Commit Graph

  • d2ff04a4f2 fix format (#1539) Awni Hannun 2024-10-28 18:29:14 -07:00
  • 015c247393 change wino dispatch conditoin (#1534) Awni Hannun 2024-10-28 11:13:44 -07:00
  • d3cd26820e Faster bits and bernoulli (#1535) Awni Hannun 2024-10-28 11:11:00 -07:00
  • 91f6c499d7 fix (#1529) Awni Hannun 2024-10-25 19:25:35 -07:00
  • 35e9c87ab9 patch bump (#1528) v0.19.1 Awni Hannun 2024-10-25 13:13:23 -07:00
  • 8e88e30d95 BFS graph evaluation order (#1525) Awni Hannun 2024-10-25 10:27:19 -07:00
  • 0eb56d5be0 Wired (#1510) Awni Hannun 2024-10-25 09:35:33 -07:00
  • f70764a162 Fix typo in build docs (#1522) Paul Hansel 2024-10-24 20:55:06 -07:00
  • dad1b00b13 fix (#1523) Awni Hannun 2024-10-24 19:17:46 -07:00
  • 430ffef58a [Feature] Added Sparse Initialization (#1498) Venkata Naga Aditya Datta Chivukula 2024-10-24 12:31:24 -07:00
  • 3d17077187 Add mx.array.__format__ (#1521) Alex Barron 2024-10-24 11:11:39 -07:00
  • c9b41d460f Working 64-bit scans (#1506) Angelos Katharopoulos 2024-10-24 11:05:46 -07:00
  • 32972a5924 C++20 compatibility for fmt (#1519) xnorai 2024-10-24 08:54:51 -07:00
  • f6afb9c09b Remove use of vector<const T> (#1514) Dhruv Govil 2024-10-22 16:31:52 -07:00
  • 3ddc07e936 Eigenvalues and eigenvectors (#1334) Kashif Rasul 2024-10-22 21:18:48 +02:00
  • c26208f67d Remove Hazard tracking with Fences (#1509) Awni Hannun 2024-10-21 19:33:32 -07:00
  • d15fa13daf Batched Quantized Matmul + Fast Small QMV (#1503) Alex Barron 2024-10-21 16:23:17 -07:00
  • 58a855682c v0.19.0 (#1502) v0.19.0 Awni Hannun 2024-10-18 11:55:18 -07:00
  • 92d7cb71f8 Fix compile (#1501) Awni Hannun 2024-10-18 11:06:40 -07:00
  • 50d8bed468 Fused attention for single query (#1497) Angelos Katharopoulos 2024-10-18 00:58:52 -07:00
  • 9dd72cd421 fix gumbel (#1495) Awni Hannun 2024-10-17 13:52:39 -07:00
  • 343aa46b78 No more 3.8 (#1493) Awni Hannun 2024-10-16 17:51:38 -07:00
  • b8ab89b413 Docs in ci (#1491) Awni Hannun 2024-10-15 17:40:00 -07:00
  • f9f8c167d4 fix submodule stubs (#1492) Awni Hannun 2024-10-15 16:23:37 -07:00
  • 3f86399922 Real and Imag (#1490) Awni Hannun 2024-10-15 16:23:15 -07:00
  • 2b8ace6a03 Typing the dropout. (#1479) LastWhisper 2024-10-15 21:45:46 +08:00
  • 0ab8e099e8 Fix cpu segfault (#1488) Awni Hannun 2024-10-14 16:17:03 -07:00
  • 020f048cd0 A few updates for CPU (#1482) Awni Hannun 2024-10-14 12:45:49 -07:00
  • 881615b072 Faster metal compiled kernels + some fixes (#1486) Awni Hannun 2024-10-14 12:45:38 -07:00
  • 0eef4febfd bump mac tests to use py39 (#1485) Awni Hannun 2024-10-14 10:40:32 -07:00
  • b54a70ec2d Make push button linux distribution (#1476) Awni Hannun 2024-10-14 06:21:44 -07:00
  • bf6ec92216 Make the GPU device more thread safe (#1478) Awni Hannun 2024-10-12 17:49:15 -07:00
  • c21331d47f version bump (#1477) v0.18.1 Awni Hannun 2024-10-10 13:05:17 -07:00
  • e1c9600da3 Add mx.random.permutation (#1471) Awni Hannun 2024-10-08 19:42:19 -07:00
  • 1fa0d20a30 consistently handle all -inf in softmax (#1470) Awni Hannun 2024-10-08 09:54:02 -07:00
  • 3274c6a087 Fix array is_available race cases (#1468) Awni Hannun 2024-10-07 19:13:50 -07:00
  • 9b12093739 Add the roll op (#1455) Angelos Katharopoulos 2024-10-07 17:21:42 -07:00
  • f374b6ca4d Bump nanobind to 2.2 (#1461) Awni Hannun 2024-10-07 16:52:40 -07:00
  • 0070e1db40 Fix deep recursion with siblings (#1462) Awni Hannun 2024-10-07 06:15:33 -07:00
  • 95d04805b3 Fix complex power on Metal (#1460) Awni Hannun 2024-10-06 19:58:30 -07:00
  • e4534dac17 Conv grad with groups + bugfix (#1449) Awni Hannun 2024-10-06 07:08:53 -07:00
  • fef3c4ec1d Fix mpi test in CI (#1456) Angelos Katharopoulos 2024-10-06 06:09:17 -07:00
  • 1bdc038bf9 fix argpartition + faster {arg} sorts / partitions (#1453) Awni Hannun 2024-10-03 14:21:25 -07:00
  • 5523d9c426 faster cpu indexing (#1450) Awni Hannun 2024-10-03 13:53:47 -07:00
  • d878015228 Fix normalization check_input (#1452) Angelos Katharopoulos 2024-10-03 13:26:56 -07:00
  • 5900e3249f Fix building on Linux (#1446) Cheng 2024-09-30 23:00:39 +09:00
  • bacced53d3 Fix row reduce with very few rows (#1447) Angelos Katharopoulos 2024-09-29 20:00:35 -07:00
  • 4a64d4bff1 Add support for grouped 1D convolutions to the nn API (#1444) Lucas Newman 2024-09-28 06:41:07 -07:00
  • b1e2b53c2d bump (#1445) v0.18.0 Awni Hannun 2024-09-27 13:53:02 -07:00
  • 11354d5bff Avoid io timeout for large arrays (#1442) Awni Hannun 2024-09-27 13:32:14 -07:00
  • 718aea3f1d allow take to work with integer index (#1440) Awni Hannun 2024-09-26 15:58:03 -07:00
  • 5b6f38df2b Faster cpu ops (#1434) Awni Hannun 2024-09-26 09:19:13 -07:00
  • 0b4a58699e Some overhead reductions in mx.fast.metal_kernel (#1437) Awni Hannun 2024-09-25 17:25:21 -07:00
  • 4f9f9ebb6f Faster Metal unary and binary for general case (#1431) Awni Hannun 2024-09-25 12:07:43 -07:00
  • afc9c0ec1b dtype is copy assignable (#1436) Awni Hannun 2024-09-25 12:07:13 -07:00
  • 195b429d99 Put along axis + fixe for partition grad (#1430) Awni Hannun 2024-09-23 10:03:38 -07:00
  • 2b878e9dd7 Create CITATION.cff (#1425) Luke Carlson 2024-09-20 14:39:46 -04:00
  • 67b6bf530d Optimization for general ND copies (#1421) Awni Hannun 2024-09-17 17:59:51 -07:00
  • 6af5ca35b2 feat: add cross_product (#1252) Nripesh Niketan 2024-09-17 21:12:43 +01:00
  • 4f46e9c997 More fixes for arrays with large sizes (#1405) Awni Hannun 2024-09-17 12:46:31 -07:00
  • c6739ba7f3 Faster RNN layers (#1419) Awni Hannun 2024-09-17 06:04:19 -07:00
  • 914409fef9 Data parallel helper (#1407) Angelos Katharopoulos 2024-09-16 18:17:21 -07:00
  • 8d68a3e805 remove fmt dependencies from MLX install (#1417) jjuang-apple 2024-09-16 13:32:28 -07:00
  • 6bbcc453ef avoid using find_library to make install truly portable (#1416) jjuang-apple 2024-09-16 13:21:32 -07:00
  • d5ed4d7a71 override class function (#1418) Awni Hannun 2024-09-16 13:21:04 -07:00
  • 669c27140d Chore: add pre-commit hook for cmake (#1362) Nripesh Niketan 2024-09-16 20:53:01 +01:00
  • adcc88e208 Conv cpu improvements (#1410) Max-Heinrich Laves 2024-09-16 03:45:10 +02:00
  • d6492b0163 fix clip (#1415) Awni Hannun 2024-09-14 16:09:09 -07:00
  • b3f52c9fbe ensure io/comm streams are active before eval (#1412) Awni Hannun 2024-09-14 06:17:36 -07:00
  • bd8396fad8 Fix typo in transformer docs (#1414) c0g 2024-09-14 06:05:15 -07:00
  • d0c58841d1 Patch bump (#1408) v0.17.3 Angelos Katharopoulos 2024-09-12 16:44:23 -07:00
  • 881f09b2e2 Allow querying the allocator for the buffer size (#1404) Angelos Katharopoulos 2024-09-11 21:02:16 -07:00
  • 8b30acd7eb fix module attribute set, reset, set (#1403) Awni Hannun 2024-09-11 16:30:42 -07:00
  • 02efb310ca Xcode 160 (#1384) Awni Hannun 2024-09-10 15:15:17 -07:00
  • e7e59c6f05 Fix copying scalars by adding fill_gpu (#1402) Awni Hannun 2024-09-09 15:54:08 -07:00
  • 3ae6aabe9f throw for certain cases of non captured inputs in compile (#1401) Awni Hannun 2024-09-09 14:54:31 -07:00
  • dc627dcb5e Replace the use of result_of_t with invoke_result_t (#1397) xnorai 2024-09-06 19:52:57 -07:00
  • efeb9c0f02 Transposed Convolution (#1245) Max-Heinrich Laves 2024-09-07 04:52:38 +02:00
  • ba3e913c7a Simplifications for MLX C (#1396) Awni Hannun 2024-09-06 19:16:50 -07:00
  • 7cca1727af Fix slice data size (#1394) Awni Hannun 2024-09-04 19:10:43 -07:00
  • 11371fe251 Test to prevent bugs like #1386 (#1391) Bhargav Yagnik 2024-09-04 20:24:30 -04:00
  • 41c603d48a fix jit reduce (#1395) Awni Hannun 2024-09-04 14:03:10 -07:00
  • 969337345f Fix reduce edge case (#1389) Angelos Katharopoulos 2024-09-01 21:37:51 -07:00
  • 9592766939 add std as method (#1387) Awni Hannun 2024-09-01 19:49:16 -07:00
  • 58dca7d846 Fix copy in the sort primitive (#1383) Angelos Katharopoulos 2024-08-31 08:32:14 -07:00
  • 0d302cd25b Fix compiel with byte sized constants (#1381) Awni Hannun 2024-08-30 17:24:35 -07:00
  • da691257ec Fix overflow in quantize/dequantize (#1379) Alex Barron 2024-08-30 13:32:41 -07:00
  • 1600092e92 Patch bump (#1376) v0.17.2 Angelos Katharopoulos 2024-08-29 16:54:30 -07:00
  • dba2bd1105 Even Even Faster IO (#1374) Awni Hannun 2024-08-29 16:05:40 -07:00
  • 28be4de7c2 Fix JIT reductions (#1373) Alex Barron 2024-08-28 16:39:11 -07:00
  • a6c3b38fba Async load (#1372) Awni Hannun 2024-08-28 14:21:55 -07:00
  • fcb65a3897 Even Faster I/O (#1369) Awni Hannun 2024-08-28 11:49:07 -07:00
  • 4e22a1dffe In continuation to PR1243 to solve issue #1240 (#1365) Saanidhya 2024-08-28 14:40:41 -04:00
  • 291cf40aca Some fixes to typing (#1371) Awni Hannun 2024-08-28 11:16:19 -07:00
  • bd47e1f066 Fix neon_fast_exp and add more softmax tests (#1367) Jeethu Rao 2024-08-28 07:42:42 +01:00
  • e6b223df5f Pinv (#875) Aditya Dhulipala 2024-08-27 23:06:12 -07:00
  • e64349bbdd Make eval just wait if all arrays are scheduled (#1368) Angelos Katharopoulos 2024-08-27 17:01:22 -07:00
  • cdb59faea6 Adds send/recv ops in distributed (#1366) Angelos Katharopoulos 2024-08-26 23:01:37 -07:00
  • 1d94ac3f90 Add optional headers to `mx.fast.metal_kernel` (#1358) Alex Barron 2024-08-26 21:45:45 -07:00
  • 5f7d19d1f5 MPI ops in GPU stream for faster comms (#1356) Awni Hannun 2024-08-26 15:12:50 -07:00