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

• a25abf55e3 retinanet only call postprocess_detections with RUNMLPERF (#10017) chenyuandGitHub 2025-04-23 20:45:38 -04:00
• 7f53e80db9 hotfix: amd mmio on mi300 (#10016) nimlgenandGitHub 2025-04-24 01:08:18 +03:00
• 1c5e353249 am: use mmio iface (#10012) nimlgenandGitHub 2025-04-24 00:27:04 +03:00
• 65faa1d94b explicit device in mlperf scripts (#10015) chenyuandGitHub 2025-04-23 17:11:52 -04:00
• a3f938dbee remove retinanet INITMLPERF from beam script (#10011) chenyuandGitHub 2025-04-23 14:32:54 -04:00
• cc52b9c528 am: add entry() to PT (#10010) nimlgenandGitHub 2025-04-23 20:45:52 +03:00
• c952cb965e amd: use mmio iface (#9997) nimlgenandGitHub 2025-04-23 20:13:00 +03:00
• 5542aeb0e4 RetinaNet MLPerf flag updates (#10009) Francis LataandGitHub 2025-04-23 13:00:34 -04:00
• de0504276b pop 0 is slow [pr] (#10007) George HotzandGitHub 2025-04-23 17:00:59 +01:00
• d3a8d5c128 print postprocess_detections time in retinanet eval (#10005) chenyuandGitHub 2025-04-23 11:39:56 -04:00
• 2ed3acd767 toposort is a function [pr] (#10004) George HotzandGitHub 2025-04-23 16:25:03 +01:00
• 0730ff0e50 Skip test that requires lru if device's allocator isn't lru (#10003) uuuvnandGitHub 2025-04-23 20:12:56 +05:00
• 954cb06957 deepwalk without recursion [pr] (#10002) George HotzandGitHub 2025-04-23 15:57:50 +01:00
• 9de73ccc22 Skip test that requires python 3.12 on older versions (#10001) uuuvnandGitHub 2025-04-23 19:09:26 +05:00
• 71ecc7fa1a use a pattern matcher for upcast [pr] (#10000) George HotzandGitHub 2025-04-23 14:24:23 +01:00
• cc1087d2ec move simplify into views_to_indexed_uops (#9999) George HotzandGitHub 2025-04-23 13:50:27 +01:00
• c39128133c retinanet green scripts (#9996) chenyuandGitHub 2025-04-23 08:28:03 -04:00
• a4a5f2d54a faster block order [pr] (#9998) George HotzandGitHub 2025-04-23 13:11:30 +01:00
• 61bfd23881 update mlperf-logging version (#9995) chenyuandGitHub 2025-04-22 19:32:39 -04:00
• dbbd755cba FP8s truncate (#9937) 3 people 2025-04-23 01:12:49 +02:00
• 58180caad3 schedule linearize small cleanups [pr] (#9994) qazalandGitHub 2025-04-23 00:42:29 +03:00
• f4ec57baff new schedule linearizer enqueues KERNEL UOps [pr] (#9993) qazalandGitHub 2025-04-23 00:17:58 +03:00
• d1f6701eb7 hotfix: lower amd threshold + improve block reorder test geohot 2025-04-22 20:43:22 +01:00
• db51133537 rename HWInterface -> FileIOInterface (#9989) nimlgenandGitHub 2025-04-22 22:18:57 +03:00
• c1539b0319 putting add first orders loads as expected (#9991) George HotzandGitHub 2025-04-22 20:12:05 +01:00
• bd580d8ea4 hcq: use mmio interface in nv (#9986) nimlgenandGitHub 2025-04-22 21:58:12 +03:00
• feee6986c9 faster block reorder (#9990) George HotzandGitHub 2025-04-22 19:18:57 +01:00
• 6cb2d18c03 refactor schedule linearize to defaultdict [pr] (#9984) qazalandGitHub 2025-04-22 19:00:23 +03:00
• 9e5e371999 make DISABLE_COMPILER_CACHE a ContextVar [pr] (#9983) chenyuandGitHub 2025-04-22 10:32:54 -04:00
• bbc324f5dc remove CAST_AFTER_EXPAND (#9980) qazalandGitHub 2025-04-22 16:06:11 +03:00
• c519b553db non recursive toposort is 2x+ faster (#9979) George HotzandGitHub 2025-04-22 13:59:38 +01:00
• 0d9014d021 place create_ast last, type_verify in the end (once) [pr] (#9977) qazalandGitHub 2025-04-22 15:15:23 +03:00
• fb89d9a584 retinanet eval combine output on GPUS[0] (#9966) chenyuandGitHub 2025-04-22 07:43:51 -04:00
• 7b55846e08 prep STORE UOp creation for multi output [pr] (#9975) qazalandGitHub 2025-04-22 14:34:52 +03:00
• e358e0a0c6 move metadata set to tensor [pr] (#9976) George HotzandGitHub 2025-04-22 12:30:35 +01:00
• f6271515fe refactor UOp.st [pr] (#9973) qazalandGitHub 2025-04-22 13:46:56 +03:00
• f5dc70c624 microbenchmarks + micro speed ups (#9972) George HotzandGitHub 2025-04-22 11:30:46 +01:00
• 1cf4e24ca5 fix kernelize usage with pm_gradient (#9953) qazalandGitHub 2025-04-22 12:26:05 +03:00
• 32bbff942c amd: add nbio 7.2.0 for some rdna2 (#9964) 3 people 2025-04-22 05:10:48 -04:00
• 0e79aee706 use_tensor_cores bugfix (#9969) Ignacio SicaandGitHub 2025-04-21 22:58:17 -03:00
• 5294c32279 dev scripts for retinanet (#9968) chenyuandGitHub 2025-04-21 17:54:56 -04:00
• 4340197132 am: download fw from web (#9956) nimlgenandGitHub 2025-04-21 23:26:33 +03:00
• 7244ca863c am: fix double read of sdma fw (#9965) nimlgenandGitHub 2025-04-21 23:04:34 +03:00
• b35f94b6ec Don't hardcode default CLOUDDEV (#9935) uuuvnandGitHub 2025-04-21 22:46:55 +05:00
• defa1e77f6 get the proper dataset count (#9962) Francis LataandGitHub 2025-04-21 12:11:37 -04:00
• 36ed3c3253 fix kernelize with VIEW children (#9961) qazalandGitHub 2025-04-21 18:38:46 +03:00
• 757533cbe6 Less verbose cloud multiprocessing start (#9960) uuuvnandGitHub 2025-04-21 20:19:54 +05:00
• d7e247f329 RetinaNet INITMLPERF support (#9950) Francis LataandGitHub 2025-04-21 10:32:05 -04:00
• 014f870733 rm (#9959) 3 people 2025-04-21 17:23:45 +03:00
• f68c7041c4 doc fix is_floating_point dtype.float -> dtypes.float (#9958) chenyuandGitHub 2025-04-21 09:23:59 -04:00
• 2d423e6737 fix assertion message for supported device in export_model (#9957) akhuntsariaandGitHub 2025-04-21 15:23:44 +02:00
• 783a191925 rm mul from _masked_setitem (#9951) ttomsaandGitHub 2025-04-21 11:41:50 +01:00
• 46469f00a2 am: tiny changes in psp load (#9952) nimlgenandGitHub 2025-04-21 11:52:02 +03:00
• 0bee225a58 Tensor.kernelize docs (#9946) qazalandGitHub 2025-04-21 11:34:03 +03:00
• ea4cb2c715 small cleanups (#9947) Francis LataandGitHub 2025-04-20 20:33:20 -04:00
• e8910540f6 Kernelize can be called multiple times on a Tensor (#9949) qazalandGitHub 2025-04-21 01:28:47 +03:00
• 1d90be2cff match kernelize API in process replay (#9948) qazalandGitHub 2025-04-21 00:23:41 +03:00
• 343a5eb588 dedup assigns in grouper VIZ name function [pr] (#9942) qazalandGitHub 2025-04-20 16:42:25 +03:00
• e20ef7196a Tensor.kernelize (#9845) qazalandGitHub 2025-04-20 15:53:49 +03:00
• dd16087f62 fold double ASSIGN to same target (#9941) qazalandGitHub 2025-04-20 14:06:38 +03:00
• 9a9aba4cd5 setitem tests (some failing) from kernelize (#9940) qazalandGitHub 2025-04-20 13:47:55 +03:00
• 6c30948df6 hand_coded_optimizations returns list[Opt] [pr] (#9938) chenyuandGitHub 2025-04-19 20:26:59 -04:00
• 720f20865b remove required_optimizations (#9848) chenyuandGitHub 2025-04-19 16:51:16 -04:00
• 218e01833d update scheduler section for abstractions2.py [pr] (#9927) qazalandGitHub 2025-04-19 12:09:14 +03:00
• 3fdba48fc7 update bert green and README (#9934) chenyuandGitHub 2025-04-18 21:21:28 -04:00
• b359125ebf rewrite the linearizer (#9885) George HotzandGitHub 2025-04-18 22:35:40 +01:00
• 023b1c28a2 test_tensor_cores_padded refactor (#9724) Ignacio SicaandGitHub 2025-04-18 17:05:54 -03:00
• afff82ba0f fix ptx linearizer bug [pr] (#9926) Ignacio SicaandGitHub 2025-04-18 13:48:43 -03:00
• 617b45748f fuse embedding for bert on red (#9925) chenyuandGitHub 2025-04-18 07:20:25 -04:00
• b58decac0c fix diamond assigns before mapping tensors UOps to assigns (#9855) qazalandGitHub 2025-04-18 14:17:43 +03:00
• a37d921917 get name from SINK in process replay (#9924) qazalandGitHub 2025-04-18 13:51:11 +03:00
• aa98aff4cd don't use ops name, just keep sink (#9922) George HotzandGitHub 2025-04-18 08:59:18 +01:00
• 8919370c76 hotfix: fix test_save_all_dtypes on METAL geohot 2025-04-18 08:42:31 +01:00
• 16dfe0a902 upstream remu (#9921) qazalandGitHub 2025-04-18 01:57:36 +03:00
• d287afe3b1 remove shapeless const check in full_shape [pr] (#9911) qazalandGitHub 2025-04-18 00:00:26 +03:00
• fe6a482f1d pin hypothesis version to 6.131.0 (#9920) chenyuandGitHub 2025-04-17 16:34:10 -04:00
• f5256e0020 Kernel.apply_opts [pr] (#9917) chenyuandGitHub 2025-04-17 08:00:56 -04:00
• e2ed673c94 FUSE_ARANGE_UINT to not fuse uint (#9915) chenyuandGitHub 2025-04-16 18:49:38 -04:00
• 497daa658a hotfix: edge-labels go above the overlay (#9910) qazalandGitHub 2025-04-16 18:38:12 +03:00
• e8e43c6dad ensure edge labels are always on top (#9908) qazalandGitHub 2025-04-16 16:08:06 +03:00
• 5265f25088 add counter for incoming edges in viz (#9907) qazalandGitHub 2025-04-16 15:14:14 +03:00
• 2c7c205bc5 Fix dtype comparisons in vectorized transcendental + tests (#9794) Eitan TurokandGitHub 2025-04-16 15:06:12 +03:00
• 929e5a9905 do not construct GrouperContext [pr] (#9906) qazalandGitHub 2025-04-16 13:26:31 +03:00
• 047c8fd70d Add amax support to Tensor operations in Torch Backend (#9905) XingyuandGitHub 2025-04-16 17:35:50 +08:00
• d7f623dac2 Use Buffer in cloud server instead of opaques (#9875) uuuvnandGitHub 2025-04-16 14:17:32 +05:00
• 05334e0f3f construct children from UOp.toposort [pr] (#9882) qazalandGitHub 2025-04-16 11:55:59 +03:00
• 4e8f25109a Revert "ONNX add output shape validation (#9720)" (#9904) geohotstanandGitHub 2025-04-16 15:15:56 +08:00
• e8024c8281 faster bert global_norm (#9901) chenyuandGitHub 2025-04-15 18:24:44 -04:00
• 91ccf1c343 Off by one error in start_pos (#9792) Sieds LyklesandGitHub 2025-04-15 21:07:13 +02:00
• 5849c43382 FP8s part 1 (#9887) 3 people 2025-04-15 17:20:02 +02:00
• 31483050c0 add eval_freq flag (#9894) Francis LataandGitHub 2025-04-15 06:42:40 -04:00
• 83ae83d871 compare amd and am to cpu as well (#9896) nimlgenandGitHub 2025-04-15 13:32:18 +03:00
• 23a95dd84d script to compare amd and am kerns (#9889) nimlgenandGitHub 2025-04-15 00:11:22 +03:00
• ce454793e6 support specifying dtype for Tensor.linear (#9886) chenyuandGitHub 2025-04-14 13:55:11 -04:00
• e8a0aee88d add arch to AMDLLVMRenderer (#9884) 3 people 2025-04-15 00:59:22 +08:00
• 44e4934167 fast pattern matcher [pr] (#9737) George HotzandGitHub 2025-04-14 15:24:41 +01:00
• 43d3a75d6c increase bert max train_steps (#9883) chenyuandGitHub 2025-04-14 08:53:44 -04:00
• bf099520a4 add names to grouper rewrites + cleanups [pr] (#9881) qazalandGitHub 2025-04-14 19:47:36 +08:00
• ca8aaadd00 clean up some patterns [pr] (#9880) George HotzandGitHub 2025-04-14 11:33:22 +01:00
• 355739fc94 switch to universal match [pr] (#9879) George HotzandGitHub 2025-04-14 09:15:37 +01:00