Commit Graph

  • eda7a7b43e Do not join threads during process exit on Windows (#1738) Cheng 2025-03-25 22:33:08 +09:00
  • 022eabb734 Remove unused import (#1987) Chunyang Wen 2025-03-25 11:19:32 +08:00
  • aba899cef8 patch bump (#2000) v0.24.1 Awni Hannun 2025-03-24 12:47:05 -07:00
  • 6a40e1c176 Fix looping limit in causal attention (#1999) Jagrit Digani 2025-03-24 12:28:00 -07:00
  • 9307b2ab8b Fixed 32-bit platform support for distributed/ring implementation (#1996) Jesper Stemann Andersen 2025-03-24 16:08:40 +01:00
  • 522d8d3917 Added missing netinet/in.h include that fixes build on FreeBSD (#1997) Jesper Stemann Andersen 2025-03-24 16:07:34 +01:00
  • a84cc0123f promote mask when needed (#1998) Awni Hannun 2025-03-23 19:58:28 -07:00
  • f018e248cd fix(backend): Include algorithm library in Allocator (#1992) Andrey Velichkevich 2025-03-23 04:27:51 +00:00
  • cfd7237a80 fix docs (#1991) Awni Hannun 2025-03-21 19:58:53 -07:00
  • 4eef8102c9 Distributed layers (#1270) Angelos Katharopoulos 2025-03-21 13:52:17 -07:00
  • 69e4dd506b Add a ring all gather (#1985) Angelos Katharopoulos 2025-03-21 13:36:51 -07:00
  • 25814a9458 Disable mpi on version mismatch (#1989) Angelos Katharopoulos 2025-03-21 13:36:26 -07:00
  • 2a980a76ce Add stats and limit to common allocator and enable tests (#1988) Awni Hannun 2025-03-21 12:28:36 -07:00
  • d343782c8b Cross platform libmpi loading (#1975) Angelos Katharopoulos 2025-03-21 11:23:10 -07:00
  • 4e1994e9d7 move memory APIs into top level mlx.core (#1982) Awni Hannun 2025-03-21 07:25:12 -07:00
  • 65a38c452b update the formula of smooth_l1_loss (#1986) jiyzhang 2025-03-21 21:25:23 +08:00
  • 7b7e2352cd fix malloc or wait deadlock (#1976) Awni Hannun 2025-03-20 16:48:43 -07:00
  • 1177d28395 patch bump (#1981) v0.24.0 Awni Hannun 2025-03-20 15:12:22 -07:00
  • 005e7efa64 fix mask in sdpa (#1980) Awni Hannun 2025-03-20 14:53:12 -07:00
  • b42d13ec84 Update attention tests to show diff, disable array masks (#1978) Jagrit Digani 2025-03-20 14:25:38 -07:00
  • 9adcd1a650 Support fused masking in Attention (#1924) Jagrit Digani 2025-03-20 11:01:32 -07:00
  • 3c164fca8c Fix multistream GPU deadlock (#1969) Awni Hannun 2025-03-20 07:19:47 -07:00
  • 95e335db7b Update smooth_l1_loss in losses.py (#1974) jiyzhang 2025-03-20 11:19:02 +08:00
  • f90206ad74 Guard nullptr dereference (#1972) Awni Hannun 2025-03-19 16:24:10 -07:00
  • 3779150750 refactor: all use schedule (#1973) Chunyang Wen 2025-03-20 02:24:04 +08:00
  • 688e421184 only interrupt during an eval interrupt_eval Awni Hannun 2025-03-19 07:56:26 -07:00
  • 9ffe88841c interruptable eval Awni Hannun 2025-03-18 17:23:31 -07:00
  • 0a9777aa5c Do not define MLX_VERSION globally (#1966) Cheng 2025-03-18 23:12:40 +09:00
  • 45ad06aac8 Fix typo; Fix lint warning when reuse the same name (#1968) Chunyang Wen 2025-03-18 22:12:24 +08:00
  • c6ea2ba329 Use same accumulation precision in gemv as gemm (#1962) Awni Hannun 2025-03-16 07:13:24 -07:00
  • 2770a10240 fix grad with inplace updates (#1961) Awni Hannun 2025-03-13 19:13:09 -07:00
  • d2a94f9e6a Only compile warnings as errors for circle (#1957) Awni Hannun 2025-03-12 13:08:19 -07:00
  • 32da94507a fix vmap for flatten (#1955) Awni Hannun 2025-03-11 10:42:22 -07:00
  • 736a340478 reduce binary size (#1952) Awni Hannun 2025-03-11 06:30:44 -07:00
  • 117e1355a2 fix copy for large arrays (#1953) Awni Hannun 2025-03-10 15:04:25 -07:00
  • 3c3e558c60 Support transposed head/seq for kv (#1950) Awni Hannun 2025-03-10 10:53:45 -07:00
  • cffceda6ee Add type hint for _extra_repr (#1948) Chunyang Wen 2025-03-10 21:05:36 +08:00
  • 048805ad2c Remove unused modules (#1949) Chunyang Wen 2025-03-10 21:05:26 +08:00
  • d14c9fe7ea Add file info when raising errors in save (#1943) Chunyang Wen 2025-03-09 06:51:04 +08:00
  • 5db90ce822 Fix obsured warning (#1944) Chunyang Wen 2025-03-09 06:50:39 +08:00
  • d699cc1330 Fix unreachable warning (#1939) Chunyang Wen 2025-03-08 09:23:04 +08:00
  • c4230747a1 redesign for faster cpu/gpu synch (#1869) Awni Hannun 2025-03-06 19:23:38 -08:00
  • 5245f12a46 always use json (#1938) Awni Hannun 2025-03-06 15:35:56 -08:00
  • a198b2787e Remove unused modules (#1936) Chunyang Wen 2025-03-07 06:20:27 +08:00
  • 04edad8c59 Add doc string for path (#1937) Chunyang Wen 2025-03-07 06:20:09 +08:00
  • 392b3060b0 Fix typo in randint docstring (#1932) David Wisdom 2025-03-05 21:48:00 -08:00
  • 85b34d59bc Clean unused sys (#1929) Chunyang Wen 2025-03-06 05:48:03 +08:00
  • f599c11bc8 bump (#1931) v0.23.2 Awni Hannun 2025-03-05 13:16:53 -08:00
  • 0792ff02ff Only fail when 10 consecutive socket errors occur (#1928) Angelos Katharopoulos 2025-03-05 13:16:19 -08:00
  • fd0d63ba5b Affine quant always in fp32 (#1925) Alex Barron 2025-03-05 01:50:19 +00:00
  • 3835a428c5 Adds nuclear norm support (#1894) Abe Leininger 2025-03-04 15:26:02 -06:00
  • 9680f72cca Add a multi optimizer (#1916) Angelos Katharopoulos 2025-03-04 13:16:35 -08:00
  • a0737273d3 Allow debugging in distributed mode (#1920) Angelos Katharopoulos 2025-03-04 13:01:10 -08:00
  • e613d0eaf0 SDPA support for small batch (over sequence) queries (#1922) Awni Hannun 2025-03-04 10:59:04 -08:00
  • 6bcd6bcf70 fix donation in scan (#1917) Awni Hannun 2025-03-03 11:30:59 -08:00
  • ba12e4999a Use a heap for small sizes (#1911) Awni Hannun 2025-03-03 06:50:57 -08:00
  • 4e7cd31d12 Fix slice data size (#1913) Awni Hannun 2025-03-02 21:50:42 -08:00
  • 5e6c130d93 RMS norm without scaling (#1915) Angelos Katharopoulos 2025-02-28 20:26:57 -08:00
  • 5d68082881 Ring docs (#1829) Angelos Katharopoulos 2025-02-28 11:34:21 -08:00
  • 607181644f Add mlx.distributed_config script (#1902) Angelos Katharopoulos 2025-02-28 11:16:39 -08:00
  • 89d327075f Enabling fused attention for head dim 128 (#1899) Jagrit Digani 2025-02-26 10:02:06 -08:00
  • 6bf00ef631 Fix ring of 2 and allow scalars in API (#1906) Angelos Katharopoulos 2025-02-25 17:03:01 -08:00
  • 7d042f17fe Double for lapack (#1904) Awni Hannun 2025-02-25 11:39:36 -08:00
  • 28b8079e30 fix double type promotion (#1901) Awni Hannun 2025-02-25 06:00:53 -08:00
  • 7face5d9fd fix cpu compile (#1897) Awni Hannun 2025-02-24 14:10:30 -08:00
  • a44dc4bdb0 fix leaking objc (#1898) Awni Hannun 2025-02-24 13:57:59 -08:00
  • 2d0f384b6f fix simd erf_inv (#1896) Awni Hannun 2025-02-24 13:57:47 -08:00
  • 8ff84b5c43 fix version and expose command queue getter (#1892) Awni Hannun 2025-02-20 15:25:15 -08:00
  • 10b271d963 Ring update (#1885) Angelos Katharopoulos 2025-02-20 14:32:31 -08:00
  • 0ebc8a3d25 Fixed issue where Clang on FreeBSD failed to compile mlx/backend/cpu/quantized.cpp (#1890) Jesper Stemann Andersen 2025-02-20 21:02:12 +01:00
  • bbda0fdbdb Allow non-square lu (#1889) Awni Hannun 2025-02-20 08:13:23 -08:00
  • c86422bdd4 Added mlx::core::version() returning std::string(MLX_VERSION) (#1819) Jesper Stemann Andersen 2025-02-20 05:30:19 +01:00
  • c707b2b0a6 Limit compile buffers (#1887) Awni Hannun 2025-02-19 20:28:13 -08:00
  • 78ba24c37d Raise an exception in the rope op if input is integer (#1884) Angelos Katharopoulos 2025-02-19 14:43:39 -08:00
  • 1a2cb72030 Ensure linspace always contains start and stop (#1883) Angelos Katharopoulos 2025-02-19 13:53:20 -08:00
  • 344a29506e Enforce triangular matrix form in tri_inv (#1876) Abe Leininger 2025-02-19 14:42:33 -06:00
  • 71de73a668 Fix convs by reverting #1803 (#1882) v0.23.1 Angelos Katharopoulos 2025-02-18 14:36:34 -08:00
  • 4c1dfa58b7 xor op on arrays (#1875) Alex Barron 2025-02-17 00:24:53 -08:00
  • 5274c3c43f compiler warnings are errors (#1870) Awni Hannun 2025-02-17 00:07:49 -08:00
  • 1762793989 Remove unused uniform (#1867) Angelos Katharopoulos 2025-02-14 15:51:41 -08:00
  • 6cec78d8f2 bump (#1866) v0.23.0 Awni Hannun 2025-02-14 13:09:34 -08:00
  • 2dc307f2e6 Winograd Update for Small batches (#1803) Jagrit Digani 2025-02-14 13:08:13 -08:00
  • 7aea5b1895 Allow dynamic ops per buffer based on dispatches and memory (#1864) Awni Hannun 2025-02-13 19:18:22 -08:00
  • 9733e16496 fix function pointer (#1865) Ronan Collobert 2025-02-13 18:46:11 -08:00
  • 7f2d1024f3 add f8_e4m3 loading (#1859) Alex Barron 2025-02-13 17:10:03 -08:00
  • 428f589364 Revert "More buffer donation in some cases (#1858)" (#1863) Awni Hannun 2025-02-13 14:21:44 -08:00
  • 5cd97f7ffe Bitwise Inverse (#1862) Alex Barron 2025-02-13 08:44:14 -08:00
  • e425dc00c0 Faster small batch qmv (#1861) Awni Hannun 2025-02-12 22:02:36 -08:00
  • d274ae77f2 More buffer donation in some cases (#1858) Awni Hannun 2025-02-12 19:41:37 -08:00
  • 55c5ac7820 fix int64 bug (#1860) Alex Barron 2025-02-12 19:23:46 -08:00
  • 0145911bea Fixes output donation for IO ops on the GPU (#1857) Angelos Katharopoulos 2025-02-12 10:52:30 -08:00
  • 0a5215693e Fix grad copies (#1854) Awni Hannun 2025-02-11 15:26:42 -08:00
  • 2a45056ba8 Cycle leak break (#1856) Awni Hannun 2025-02-11 14:45:02 -08:00
  • 142b77751d Fix compilation error on Windows (#1844) Cheng 2025-02-11 12:53:05 +09:00
  • a5ededf1c3 CPU LU factorization and linear solvers (#1451) Abe Leininger 2025-02-10 14:32:24 -06:00
  • 7df3f792a2 Ensure Conv2D and Conv3D's kernel sizes aren't trimmed (#1852) Franck Verrot 2025-02-10 06:27:01 -08:00
  • 9eb7d7362f Fix Split::vmap (#1845) Angelos Katharopoulos 2025-02-08 09:22:13 -08:00
  • 1c0c118f7c Fp64 on the CPU (#1843) Awni Hannun 2025-02-07 15:52:22 -08:00
  • 1a1b2108ec bump (#1840) v0.22.1 Awni Hannun 2025-02-06 11:53:24 -08:00
  • b6c6552d20 Add missing #pragma once (#1838) Jagrit Digani 2025-02-06 11:11:22 -08:00