L-electron-Rare/mlx
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Commit Graph

  • 73f33e92cd fix(ci): use branch_to_build input for checkout and artifact naming main electron-rare 2026-05-10 13:41:03 +02:00
  • 765ebd7ef2 ci: add macOS arm64 wheel build workflow electron-rare 2026-05-10 13:13:24 +02:00
  • 84961223c0 [CUDA] Separate main loop into a function in qmm (#3443) Cheng 2026-05-09 11:11:25 +09:00
  • 662115c1f0 Do not use prebuilt cpu compile preamble when headers are installed (#3463) Cheng 2026-05-09 09:02:12 +09:00
  • a1c0b6f9ac Compute contiguity from the actual occupied data (#3475) Valeriy Sofin 2026-05-08 11:29:29 +03:00
  • bf6421bfb0 Add test fix-rope Angelos Katharopoulos 2026-05-08 00:25:27 -07:00
  • b545a35b9b Fix rope Angelos Katharopoulos 2026-05-08 00:17:59 -07:00
  • 2ecf184f91 Add a test qvm-split-k Angelos Katharopoulos 2026-05-07 23:53:34 -07:00
  • 1ea24e11f0 Fix qvm split k with batch dim Angelos Katharopoulos 2026-05-07 23:50:12 -07:00
  • c9aa560577 Make device_count() return 0 when there is no GPU (#3486) Cheng 2026-05-08 08:33:55 +09:00
  • ec59531c02 Add free threaded build test/free-threading Cheng 2026-05-06 10:24:29 +09:00
  • ff57d875ea Fix indexing bug in slice update with op (#3483) Angelos Katharopoulos 2026-05-06 17:04:50 -07:00
  • 1091e3dd0a Use uv in macOS CI test/mac-uv Cheng 2026-05-06 11:57:01 +09:00
  • 80bcd1c658 [CUDA] Fix half type matmul in cutlass kernels (#3469) Cheng 2026-05-06 08:35:53 +09:00
  • 1fdd4e23c2 Clearer error when shape dimension overflows int32 (#3425) serenposh 2026-05-05 06:23:36 +05:30
  • b43965925f Define ST_F8_E8M0 (#3448) Pedro Cuenca 2026-05-05 02:22:23 +02:00
  • 0938db7e54 Add determinant and sign-log-determinant functions to mlx.core.linalg (#3416) Abhilash Shankarampeta 2026-05-04 17:06:23 -07:00
  • e8ebdebeeb Add barrier to JACCL (#3459) Irakli Salia 2026-04-28 20:39:56 +04:00
  • d7d0992d75 Reuse nightly build's ccache for release (#3458) Cheng 2026-04-28 10:54:41 +09:00
  • bdb6ff8881 Keep gguflib input-validation asserts active in release builds (#3436) Kimon N. 2026-04-27 01:46:57 +02:00
  • 894c948773 [CUDA] Fix qmm_naive K-tail dispatch for FP quantized kernels (#3445) Long Yixing 2026-04-27 07:40:14 +08:00
  • 73a6e13b18 Start fixing ring jaccl-sum-scatter Angelos Katharopoulos 2026-04-22 11:38:02 -07:00
  • 892fdd1207 Add reduce scatter cpu implementation Angelos Katharopoulos 2026-04-22 11:38:02 -07:00
  • 211e57be53 Bump minor (#3438) Angelos Katharopoulos 2026-04-22 11:09:30 -07:00
  • 46aeba02e7 First pass of sum_scatter Angelos Katharopoulos 2026-04-22 10:57:35 -07:00
  • 5ae36f2c08 Tentative JACCL examples jaccl-examples Angelos Katharopoulos 2026-04-22 01:29:40 -07:00
  • c284e0a231 Enable swap for all CI building CUDA (#3437) Cheng 2026-04-22 13:13:24 +09:00
  • b9b1bfb9a5 Generate qmm implementaions with cmake (#3424) Cheng 2026-04-22 13:11:55 +09:00
  • 68cf2fddd8 Fix mx.prod vjp for complex types (#3433) v0.31.2 Cameron Churchwell 2026-04-21 19:35:20 -05:00
  • c594e6ec38 Fix use after move in Compiled primitive (#3427) Doğukan Veziroğlu 2026-04-22 01:22:45 +03:00
  • 7d40a4fd5a Throw meaningful error when Metal device is not found (#3428) Doğukan Veziroğlu 2026-04-22 01:21:08 +03:00
  • 5f519ef6f9 Fix bytes_per_key truncation in random kernels (Metal + CUDA) (#3432) Doğukan Veziroğlu 2026-04-22 01:15:11 +03:00
  • 705c828feb Fix synchronize for ThreadLocalStream (#3429) Angelos Katharopoulos 2026-04-20 11:29:49 -07:00
  • b4ddf9b374 Fix flaky TestVmap.test_vmap_masked_scatter (#3421) Cheng 2026-04-20 17:19:20 +09:00
  • 1f5a413a27 Make Scheduler::enqueue thread safe (#3423) Cheng 2026-04-20 14:30:05 +09:00
  • a6222f53d5 Speed up NAX split-K by better tuning and routing and fix NAX addmm (#3422) Angelos Katharopoulos 2026-04-19 18:05:39 -07:00
  • fa4320d5fa [CUDA] Handle residue k in qmm_naive (#3379) Cheng 2026-04-18 13:30:07 +09:00
  • 859f22fbb0 [CUDA] GatherQMM matrix-matrix sm80/naive path (#3417) Long Yixing 2026-04-18 09:59:47 +08:00
  • d142de6a20 [CUDA] gather_mm (#3414) Cheng 2026-04-17 16:53:44 +09:00
  • 940ba473fe Segmented mm nax kernel (#3419) Angelos Katharopoulos 2026-04-16 17:26:29 -07:00
  • 8e649be4d0 Fix jaccl init bug (#3418) Angelos Katharopoulos 2026-04-16 01:23:35 -07:00
  • dec6b4d10f ThreadLocalStream in C++ (#3405) Cheng 2026-04-16 07:46:11 +09:00
  • fd8e849e26 Document sort stability and NaN handling (#3400) NeuralNoble 2026-04-16 03:02:42 +05:30
  • 50ae31241a Validate safetensors data offsets against file boundaries (#3410) Matias Insaurralde 2026-04-15 18:30:55 -03:00
  • 6cef1e995e Validate safetensors data offsets (#3364) Dan Anderson 2026-04-15 03:52:42 -04:00
  • 57bcced8cb Fixes for CUDA CI (#3413) Cheng 2026-04-15 15:52:52 +09:00
  • 4400504ad5 Jaccl refactor (#3412) Angelos Katharopoulos 2026-04-14 23:52:21 -07:00
  • 1fa764fbec Update nanobind version to v2.12.0 (#3396) jrp2014 2026-04-15 01:21:00 +01:00
  • 435f0b6cdb Add clear_streams API for cleanup before exit (#3395) Cheng 2026-04-14 18:41:32 +09:00
  • 520cea2bec Avoid joining threads on exit (#3388) Cheng 2026-04-11 09:22:34 +09:00
  • a33b791615 Fix int16 overflow in SDPA NAX mask indexing for KV sequences > 32K (#3361) Clydingus 2026-04-10 15:01:47 +08:00
  • d6d9b24801 Conjugate VJP and JVP support (#3386) Cameron Churchwell 2026-04-09 17:04:46 -05:00
  • 8332e228e4 Fix test "test get streams" missing initialization (#3376) Daniil Seredkin 2026-04-09 02:29:04 +03:00
  • 4403165843 [CUDA] Thread safety (#3367) Cheng 2026-04-09 08:18:00 +09:00
  • a8776b7bbd Fix: Correct cross-attention query routing in Post-LN TransformerDecoderLayer (#3382) Shantanu Suryawanshi 2026-04-07 12:16:12 -04:00
  • b98831ad0e fix: fail build when Metal compiler header resolution fails (#3332) Doğukan Veziroğlu 2026-04-06 22:49:25 +03:00
  • d025111b1d [CUDA] Add GatherQMM for quantized gather matmul (#3321) Long Yixing 2026-04-07 03:48:18 +08:00
  • 9239808225 Fix CMake finding wrong Python during pip install (#3375) Harrison Powers 2026-04-06 15:32:16 -04:00
  • 6a9a121d09 Add a convenience for making local streams in python (#3355) Angelos Katharopoulos 2026-04-02 18:43:02 -07:00
  • befe42d303 Add printoptions (#3333) Christophe Prat 2026-04-02 07:24:48 +02:00
  • 80a1c206f9 Use metal as the front-end for the metal linker (#3354) Valentin Roussellet 2026-04-01 16:52:07 -07:00
  • b0748ad8de Fix regression in array creation (#3353) Angelos Katharopoulos 2026-04-01 11:30:36 -07:00
  • 2ffafe07f4 [CUDA] 3/5/6-bit quants for qmm_naive (#3352) Cheng 2026-04-01 20:13:01 +09:00
  • 5e2c44259f Make CommandEncoder thread local (#3348) Cheng 2026-04-01 18:42:49 +09:00
  • 1c9ee2f655 [CUDA] Fallback QMM (#3315) Cheng 2026-04-01 12:41:26 +09:00
  • 7cd73c4202 [Metal] Support sorting complex numbers (#3314) Long Yixing 2026-04-01 11:40:50 +08:00
  • 2105df91da Add fftfreq, rfftfreq and scalar axes for fftshift/ifftshift (#3298) declanhealy2 2026-04-01 02:29:16 +01:00
  • 1944cf67a2 Add vmap for BroadcastAxes (#3344) Angelos Katharopoulos 2026-03-31 17:08:56 -07:00
  • 939e425c7a Decouple CommandEncoder from Device (#3316) Cheng 2026-04-01 08:51:17 +09:00
  • 8439b1f501 Fix use after move (#3343) Angelos Katharopoulos 2026-03-31 10:37:40 -07:00
  • 117b4f1806 Bump actions/deploy-pages from 4 to 5 (#3334) dependabot[bot] 2026-03-31 09:04:19 +09:00
  • 66f58032dc Remove no longer needed const_cast (#3325) Cheng 2026-03-31 08:10:49 +09:00
  • 8a6d28713c Fix np bfloat16 misinterpreted as complex (#3146) Kellen Sun 2026-03-30 19:04:55 -04:00
  • 0ff1115a46 [CUDA] Implement BlockMaskedMM (#3299) Long Yixing 2026-03-27 05:57:26 +08:00
  • df7f7db943 Make each thread have its own default stream (#3281) Cheng 2026-03-25 15:48:49 +09:00
  • 57c813f042 Add norm parameter to FFT transforms (#3287) Sheldon Aristide 2026-03-25 00:27:40 -04:00
  • f8eda2c61b [CUDA] support sorting complex numbers (#3286) Long Yixing 2026-03-25 11:35:02 +08:00
  • 282174dd03 Manage Metal objects with smart pointers (#3282) Cheng 2026-03-25 11:19:20 +09:00
  • bd200d6267 Add output_shapes for AddMM (#3262) Pranav Hari 2026-03-25 07:16:15 +05:30
  • d01b83dfe7 Use nb::ndarray for checking arrays (#3283) Cheng 2026-03-25 10:44:54 +09:00
  • 1b1c56352a Fix moved-from shape bug in broadcast_arrays causing vmap bus error (#3310) Sheldon Aristide 2026-03-24 20:02:31 -04:00
  • e18d4e97f6 Fix vmap + floor_divide: preserve integer dtype (#3292) Robert Johansson 2026-03-25 00:10:37 +01:00
  • 9ab3913567 logo files (#3308) Ronan Collobert 2026-03-24 15:08:06 -07:00
  • 81530c261b Implement Pad::vmap (#3304) Sheldon Aristide 2026-03-24 18:02:31 -04:00
  • 604c825538 Fix stale transform copy-chain leaks (#3290) LongYinan 2026-03-24 14:15:23 -07:00
  • e40ada3fe2 [Metal] Fix depthwise conv 1D kernel name for large variant (#3289) LongYinan 2026-03-23 16:20:13 -07:00
  • 38ad257088 [Metal][Performance]: Add split-K for quantized matmul (small M) (#3120) Ziqiao-git 2026-03-20 20:15:48 -07:00
  • 70a0da6fca Use thread local storage for frontend compile cache (#3280) Cheng 2026-03-20 07:44:45 +09:00
  • 82809ebd12 Fix sort NaN handling for float16 and bfloat16 (#3269) Long Yixing 2026-03-20 06:19:41 +08:00
  • 5fa1a8d59f Support indexing with any type which implmented __index__ (#3210) AN Long 2026-03-20 06:19:08 +08:00
  • 21c11fc9b0 Create default random key lazily (#3278) Cheng 2026-03-19 20:22:52 +09:00
  • e1cbac9cf4 [CUDA] Search system-installed CUDA toolkit for headers (#3277) Cheng 2026-03-19 20:09:05 +09:00
  • c8292ea11c Merge DeviceStream into CommandEncoder (#3264) Cheng 2026-03-19 19:39:30 +09:00
  • 45af0df90b Fix repr of conv layers (#3275) Angelos Katharopoulos 2026-03-18 22:47:38 -07:00
  • dbfbc0f65a [CUDA] fp and int4 quants for qmm_sm80 (#3268) Cheng 2026-03-19 09:38:55 +09:00
  • 75f74ea9bc Fix building with CUDA toolkit 13.2 (#3273) Cheng 2026-03-19 08:31:44 +09:00
  • b41b349b67 Nax Refactor (#3271) Jagrit Digani 2026-03-18 10:26:49 -07:00
  • 7bc61cceed Slice update with operation (#3266) Angelos Katharopoulos 2026-03-18 06:18:02 -07:00
  • e353be8235 tests: harden memory leak check in test_siblings_without_eval (#3088) Ihar Hrachyshka 2026-03-17 03:03:10 -04:00
  • 1e855446b2 [CUDA] Pipelined QMM (#3255) Cheng 2026-03-17 07:10:12 +09:00