Commit Graph

  • 7bdb3adbbf viz/cli: simplification and reordering (#15785) qazal 2026-04-17 15:16:07 +03:00
  • e1d13bc4fe add GGUF IQ4_XS support (#15766) George Hotz 2026-04-17 14:43:39 +08:00
  • 9e60e4a7e7 llama: native fp8 (#15733) wozeparrot 2026-04-17 13:16:05 +08:00
  • a9b6cfece0 refactor llm into files (#15780) George Hotz 2026-04-17 12:33:11 +08:00
  • 1fac03ce54 softmax and friends to mixin (#15778) chenyu 2026-04-16 23:03:37 -04:00
  • ec00cefa5b llm is the only app (#15779) George Hotz 2026-04-17 10:44:48 +08:00
  • 0e69388f6b viz/cli: add DEBUG, optional number of rows (#15777) qazal 2026-04-17 04:36:47 +03:00
  • 2d196fb9bb move Tensor.size to mixin (#15775) chenyu 2026-04-16 17:56:17 -04:00
  • 9f4b7bed25 add pickled jit regression test (#15774) Christopher Milan 2026-04-16 13:59:09 -07:00
  • 6d9320ffb3 add NO_COLOR (#15765) qazal 2026-04-16 22:44:55 +03:00
  • 12c653a743 remove opts arg in get_program, everything uses opts_to_apply [pr] (#15767) qazal 2026-04-16 22:42:43 +03:00
  • f0c12a2004 another form of assign to itself (#15770) chenyu 2026-04-16 15:17:19 -04:00
  • 4e88d875ba llm: glm 4.7 flash (#15738) b1tg 2026-04-16 22:42:04 +08:00
  • d147e2a549 update test_nested_after_contiguous_store (#15763) chenyu 2026-04-16 09:59:26 -04:00
  • 126cda45f8 viz/cli: cleanups, add memory printer (#15762) qazal 2026-04-16 16:44:47 +03:00
  • f57380cbc2 simplify GatedDeltaNetBlock using two state tensors (#15704) George Hotz 2026-04-16 21:14:00 +08:00
  • c04f3eaa70 jit: capturedjit is linear (#15743) nimlgen 2026-04-16 14:54:39 +03:00
  • d1cce7a476 put the ranges on store instead of after (#15759) George Hotz 2026-04-16 19:06:40 +08:00
  • d24466c844 CALL with return value is FUNCTION (#15758) George Hotz 2026-04-16 13:25:07 +08:00
  • e336f3cf8c CALL with return value is FUNCTION param_call_is_function George Hotz 2026-04-16 12:36:14 +08:00
  • 218d6b8988 delete old UOp.size [pr] (#15756) chenyu 2026-04-15 23:21:00 -04:00
  • d090732270 usbgpu: reset endpoint for custom fw (#15754) wozeparrot 2026-04-16 11:01:27 +08:00
  • 983a7bb576 exclude __del__ from TRACEMETA wrapping (#15747) Muzammil 2026-04-16 08:19:55 +05:30
  • 8bd4fead26 UOp.size -> prod(max_shape) (#15755) chenyu 2026-04-15 22:41:30 -04:00
  • 10c262ced8 update tests that use UOp.size (#15753) chenyu 2026-04-15 21:58:27 -04:00
  • 96092d110c fix process_replay Ops.BEAM [pr] (#15752) qazal 2026-04-16 01:35:28 +03:00
  • 41421c3b48 BUFFER size is their arg (#15750) chenyu 2026-04-15 18:08:29 -04:00
  • be8005c5dc DEV: secondary targets (#15748) Christopher Milan 2026-04-15 14:26:20 -07:00
  • 507c02cecb fix symbolic contiguous_view_offset (#15749) chenyu 2026-04-15 16:54:38 -04:00
  • 164495678c test_graph to use uops (#15746) nimlgen 2026-04-15 21:59:41 +03:00
  • 1f26584b2e viz/cli: cleanups from linter (#15745) qazal 2026-04-15 21:36:24 +03:00
  • 7cbfa1896a comment out unused arm, triton in toml (#15741) chenyu 2026-04-15 10:05:19 -04:00
  • 1c36878008 DEV: suggest alternatives (#15732) Christopher Milan 2026-04-14 20:42:32 -07:00
  • 1ae6528bb6 move schedule into schedule (#15736) George Hotz 2026-04-15 11:03:25 +08:00
  • 3721c60bef llama: bs 16 (#15737) wozeparrot 2026-04-15 10:52:03 +08:00
  • 150ee9eb6d sched docs move_schedule George Hotz 2026-04-15 10:34:47 +08:00
  • 5dcdfb0d75 callify to root George Hotz 2026-04-15 10:20:38 +08:00
  • bb69860d41 move schedule into schedule George Hotz 2026-04-15 10:17:31 +08:00
  • 480ad264a4 llama: per device amax (#15735) wozeparrot 2026-04-15 10:01:17 +08:00
  • adc96cd724 qcom: synchronize for copyin (#15731) Christopher Milan 2026-04-14 15:31:15 -07:00
  • 3394d18066 size*itemsize -> nbytes (#15729) chenyu 2026-04-14 16:27:54 -04:00
  • e9ecc990ea amd: add r9700 devid (#15721) nimlgen 2026-04-14 20:15:00 +03:00
  • 2450c8cba8 rename to callify + fix mypy (#15727) George Hotz 2026-04-14 23:43:19 +08:00
  • 528faa18ec update env_vars.md (#15722) chenyu 2026-04-14 09:13:35 -04:00
  • 359b1582d6 amd: EMU DPP support (#15719) George Hotz 2026-04-14 14:58:41 +08:00
  • 2b8d303f75 allreduce in precast dtype (#15689) wozeparrot 2026-04-14 11:24:12 +08:00
  • 5683126844 llm: support for tekken tokenizer (#15720) George Hotz 2026-04-14 10:52:07 +08:00
  • 70883a6950 cat the stack to mixin (#15715) chenyu 2026-04-13 18:44:39 -04:00
  • 355e2729d3 viz: keep program UOp in data (#15714) qazal 2026-04-14 01:04:16 +03:00
  • 905b8adc97 viz: cli and server cleanups (#15713) qazal 2026-04-14 00:42:29 +03:00
  • d83707ec29 autogen: explicit types (#15679) Christopher Milan 2026-04-13 13:54:39 -07:00
  • ac41f15fc1 cumsum to mixin (#15712) chenyu 2026-04-13 15:06:08 -04:00
  • eac481b67f mlx: fix ctypes (#15711) nimlgen 2026-04-13 20:43:56 +03:00
  • b370f5c5ac hcq: call free for unmap (#15710) nimlgen 2026-04-13 20:30:21 +03:00
  • 931d6cc62a basic getitem to mixin (#15697) chenyu 2026-04-13 13:04:36 -04:00
  • 7610bdc59e block multistore, it's not supported (#15708) George Hotz 2026-04-13 20:57:59 +08:00
  • 84d64b5835 hotfix: abstractions4 works in mock except asm George Hotz 2026-04-13 20:57:00 +08:00
  • 16f50a40a5 remove REMU from tree (#15706) George Hotz 2026-04-13 20:43:08 +08:00
  • ac027055ef viz: no global state (#15705) qazal 2026-04-13 15:35:20 +03:00
  • 4c1fb18a09 Revert "Revert "Tests for GatedDeltaNetBlock + fix multi after assign issue (…" (#15703) George Hotz 2026-04-13 19:09:38 +08:00
  • 0cec42db71 Revert "Tests for GatedDeltaNetBlock + fix multi after assign issue (#15700)" (#15702) George Hotz 2026-04-13 19:06:44 +08:00
  • 6f5d756282 Tests for GatedDeltaNetBlock + fix multi after assign issue (#15700) George Hotz 2026-04-13 18:43:23 +08:00
  • 19c0e4a11d gemma4 from GPT 5.4 gemma4_gpt George Hotz 2026-04-07 16:51:46 +08:00
  • 2b5ba0095d qwen3.5 (#15210) b1tg 2026-04-13 15:35:24 +08:00
  • 2ada38f777 viz: execv after all producers complete (#15696) qazal 2026-04-13 02:15:47 +03:00
  • f7ff480fa6 start mixin getitem tests (#15695) chenyu 2026-04-12 18:54:33 -04:00
  • 77385ccb37 more trivial stuff to mixin (#15693) chenyu 2026-04-12 15:17:16 -04:00
  • ff1de5ae13 normalize logsumexp contiguous_backward to mixin (#15692) chenyu 2026-04-12 13:13:00 -04:00
  • 0254cfe642 move usum and uprod to mixin (#15690) chenyu 2026-04-12 11:42:24 -04:00
  • e9b2e156b4 add jitbeam to tinygpu docs (#15691) nimlgen 2026-04-12 18:20:26 +03:00
  • e706f408cb suppress test warnings from numpy (#15688) chenyu 2026-04-11 22:33:20 -04:00
  • 938cba4fdf amd: a bit faster usb, skip interrupts on sync (#15686) nimlgen 2026-04-11 17:26:36 +03:00
  • 054d78e6ff fix llama profile.sh NULL source (#15685) qazal 2026-04-11 16:56:05 +03:00
  • 4ca844e96b add Q1_0 gguf type (#15683) Graham Robbins 2026-04-11 05:17:24 -05:00
  • 5156a04cf5 add support for AM_POWER_LIMIT (#15684) George Hotz 2026-04-11 17:14:54 +08:00
  • 457508d5a0 llama: save more 2 (#15681) wozeparrot 2026-04-11 16:03:36 +08:00
  • 29238b772f AMD USB: support for 0xF3 power toggle George Hotz 2026-04-11 13:03:56 +08:00
  • b5a9465b13 llm: add support for moonlight (deepseek MLA) (#15466) George Hotz 2026-04-11 10:32:48 +08:00
  • 590464c8d8 llama: only support wqkv path + cleanups (#15680) wozeparrot 2026-04-11 07:39:27 +08:00
  • aa012d6f08 usb: faster custom (#15678) nimlgen 2026-04-10 23:00:31 +03:00
  • 58646f9569 usb fast copyout (#15677) nimlgen 2026-04-10 21:04:49 +03:00
  • 0d5cdc9600 viz: split draw loop (#15676) qazal 2026-04-10 17:25:50 +03:00
  • e1334d3852 move canonicalize_device to device.py (#15675) chenyu 2026-04-10 09:43:56 -04:00
  • 8e7fcc8ca3 remove _include_initial in _cumalu (#15674) chenyu 2026-04-10 08:33:30 -04:00
  • 9092f2a8c0 llm: add shared_expert and rope_dim support from qwen35 (#15673) George Hotz 2026-04-10 19:18:27 +08:00
  • 9ab1415937 llm: fix streaming UTF-8 decode (#15653) b1tg 2026-04-10 17:01:02 +08:00
  • 55bcd7cc9e llama amax outside (#15670) wozeparrot 2026-04-10 14:08:03 +08:00
  • 16f3448b26 Add HIP to abstractions4 (#15672) George Hotz 2026-04-10 14:05:52 +08:00
  • ed2a72bb23 work on abstractions4 (#15671) George Hotz 2026-04-10 13:25:11 +08:00
  • 4b960a62e2 work abstractions_4 George Hotz 2026-04-10 05:21:31 +00:00
  • 20a3783d14 cleanup George Hotz 2026-04-10 05:12:35 +00:00
  • ea358fdca6 RAND George Hotz 2026-04-10 04:51:51 +00:00
  • bf894a8285 assembly works George Hotz 2026-04-10 04:46:30 +00:00
  • 2de0f6a9aa offst George Hotz 2026-04-10 04:23:30 +00:00
  • 0645e48de3 works George Hotz 2026-04-10 04:13:50 +00:00
  • 8239301df0 work on abstractions4 George Hotz 2026-04-10 11:32:11 +08:00
  • dbc23e8a1b move HCQ_VISIBLE_DEVICES into DEV (#15668) Christopher Milan 2026-04-09 19:01:35 -07:00
  • fa02105546 hotfix: pin amd isa xml version George Hotz 2026-04-10 06:47:00 +08:00
  • 057dc173ab beam uop (#15660) nimlgen 2026-04-09 19:13:03 +03:00
  • 0ff30b003d am: reset queues from spi (#15664) nimlgen 2026-04-09 18:25:50 +03:00