Commit Graph

  • 9442442cb1 update variable names in search [pr] (#12979) chenyu 2025-10-28 15:37:52 -04:00
  • d66c997a39 feat: thunderkittens fa2 (#12955) wozeparrot 2025-10-28 11:27:45 -07:00
  • bb307b9e81 fix fp8 vectorization (#12977) b1tg 2025-10-29 01:55:30 +08:00
  • c11dd56956 amd: cleanup import urls (#12976) nimlgen 2025-10-29 00:43:02 +08:00
  • bb7aa19d67 Merge branch 'master' into replace_if_with_range replace_if_with_range George Hotz 2025-10-28 23:02:15 +08:00
  • 5e01cc299b zero len ranges fail (#12974) George Hotz 2025-10-28 22:49:55 +08:00
  • e936aa7974 cleanups from if range branch (#12973) George Hotz 2025-10-28 20:58:47 +08:00
  • f55f488543 cleanups from if range branch if_range_cleanups George Hotz 2025-10-28 19:38:45 +08:00
  • 117f37ae5f don't remove the gate George Hotz 2025-10-28 19:28:38 +08:00
  • ba23b097c1 fix image George Hotz 2025-10-28 19:22:59 +08:00
  • 3341272771 cleanup patterns George Hotz 2025-10-28 19:15:01 +08:00
  • 975f5ccc99 tests pass George Hotz 2025-10-28 19:12:21 +08:00
  • 901d27b3ba viz: optional text dims try 2 (#12971) qazal 2025-10-28 18:54:28 +08:00
  • d0de209ad0 don't brick on that George Hotz 2025-10-28 18:49:00 +08:00
  • 2d87d89202 replace if with range George Hotz 2025-10-28 18:30:11 +08:00
  • f5a3b33d33 add fun with nhwc convs George Hotz 2025-10-28 17:12:22 +08:00
  • 907499b02c clean up GROUP/SINK (#12969) George Hotz 2025-10-28 16:08:10 +08:00
  • e22c5e7e73 process_replay uses opts argument for KernelInfo.opts_to_apply (#12946) Sieds Lykles 2025-10-28 09:00:28 +01:00
  • 6c9560a846 more syntactic sugar for pyrender (#12968) George Hotz 2025-10-28 15:24:33 +08:00
  • c24ac16841 more syntactic sugar for pyrender better_spec George Hotz 2025-10-28 15:15:06 +08:00
  • b0da173f2f add unique to const, fix longstanding bug (#12965) George Hotz 2025-10-28 15:11:37 +08:00
  • e110f4632a split cat (on cpu) (#12864) Sieds Lykles 2025-10-28 07:55:19 +01:00
  • 3b82dee625 viz: match DEBUG=2 for exec item metadata (#12966) qazal 2025-10-28 14:53:57 +08:00
  • 99589dea81 move viz edge tagging to UOp graph (#12964) qazal 2025-10-28 12:46:23 +08:00
  • bbe0bebbf3 no range tags in kernels (#12962) George Hotz 2025-10-28 12:33:48 +08:00
  • 39c2117dea cleanup pyrender (#12961) George Hotz 2025-10-28 10:47:39 +08:00
  • 2832954bcb test with IGNORE_OOB=0 (#12960) George Hotz 2025-10-28 10:32:19 +08:00
  • 7aaa9b577c test with IGNORE_OOB=0 ignore_oob George Hotz 2025-10-28 10:15:13 +08:00
  • 7784cec48e pytest-split on spec (#12959) George Hotz 2025-10-28 10:09:01 +08:00
  • 4d817a289e simplify spec (#12958) George Hotz 2025-10-28 09:52:32 +08:00
  • 77aadcb01d more simp_spec George Hotz 2025-10-28 09:41:58 +08:00
  • 7d26342ab6 simplify spec George Hotz 2025-10-28 09:36:09 +08:00
  • 62e62d8760 move verify to spec / cleanup (#12956) George Hotz 2025-10-28 08:58:10 +08:00
  • 99fd9ff799 more explicit spec_cleanup George Hotz 2025-10-28 08:47:20 +08:00
  • f4f409290f lil George Hotz 2025-10-28 08:37:05 +08:00
  • 43019cbd6e move verify to spec / cleanup George Hotz 2025-10-28 08:32:18 +08:00
  • 24884c6768 fix: don't use KITTENS_HOPPER for 4090 (#12954) wozeparrot 2025-10-27 17:19:53 -07:00
  • 372d9e5753 hcq: helper for visible devices (#12950) nimlgen 2025-10-28 02:27:56 +08:00
  • f2ffe9c8cf Apply an override for nbio 7.3.0 to 7.2.0. (#12949) Justin Erenkrantz 2025-10-27 11:10:10 -07:00
  • 63484d837e Revert "viz graph drawing cleanups (#12933)" (#12947) qazal 2025-10-28 00:39:37 +08:00
  • a79832b01f control_flow.py -> linearizer.py [pr] (#12948) chenyu 2025-10-27 12:38:13 -04:00
  • 45e2f916a3 add quantize fp8 in llama3 (#12893) b1tg 2025-10-27 22:22:57 +08:00
  • 25c2da1579 check SPEC=2 in CI (#12945) George Hotz 2025-10-27 21:53:57 +08:00
  • 072f7c35c5 fix in/outs calculation in ProgramSpec (#12937) Sieds Lykles 2025-10-27 12:31:41 +01:00
  • e93c9bf6a7 viz: extend main code block to full height (#12944) qazal 2025-10-27 18:43:49 +08:00
  • 273b1f914d new pyrender, tested with SPEC=2 (#12934) George Hotz 2025-10-27 18:41:51 +08:00
  • 36918a7d7a cleanups real_pyrender George Hotz 2025-10-27 18:29:17 +08:00
  • c5a420ed7e optional forced_reshape George Hotz 2025-10-27 18:22:11 +08:00
  • f55056af48 viz always George Hotz 2025-10-27 18:19:13 +08:00
  • 0b81d9b0d6 pyrender George Hotz 2025-10-27 18:04:29 +08:00
  • ab7d8c24be Merge branch 'master' into real_pyrender George Hotz 2025-10-27 17:39:37 +08:00
  • 533f81bfc8 kernelize George Hotz 2025-10-27 17:38:30 +08:00
  • 701a632907 move VECTORIZE/CONST (#12942) George Hotz 2025-10-27 17:37:13 +08:00
  • 95748a4518 nv: map vram after resets (#12938) nimlgen 2025-10-27 17:17:07 +08:00
  • 8084aa06b2 Merge branch 'master' into real_pyrender George Hotz 2025-10-27 17:08:07 +08:00
  • 8fb545c475 don't late simplify on marg (#12941) George Hotz 2025-10-27 17:07:41 +08:00
  • 1a3498e6dc v const George Hotz 2025-10-27 17:07:32 +08:00
  • 8f9b09b4f6 .sintify George Hotz 2025-10-27 17:00:02 +08:00
  • 8743a21d3f Merge branch 'master' into real_pyrender George Hotz 2025-10-27 16:57:19 +08:00
  • 017ba454af Merge branch 'master' into no_late_simplify_marg no_late_simplify_marg George Hotz 2025-10-27 16:57:02 +08:00
  • 7139e036c5 bugfixes from pyrender (#12940) George Hotz 2025-10-27 16:56:53 +08:00
  • 0b85d31a68 Merge branch 'master' into real_pyrender George Hotz 2025-10-27 16:53:54 +08:00
  • 804133cffd rename RECIP to RECIPROCAL (#12939) George Hotz 2025-10-27 16:53:13 +08:00
  • 7563e20809 don't late simplify on marg George Hotz 2025-10-27 16:51:49 +08:00
  • 4c63cf3914 work George Hotz 2025-10-27 16:40:50 +08:00
  • f4da94af28 system: reset is a method of pcidevice (#12936) nimlgen 2025-10-27 16:21:10 +08:00
  • 97a4296d8b work George Hotz 2025-10-27 16:04:19 +08:00
  • b987b8b22a work George Hotz 2025-10-27 15:29:26 +08:00
  • af3211f73c test pyrender George Hotz 2025-10-27 14:47:35 +08:00
  • 6b54378eba working kitten matmul (#12935) wozeparrot 2025-10-26 23:40:49 -07:00
  • dc6d667941 pyrender always works with SPEC=3 George Hotz 2025-10-27 14:18:02 +08:00
  • 189582db5e viz graph drawing cleanups (#12933) qazal 2025-10-27 13:59:32 +08:00
  • f76a6c8845 unused spec_3 George Hotz 2025-10-27 13:31:52 +08:00
  • 33013db092 fix nan George Hotz 2025-10-27 13:21:43 +08:00
  • b7436f600d works George Hotz 2025-10-27 13:05:36 +08:00
  • 67183049c1 SPEC=3 works George Hotz 2025-10-27 12:51:56 +08:00
  • 70ba84eb04 viz: generic node label centering (#12925) qazal 2025-10-27 12:02:34 +08:00
  • 9cdb45f410 spec 3 works George Hotz 2025-10-27 11:59:33 +08:00
  • 46914e2f40 only check it there George Hotz 2025-10-27 11:50:33 +08:00
  • 1eb982e01f SPEC=3 tests pyrender George Hotz 2025-10-27 11:11:00 +08:00
  • eaeaea2f9c pyrender Ops.SPECIAL and use correct dtype for Ops.RANGE rendering (#12931) Sieds Lykles 2025-10-27 03:21:34 +01:00
  • 8c1368cab6 system: class PCIBarInfo (#12930) nimlgen 2025-10-27 03:57:42 +08:00
  • f00009c731 hcq: drivers take pcidev (#12929) nimlgen 2025-10-26 20:43:51 +08:00
  • 99a519f068 linearizer cleanup (#12923) ttomsa 2025-10-26 10:30:12 +00:00
  • c0c24d3a70 cleanup wmma (#12927) George Hotz 2025-10-26 18:26:47 +08:00
  • 0a32ab0006 nitpicks from typecheckers (#12926) George Hotz 2025-10-26 17:52:55 +08:00
  • db5c918215 source extra/cl_android.sh to fix opencl on android George Hotz 2025-10-26 15:27:51 +08:00
  • c94e597b3e viz ui selector cleanups (#12924) qazal 2025-10-26 14:40:47 +08:00
  • 94701d4838 clean up divide_exact order [pr] (#12919) chenyu 2025-10-25 18:47:57 -04:00
  • e18922f111 limit AND const min max to ints [pr] (#12918) chenyu 2025-10-25 16:07:52 -04:00
  • 92324172be amd: refactor usb into usbdevice (#12916) nimlgen 2025-10-26 01:00:19 +08:00
  • 3b192f5eac split viz graph rendering from dag layout (#12914) qazal 2025-10-25 15:36:44 +08:00
  • fbc7f4c12a only floats relu_pattern George Hotz 2025-10-25 13:17:33 +08:00
  • 2fec7ed6df lil symbolic pattern for relu George Hotz 2025-10-25 13:12:21 +08:00
  • 6415e3e8a7 use Ops.GROUP instead of Ops.NOOP for merging stores (#12912) George Hotz 2025-10-25 12:26:12 +08:00
  • b4f6a2c7a3 add kernel spec (#12911) George Hotz 2025-10-25 11:49:20 +08:00
  • 8a941d95a4 SPEC=2 is full spec, SPEC=1 is default (#12910) George Hotz 2025-10-25 11:10:43 +08:00
  • 5c20955c8e just use SPEC, not __debug__ full_spec George Hotz 2025-10-25 10:34:01 +08:00
  • 9cdd284008 SPEC=1 passes all tests George Hotz 2025-10-25 10:31:38 +08:00
  • 456560c1ff stateless tinyfs copyin (#12908) wozeparrot 2025-10-24 19:18:38 -07:00