly/amd: test more instructions (#14365) George HotzandGitHub 2026-01-31 12:40:22 +08:00
• c9a3ddb341 benchmark llama walltime script (#14454) George HotzandGitHub 2026-01-31 10:21:54 +08:00
• f5346d6a1a fix USE_ATOMICS for non float dtypes and make it the default (#14444) George HotzandGitHub 2026-01-31 09:44:16 +08:00
• e575dd8275 prevent UB in long decomp and more emulated tests (#14447) sirhcmandGitHub 2026-01-30 16:38:41 -08:00
• 3204f94454 correct var_vals schedule filter (#14451) chenyuandGitHub 2026-01-30 17:10:07 -05:00
• cfcd1debb5 test schedule with multiple AFTER (#14449) chenyuandGitHub 2026-01-30 15:59:00 -05:00
• 486d53d646 device: call free for external_ptr (#14448) nimlgenandGitHub 2026-01-30 23:53:17 +03:00
• e0978498dc amd: read_ptr/write_ptr/doorbells are not lists (#14445) nimlgenandGitHub 2026-01-30 23:11:57 +03:00
• 1803ee939d EMULATED_DTYPES=long works with CPU_LLVM (#14446) sirhcmandGitHub 2026-01-30 10:54:43 -08:00
• 03613e83ad update TestTensorMetadata (#14443) chenyuandGitHub 2026-01-30 12:39:01 -05:00
• cbb1eed57b hotfix: partial revert of 9eb449f88, caused llama NaN geohot 2026-01-30 17:19:27 +00:00
• 26f5c00265 move TestTensorMetadata to unit (#14442) chenyuandGitHub 2026-01-30 12:14:21 -05:00
• c05a0b85ae flip unique const src order [pr] (#14441) chenyuandGitHub 2026-01-30 11:44:18 -05:00
• ee2c78709d mlperf/llama: disable USE_ATOMICS for now geohot 2026-01-31 00:42:08 +08:00