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

• 9caceda79a amd: comgr is not required (#10128) nimlgenandGitHub 2025-05-01 13:41:44 +03:00
• c3d2e4a6e1 amd: use sdma to copy program (#10126) nimlgenandGitHub 2025-05-01 13:04:22 +03:00
• 09f5be9bcb amd: finalize device in case of failures (#10124) nimlgenandGitHub 2025-05-01 10:41:15 +03:00
• ef011ff5f9 flip Ops.COPY order [pr] (#10122) George HotzandGitHub 2025-05-01 00:26:24 -04:00
• 145e51247a split CAST and BITCAST in PYTHON [pr] (#10123) chenyuandGitHub 2025-04-30 23:27:35 -04:00
• bf5fb97498 fix AMD_LLVM bf16 tc for gfx1100 (#10102) Ignacio SicaandGitHub 2025-04-30 20:06:38 -03:00
• dd0070daab Revert "flip Ops.COPY order [pr] (#10120)" (#10121) George HotzandGitHub 2025-04-30 17:25:21 -04:00
• 984f09ac74 flip Ops.COPY order [pr] (#10120) George HotzandGitHub 2025-04-30 16:50:18 -04:00
• 17d4d258ea simple symbolic slice in llama [pr] (#10112) chenyuandGitHub 2025-04-30 14:36:35 -04:00
• b583ece8f3 amd: replace AMD_DRIVERLESS with AMD_IFACE (#10116) nimlgenandGitHub 2025-04-30 20:22:02 +03:00
• 0e1beaf44f nv: align copies + better test (#10118) nimlgenandGitHub 2025-04-30 20:09:53 +03:00
• 2941537250 cast is noop if src has dtypes.void (#10110) Ignacio SicaandGitHub 2025-04-30 13:55:41 -03:00
• fcdda4fc09 am: move boot memory to vram start (#10115) nimlgenandGitHub 2025-04-30 19:12:19 +03:00
• 844d5577d8 hcq: make copy_bufs and kernargs_size params configurable per device (#10114) nimlgenandGitHub 2025-04-30 18:43:50 +03:00
• 2ec3b722e2 nv: fix copies larger than 4g (#10117) nimlgenandGitHub 2025-04-30 18:43:17 +03:00
• d81acbeef6 multi: move shrink after copy (#10109) George HotzandGitHub 2025-04-30 10:29:51 -04:00
• 67bd8489ad grouper cleanups [pr] (#10113) qazalandGitHub 2025-04-30 13:54:47 +03:00
• b4c9a3d8f4 hcq: use mmio iface in copies (#10111) nimlgenandGitHub 2025-04-30 11:05:13 +03:00
• 5c7d004da5 hcq: refactor int ptrs to hcqbuffers (#10105) nimlgenandGitHub 2025-04-30 00:12:18 +03:00
• 573bbb9746 Revert "remove TransformerBlock contiguous in llama (#10104)" (#10108) chenyuandGitHub 2025-04-29 15:28:38 -04:00
• 4a04098389 fix llama3 with nf4 quantize (#10107) chenyuandGitHub 2025-04-29 15:14:36 -04:00
• 9c1b80499f names for graph rewrites + null device supports exp and friends (#10106) George HotzandGitHub 2025-04-29 14:28:20 -04:00
• b8d07dcc54 remove TransformerBlock contiguous in llama (#10104) chenyuandGitHub 2025-04-29 14:15:39 -04:00
• 9d5677c12c fix ptx linearizer bug 2 [pr] (#9967) Ignacio SicaandGitHub 2025-04-29 14:30:07 -03:00
• a59d18da21 hack for VIZ=1 with examples/llama (#10103) qazalandGitHub 2025-04-29 18:42:17 +03:00
• 93bf8764f2 do not open devices in lowering (#10101) qazalandGitHub 2025-04-29 18:18:16 +03:00
• c3ff308abb range has only one src now [pr] (#10100) George HotzandGitHub 2025-04-29 10:31:05 -04:00
• 427471550a hotfix: amd tflops to 74 and some external_benchmark_sdxl_softmax stuff geohot 2025-04-29 09:01:30 -04:00
• 58cf8cd493 add support for "shared_mem" for LLVM (#10093) Ignacio SicaandGitHub 2025-04-29 09:56:36 -03:00
• ad7546c931 assert in test_indexing_two_bind instead of silent fail (#10099) qazalandGitHub 2025-04-29 15:23:25 +03:00
• cee220a1ab always expand ssa on wheres (#9697) 3 people 2025-04-29 08:08:41 -04:00
• 3b67f56c02 kernelize some llama realizes (#10098) qazalandGitHub 2025-04-29 13:39:56 +03:00
• cbf7347cd6 display viz rewrites with tabbing if they are subrewrites (#10097) qazalandGitHub 2025-04-29 12:57:21 +03:00
• 73c2f6602f test sdxl softmax (#10096) George HotzandGitHub 2025-04-28 21:55:50 -04:00
• eaceafecae do fusion locally (#10095) George HotzandGitHub 2025-04-28 20:45:37 -04:00
• 3eba3d6ee9 don't pass model in convert_from_huggingface and convert_from_gguf (#10094) chenyuandGitHub 2025-04-28 20:11:19 -04:00
• a2d0684fc1 test_attention_simple_view (#10092) George HotzandGitHub 2025-04-28 20:01:22 -04:00
• bda116d773 fix use_tensor_cores propagation (#10048) Ignacio SicaandGitHub 2025-04-28 19:30:50 -03:00
• d32f5e9f3a improve rendering of shapes in viz + investigate symbolic [pr] (#10091) George HotzandGitHub 2025-04-28 16:44:09 -04:00
• dbb7aee02e Split constant in div with negative x (#10088) Sieds LyklesandGitHub 2025-04-28 22:24:14 +02:00
• 610ee79b22 cherry pick mlperf5.0 branch to master (#10089) chenyuandGitHub 2025-04-28 15:36:56 -04:00
• 459a223202 simpler Literal annotation in code_for_workitem [pr] (#10087) chenyuandGitHub 2025-04-28 14:59:25 -04:00
• dcd9a633c3 am: load minimum fw (#10083) nimlgenandGitHub 2025-04-28 21:28:05 +03:00
• ecff82a698 fixing single kernel softmax: resolve (#10086) George HotzandGitHub 2025-04-28 13:46:20 -04:00
• 4c242b0483 hotfix: tests all pass on metal local geohot 2025-04-28 12:09:00 -04:00
• 690dac79b5 don't modify the ranges on reduce rewrite (#10062) 3 people 2025-04-28 12:01:19 -04:00
• 5130759605 Make sure clang always inline batched functions (#10037) quortusandGitHub 2025-04-28 16:48:24 +02:00
• c4a50f9d89 fix full shape in kernel.py [pr] (#10085) George HotzandGitHub 2025-04-28 09:32:58 -04:00
• ac37510f60 remu: only write v_cmp result if exec is set (#10084) qazalandGitHub 2025-04-28 15:31:52 +03:00
• d6b436a815 remu bugfix with -0.0 negation (#10082) qazalandGitHub 2025-04-28 10:46:42 +03:00
• 15e4302784 am: optimize zeroing out boot structs (#10081) nimlgenandGitHub 2025-04-28 10:15:32 +03:00
• 68e5ab8552 am: fix typo in fw loading (#10080) nimlgenandGitHub 2025-04-28 09:45:00 +03:00
• e996584685 olmoe in mac benchmark (#10077) chenyuandGitHub 2025-04-27 21:07:02 -04:00
• 732e172961 don't require contiguous after fuse (#10074) George HotzandGitHub 2025-04-27 13:17:22 -04:00
• 1aed04ec12 cpu is ground truth in VALIDATE_WITH_CPU=1 [pr] (#10067) qazalandGitHub 2025-04-27 20:14:21 +03:00
• 129bddde74 lin failure from SINGLE_KERNEL_SOFTMAX (#10073) George HotzandGitHub 2025-04-27 13:02:10 -04:00
• b341296304 hotfix: save sdxl ram geohot 2025-04-27 12:09:45 -04:00
• 68c5f7ba80 load fast in sdxl (#10072) George HotzandGitHub 2025-04-27 11:58:51 -04:00
• 768eb94c3e disable debug for load_state_dict [pr] (#10070) George HotzandGitHub 2025-04-27 11:11:56 -04:00
• 4b8ef6ce78 hotfix: sdxl corealize geohot 2025-04-27 10:41:46 -04:00
• b6d2effaf5 assign is contiguous (#10066) George HotzandGitHub 2025-04-27 08:40:33 -04:00
• 1253819151 make beautiful indexing use a Variable (#10063) George HotzandGitHub 2025-04-27 08:22:38 -04:00
• a13a43c4fe yolo 416 to 640 res (#10047) Rory ClearandGitHub 2025-04-27 01:45:58 +01:00
• 4c1ce1a299 don't simplify if div folding resulted in negative numerator (#10064) chenyuandGitHub 2025-04-26 17:01:18 -04:00
• 1805403821 fix rand arange folding (#10060) George HotzandGitHub 2025-04-26 12:24:05 -04:00
• d13c100981 don't sort dims in verify_sink_dims [pr] (#10059) qazalandGitHub 2025-04-26 18:24:30 +03:00
• c80fe6d5fc handle some fancier reduces (#10057) George HotzandGitHub 2025-04-26 11:20:15 -04:00
• e08270c1ba nv: fix program init for no-args kernels (#10058) nimlgenandGitHub 2025-04-26 18:08:53 +03:00
• 11113c9d07 reduce_unparented (#10056) George HotzandGitHub 2025-04-26 09:48:16 -04:00
• ea5dddc537 reduce collapse generic (#10045) George HotzandGitHub 2025-04-26 09:13:24 -04:00
• 5cdc96409e Update outdated renderer.render calls (#10044) quortusandGitHub 2025-04-26 13:35:19 +02:00
• e055b9422f am: fix mmap failures (#10054) nimlgenandGitHub 2025-04-26 14:21:28 +03:00
• e1d2b64e92 remu new instructions (#10050) qazalandGitHub 2025-04-26 02:04:12 +03:00
• bba5d0a3e4 remu refactors (#10028) qazalandGitHub 2025-04-25 23:31:10 +03:00
• 0fc85a2b0a hcqfuzz: init (#10049) nimlgenandGitHub 2025-04-25 23:19:21 +03:00
• b30050e287 fix amdgpu_disassemble on mac [pr] (#10042) qazalandGitHub 2025-04-25 10:23:11 +03:00
• a197aa4ef3 upat reduce syntax [pr] (#10040) George HotzandGitHub 2025-04-24 22:05:28 -04:00
• 76a86735c0 hotfix amd bf16 is supported case (#10039) Ignacio SicaandGitHub 2025-04-24 21:29:27 -03:00
• b4f823acbe fix helper_tc_allclose (#9606) Ignacio SicaandGitHub 2025-04-24 18:36:40 -03:00
• 3a189fa561 More yolo processing in tinygrad (#9928) 3 people 2025-04-24 21:21:46 +01:00
• 74c6cf8be3 lint mlperf model_train (#10038) chenyuandGitHub 2025-04-24 16:19:44 -04:00
• 51ca19d061 set test_tensor_cores_padded_amd to expectedFailure (#10036) Ignacio SicaandGitHub 2025-04-24 17:11:40 -03:00
• 914d89fa0b fix tensor cores for gfx1201 (#9838) 4 people 2025-04-25 02:57:41 +08:00
• 779aa1e2e9 Enable image tests on cloud if clouddev supports image (#9903) 3 people 2025-04-24 23:30:12 +05:00
• 29a12b19ea Add macos CLOUD tests (#10033) uuuvnandGitHub 2025-04-24 23:14:13 +05:00
• 55942a8d8e [Bounty] moved index_tensor off cpu in torch_backend (#9916) Nishant RajadhyakshaandGitHub 2025-04-24 11:12:37 -07:00
• 373ca59b7f use is_dtype_supported to check dtype support in tc tests (#10035) Ignacio SicaandGitHub 2025-04-24 14:59:14 -03:00
• 93a1e9eeb9 improve bf16 case for is_dtype_supported [pr] (#10034) Ignacio SicaandGitHub 2025-04-24 14:03:57 -03:00
• 754d789f51 Fix and enable jit tests on CLOUD (#10031) uuuvnandGitHub 2025-04-24 20:39:31 +05:00
• 0b482fb824 add RDNA3 parser to remu (#10025) qazalandGitHub 2025-04-24 16:34:07 +03:00
• 0d903c9495 Print clouddev instead of cloudev's renderer (#10023) uuuvnandGitHub 2025-04-24 18:32:08 +05:00
• aec75f51ef fixup some slow CI tests [pr] (#10027) George HotzandGitHub 2025-04-24 14:20:49 +01:00
• c990aac2b1 skip flaky test_transcribe_file1_OOB (#10026) qazalandGitHub 2025-04-24 16:09:43 +03:00
• 4e2ccfddc6 ci refactor to split AMD/NVIDIA [pr] (#10024) George HotzandGitHub 2025-04-24 13:59:54 +01:00
• 0c68e44d6f Cloud properties (#10021) uuuvnandGitHub 2025-04-24 17:17:01 +05:00
• db00d88415 hotfix: handle bad z3 install like z3 import fail geohot 2025-04-24 08:09:40 -04:00
• e75be6eafc [bounty] [pr] index validation with z3 (#9981) 3 people 2025-04-24 14:06:08 +02:00
• 9e49721c47 CPUGraph support for clang (#10014) 3 people 2025-04-24 13:52:35 +02:00
• c3ad7b2a84 create randperm and support pytorch backend (#10019) Park JunandGitHub 2025-04-24 20:29:02 +09:00
• b545338e59 isin_Tensor_out added (#10018) Matthew DaiterandGitHub 2025-04-24 19:26:51 +08:00