mlx - Browse /v0.32.1 at SourceForge.net

The interactive file manager requires Javascript. Please enable it or use sftp or scp.
You may still browse the files here.

Name	Modified	Size	InfoDownloads / Week
Parent folder
README.md	2026-08-18	32.3 kB	0
v0.32.1 source code.tar.gz	2026-08-18	4.4 MB	1
v0.32.1 source code.zip	2026-08-18	4.8 MB	0
Totals: 3 Items		9.3 MB	1

What's Changed

Fix int64 type cast error when loading GGUF metadata arrays by @danlee2002 in https://github.com/ml-explore/mlx/pull/3823
Document default values in normalization layer docstrings by @Pablosinyores in https://github.com/ml-explore/mlx/pull/3819
Warn at configure time when NAX kernels are disabled by @pierre427 in https://github.com/ml-explore/mlx/pull/3824
[CUDA][Improvement] RMSNorm forward speed up by @nastya236 in https://github.com/ml-explore/mlx/pull/3850
Fix captured random state in compile by @angeloskath in https://github.com/ml-explore/mlx/pull/3828
Fix JIT preamble header filter matching project paths containing "Xcode" by @apocryphx in https://github.com/ml-explore/mlx/pull/3873
Document default value of p in dropout layer docstrings by @ayaangazali in https://github.com/ml-explore/mlx/pull/3870
[WIP] [CUDA] fsdp by @nastya236 in https://github.com/ml-explore/mlx/pull/3768
Fix triplet_loss docstring to document the reduced output shape by @ayaangazali in https://github.com/ml-explore/mlx/pull/3884
Zero-copy CPU import: mx.array(host_buffer, copy=False) on unified memory by @HaoXuAI in https://github.com/ml-explore/mlx/pull/3872
Round MLX_SDPA_BLOCKS up to a multiple of 32 by @pierre427 in https://github.com/ml-explore/mlx/pull/3875
Reuse Metal WAR tracking hash tables by @neilmehta24 in https://github.com/ml-explore/mlx/pull/3882
Use unroll_count(4) for the NAX attention Q@K.T loop by @wyanzhao in https://github.com/ml-explore/mlx/pull/3843
metal: add gemv_wide for fp16/bf16 matmuls of a few rows by @jessegross in https://github.com/ml-explore/mlx/pull/3888
Fix broken docstring rendering in Linear and RNN by @ayaangazali in https://github.com/ml-explore/mlx/pull/3890
Fix Adamax betas docstring and MultiOptimizer filters type by @ayaangazali in https://github.com/ml-explore/mlx/pull/3889
Fix incorrect nvfp4 quantized_matmul through the split-K path by @metascroy in https://github.com/ml-explore/mlx/pull/3854
[Metal] Avoid regex in custom kernel name generation by @aleroot in https://github.com/ml-explore/mlx/pull/3869
Fix prod dtype promotion when reducing a size-1 axis by @eyupcanakman in https://github.com/ml-explore/mlx/pull/3898
[CUDA] Fix grid overflow in gemm conv unfold kernels for >= 65,536 output positions by @AdamDLuz in https://github.com/ml-explore/mlx/pull/3893
[CUDA] columnwise quantize with tma by @nastya236 in https://github.com/ml-explore/mlx/pull/3157
metal: reduce NVFP4 scales per 16-lane group by @jessegross in https://github.com/ml-explore/mlx/pull/3934
Update homebrew in CI by @angeloskath in https://github.com/ml-explore/mlx/pull/3946
Make index autodiff errors explicitly recommend stop_gradient by @dogukanveziroglu in https://github.com/ml-explore/mlx/pull/3820
Making JACCL coordinator optional by @angeloskath in https://github.com/ml-explore/mlx/pull/3899
Fix docstring mismatches in the Python bindings by @ayaangazali in https://github.com/ml-explore/mlx/pull/3948
Raise a clear error for an invalid quantization mode in nn layers by @ayaangazali in https://github.com/ml-explore/mlx/pull/3914
Fix incorrect examples and outputs in the usage docs by @ayaangazali in https://github.com/ml-explore/mlx/pull/3956
Skip test_gather_qmm_sorted cpu test on M1 mac by @zcbenz in https://github.com/ml-explore/mlx/pull/3973
Fixes an axis mismatch bug in matrix norm for case -1 and 1 by @danlee2002 in https://github.com/ml-explore/mlx/pull/3827
Fix BatchNorm running variance estimator by @ishtihoss in https://github.com/ml-explore/mlx/pull/3817
Fix build error caused by TMA macro guard by @zcbenz in https://github.com/ml-explore/mlx/pull/3988
Fix Glorot/He uniform init docstrings to label the uniform bound, not sigma by @vineethsaivs in https://github.com/ml-explore/mlx/pull/3831
Fix custom metal kernel cache collision for same name, different source by @katlun-lgtm in https://github.com/ml-explore/mlx/pull/3833
Fix log_cosh_loss docstring to document the element-wise loss by @winklemad in https://github.com/ml-explore/mlx/pull/3846
Fix step activation docstring to match >= threshold behavior by @ayaangazali in https://github.com/ml-explore/mlx/pull/3902
docs: document the reduced-precision float32 default and MLX_ENABLE_TF32 by @stoyoda0012-cyber in https://github.com/ml-explore/mlx/pull/3894
Fix InstanceNorm Shape docstring to require at least 3 dimensions by @ayaangazali in https://github.com/ml-explore/mlx/pull/3903
Fix filter_and_map docstring argument order for filter_fn and is_leaf_fn by @ayaangazali in https://github.com/ml-explore/mlx/pull/3906
Export C++20 requirement to CMake consumers by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/3971
Fix implicit thread address space qualifier becoming explicit in metal 4.1 by @louen in https://github.com/ml-explore/mlx/pull/3963
Fix Transformer ignoring a custom encoder or decoder with no parameters by @ayaangazali in https://github.com/ml-explore/mlx/pull/3962
docs: remove references to removed --no-verify-script launch flag by @latent-9 in https://github.com/ml-explore/mlx/pull/3959
added eye(0) support and tests by @aaishwarymishra in https://github.com/ml-explore/mlx/pull/3952
Refactor the JACCL ring and add threads for the multiple rings by @angeloskath in https://github.com/ml-explore/mlx/pull/3900
Fix shapeless matmul with dynamic batch dimensions by @varshneydevansh in https://github.com/ml-explore/mlx/pull/3813
Fix JIT build with old macOS SDK by @metascroy in https://github.com/ml-explore/mlx/pull/3853
Fix cholesky_inv documented argument name to match the binding by @ayaangazali in https://github.com/ml-explore/mlx/pull/3950
Add new_thread_unsafe_stream to the devices and streams docs by @ayaangazali in https://github.com/ml-explore/mlx/pull/3968
Add gather_qqmm by @zcbenz in https://github.com/ml-explore/mlx/pull/3757
Fix missing printoptions doc page from autosummary filename collision by @ayaangazali in https://github.com/ml-explore/mlx/pull/3985
Document ThreadLocalStream and iinfo in the API reference by @ayaangazali in https://github.com/ml-explore/mlx/pull/3986
Make "stop" optional in arange by @aaishwarymishra in https://github.com/ml-explore/mlx/pull/3982
Add softsign to the nn functions docs by @ayaangazali in https://github.com/ml-explore/mlx/pull/3989
Fix broken all_sum and Group references in the data parallelism example by @ayaangazali in https://github.com/ml-explore/mlx/pull/3996
Fix ast.metal_kernel typo in the custom Metal kernels guide by @ayaangazali in https://github.com/ml-explore/mlx/pull/3997
Fix unresolved mx.array docstring references by @ayaangazali in https://github.com/ml-explore/mlx/pull/3990
python: fix bfloat16 buffer format itemsize mismatch by @reckylurker in https://github.com/ml-explore/mlx/pull/3975
Template Metal complex scalar lanes by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/3970
Pad 2D conv input channels to reach the specialized Metal kernel by @eyupcanakman in https://github.com/ml-explore/mlx/pull/3904
Support head dimension 96 in Metal full attention by @dhiltgen in https://github.com/ml-explore/mlx/pull/3943
Fix mx.remainder floored-mod for float16/bfloat16 on CPU by @sashko-zakharchuk in https://github.com/ml-explore/mlx/pull/3976
Fix sorted gather_mm activation row stride by @metascroy in https://github.com/ml-explore/mlx/pull/3960
Fix state corruption when a primitive throws during eval by @WindChimeRan in https://github.com/ml-explore/mlx/pull/3675
added complex support by @aaishwarymishra in https://github.com/ml-explore/mlx/pull/3984
Validate freeze and unfreeze keys against the whole model when recursing by @ayaangazali in https://github.com/ml-explore/mlx/pull/3966
Fix mlx.launch --python: flag is parsed but never forwarded to the launch script by @jonathan308 in https://github.com/ml-explore/mlx/pull/4002
Fix broken fully_shard reference in FullyShardedModule docstring by @ayaangazali in https://github.com/ml-explore/mlx/pull/4007
Use the current interpreter in the comparative benchmark runner by @ayaangazali in https://github.com/ml-explore/mlx/pull/4016
Document MLX environment variables by @XXXXRT666 in https://github.com/ml-explore/mlx/pull/4000
Fix CUDA batched GEMV grid overflow by @jasp-nerd in https://github.com/ml-explore/mlx/pull/3929
Fix all_gather benchmark collapsing its input to a scalar by @ayaangazali in https://github.com/ml-explore/mlx/pull/4017
Check threadgroup size in the 1-pass sdpa_vector dispatch by @apocryphx in https://github.com/ml-explore/mlx/pull/4018
Align example projects with the Python 3.10 minimum by @ayaangazali in https://github.com/ml-explore/mlx/pull/4024
Fix C++ benchmarks failing to build on overloaded astype by @ayaangazali in https://github.com/ml-explore/mlx/pull/4025
Fix conv_transpose maxBufferLength failures on Metal via tiled unfold by @eyupcanakman in https://github.com/ml-explore/mlx/pull/3845
Give each host a unique rank in Hostfile.from_list by @ayaangazali in https://github.com/ml-explore/mlx/pull/4027
Fix crash when reporting partial rings in mlx.distributed_config by @ayaangazali in https://github.com/ml-explore/mlx/pull/4026
docs: Do not pass MLX_METAL_FAST_SYNCH=1 by default by @katlun-lgtm in https://github.com/ml-explore/mlx/pull/4005
Fix signed-integer overflow in convolution shape arithmetic by @eyupcanakman in https://github.com/ml-explore/mlx/pull/3938
Template Metal C2C FFT scalar lanes by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/3969
Fix segfault in expand_dims for out of bounds negative axes by @Gusanidas in https://github.com/ml-explore/mlx/pull/4021
Add optional dtype parameter to zeros_like and ones_like by @reckylurker in https://github.com/ml-explore/mlx/pull/4028
Fix mx.longsumexp output shape issue. by @JasonHonKL in https://github.com/ml-explore/mlx/pull/4030
Treat backend and envs as optional when parsing a hostfile by @ayaangazali in https://github.com/ml-explore/mlx/pull/4039
Report bad hostfiles as CLI errors in mlx.launch by @ayaangazali in https://github.com/ml-explore/mlx/pull/4040
Fix ceil error message to say ceil instead of floor by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4042
Fix installed static MLX package on Windows by @yingjiacai in https://github.com/ml-explore/mlx/pull/3848
Derive the qmv fast path K alignment from bits by @jlee5814 in https://github.com/ml-explore/mlx/pull/3965
Fix eigh UPLO and zero-size eigh/svd on the CPU by @kapellirohith in https://github.com/ml-explore/mlx/pull/3834
Add metadata to exported functions by @eyupcanakman in https://github.com/ml-explore/mlx/pull/3855
Fix empty matrix case in cholesky by @prady0t in https://github.com/ml-explore/mlx/pull/4033
Reject a negative max_norm in clip_grad_norm by @ayaangazali in https://github.com/ml-explore/mlx/pull/4058
Build extensions with the interpreter running the build by @ayaangazali in https://github.com/ml-explore/mlx/pull/4057
Reject negative dimensions in broadcast_to and random shapes by @ayaangazali in https://github.com/ml-explore/mlx/pull/4046
Report bad host arguments as CLI errors instead of tracebacks by @ayaangazali in https://github.com/ml-explore/mlx/pull/4045
Fix Metal FFT for sizes above 2**20 by @adityasingh2400 in https://github.com/ml-explore/mlx/pull/4013
Request MSL 4.1 from the runtime compiler on macOS 27 by @magnuslundstedt in https://github.com/ml-explore/mlx/pull/4052
Fix concurrent Metal kernel cache lookup by @kitty-eu-org in https://github.com/ml-explore/mlx/pull/4043
Refactor CPU ArgReduce dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4066
Refactor NumberOfElements dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4064
Refactor CPU Select dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4065
Raise qmv batch limit for large matrices on M5-class GPUs by @pierre427 in https://github.com/ml-explore/mlx/pull/3791
Enable half-precision complex Metal FFT kernels by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/3981
Fix: mx.distributed.sum_scatter crashes on a scalar(0-d) by @JasonHonKL in https://github.com/ml-explore/mlx/pull/4071
Return an empty result from pinv for zero-size inputs by @ayaangazali in https://github.com/ml-explore/mlx/pull/4069
Handle zero-size inputs in eig and qr on the CPU by @ayaangazali in https://github.com/ml-explore/mlx/pull/4068
Normalize biases before encoding in gather_qmm_rhs by @dudududukim in https://github.com/ml-explore/mlx/pull/4056
Keep randint samples inside [low, high) by @adityasingh2400 in https://github.com/ml-explore/mlx/pull/4012
Allow saving empty arrays to npy and safetensors by @ayaangazali in https://github.com/ml-explore/mlx/pull/4080
Fix DeviceType annotations for device and stream arguments by @XXXXRT666 in https://github.com/ml-explore/mlx/pull/4059
Refactor CPU DivMod dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4088
Refactor CPU SegmentedMM dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4087
Refactor CPU Arange dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4085
Fix Device and Stream lexicographic ordering by @wgu9 in https://github.com/ml-explore/mlx/pull/4086
Enable complex64 scatter addition on GPU by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4078
Optimize large NVFP4 QMV on M5 Max by @dhiltgen in https://github.com/ml-explore/mlx/pull/3961
ring: fail on peer disconnect instead of hanging forever by @erwinzhang7 in https://github.com/ml-explore/mlx/pull/4060
Fix: propagate NaN when dealing with cummax and curmin by @JasonHonKL in https://github.com/ml-explore/mlx/pull/4047
Refactor CPU equal_nan dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4095
Refactor CPU Matmul dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4103
Refactor CPU Gather dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4105
Refactor generic CPU unary dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4106
Refactor generic CPU binary dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4107
Validate pooling kernel size, stride and padding by @ayaangazali in https://github.com/ml-explore/mlx/pull/4101
Only reject max and min when a reduced axis is empty by @ayaangazali in https://github.com/ml-explore/mlx/pull/4079
Fix SinusoidalPositionalEncoding silently overriding explicit scale=0.0 by @YassineMA03 in https://github.com/ml-explore/mlx/pull/4098
Fix any and all treating -0.0 as nonzero by @ayaangazali in https://github.com/ml-explore/mlx/pull/4090
Fix Eigh CPU dtype error message by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4130
Fix random distribution dtype error messages by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4131
Fix safetensors dtype error formatting by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4132
Fix no-GPU synchronize error formatting by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4133
Fix: mx.isinf doens't consider imaginary plane. by @JasonHonKL in https://github.com/ml-explore/mlx/pull/4092
Refactor CPU Copy destination dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4104
Refactor CPU GatherAxis dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4109
Refactor inexact CPU unary dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4111
Refactor CPU comparison dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4112
Replace Ring's local dtype macro with common dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4120
Refactor CPU ArgSort dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4123
Refactor CPU Softmax dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4127
Use common dtype dispatch for Metal Arange by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4128
Pick BM from rows per expert in gather_qmm_rhs_nax by @dwijenpatel in https://github.com/ml-explore/mlx/pull/4023
Enable complex64 take_along_axis backward on Metal by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4094
Add mx.searchsorted with CPU, Metal and CUDA kernels by @erwinzhang7 in https://github.com/ml-explore/mlx/pull/4035
Stop integer power from hanging on a negative exponent by @ayaangazali in https://github.com/ml-explore/mlx/pull/4100
Fix second order gradients for sort, partition, topk and cummax/cummin by @Adityaj0 in https://github.com/ml-explore/mlx/pull/4117
Fix segfault on negative out of bounds axes in take_along_axis/put_along_axis by @Adityaj0 in https://github.com/ml-explore/mlx/pull/4118
Deduplicate slow CPU convolution dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4126
Use common dispatch for compiled constants by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4129
Fix set_printoptions example by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4140
Clarify export output type diagnostic by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4141
Fix integer keys collision in tree_unflatten by @Solaris-star in https://github.com/ml-explore/mlx/pull/3878
Fix fence tracking for donated dynamic slice offsets by @michaelellis003 in https://github.com/ml-explore/mlx/pull/4099
Deduplicate distributed all-reduce construction by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4144
Deduplicate distributed socket errors by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4145
Remove redundant Metal bitwise dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4147
Deduplicate custom Metal kernel size validation by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4149
Deduplicate CUBLAS compute type cases by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4151
Refactor CPU Scatter dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4155
fix(cpu): name the right routine in LUF and SVD error messages by @devteamaegis in https://github.com/ml-explore/mlx/pull/4167
Refactor CPU ArgPartition dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4156
add complex to scalar to avoid raising warnings by linters by @aaishwarymishra in https://github.com/ml-explore/mlx/pull/4168
Refactor CPU Copy source dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4170
Deduplicate Metal scan encoder setup by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4154
[Metal] Skip empty NAX GEMM output groups by @XXXXRT666 in https://github.com/ml-explore/mlx/pull/3941
docs: fix stale CLI invocations in LLM inference example by @deBrian07 in https://github.com/ml-explore/mlx/pull/4172
Increase ccache size in CI by @zcbenz in https://github.com/ml-explore/mlx/pull/3999
Fix hadamard_transform on the GPU for n = m with no power-of-2 factor by @axiom-of-choice in https://github.com/ml-explore/mlx/pull/4054
Refactor CPU Partition dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4175
Refactor CPU ScatterAxis and MaskedScatter dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4176
docs: document softmax's precise argument by @apocryphx in https://github.com/ml-explore/mlx/pull/4178
Send out of range trig arguments to libm by @ayaangazali in https://github.com/ml-explore/mlx/pull/4157
Add reflect and symmetric padding modes to mx.pad by @katlun-lgtm in https://github.com/ml-explore/mlx/pull/3608
docs: add a usage page on growing arrays by @apocryphx in https://github.com/ml-explore/mlx/pull/4019
Fix and enable non-transposed NAX qmm by @gordofreemo in https://github.com/ml-explore/mlx/pull/4051
Add scatter reduce for JACCL by @angeloskath in https://github.com/ml-explore/mlx/pull/3901
Fix diag for zero-size input by @devteamaegis in https://github.com/ml-explore/mlx/pull/4165
Add an inverse-CDF path to categorical sampling by @deescondrillas in https://github.com/ml-explore/mlx/pull/4177
Fail when a requested RDMA device is not found by @erwinzhang7 in https://github.com/ml-explore/mlx/pull/4180
Bound GGUF tensor data offsets against the file mapping by @robertomeroni in https://github.com/ml-explore/mlx/pull/4179
Shift by the max in log_softmax by @ayaangazali in https://github.com/ml-explore/mlx/pull/4169
Shift by the max in cross_entropy by @ayaangazali in https://github.com/ml-explore/mlx/pull/4188
Fix linalg.norm(x, keepdims=True) not keeping dims by @devteamaegis in https://github.com/ml-explore/mlx/pull/4166
Refactor inexact CPU binary dtype dispatch by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4194
Implement batched matmul for large 1D dot products by @Ved235 in https://github.com/ml-explore/mlx/pull/3580
Re-disable qmm_n_nax and fix group_size < 64 by @PhilipJohnBasile in https://github.com/ml-explore/mlx/pull/4202
fix: shell-quote the working directory in the distributed launch script by @bunlongheng in https://github.com/ml-explore/mlx/pull/4189
Report when no usable GID is found by @erwinzhang7 in https://github.com/ml-explore/mlx/pull/4191
Keep double precision for python floats in float64 operations by @nileshpatil6 in https://github.com/ml-explore/mlx/pull/4173
Jaccl ring multi wire fix by @angeloskath in https://github.com/ml-explore/mlx/pull/4193
Report why creating a queue pair failed by @erwinzhang7 in https://github.com/ml-explore/mlx/pull/4209
Add antialias support to nn.Upsample for linear and cubic modes by @lyonsno in https://github.com/ml-explore/mlx/pull/3677
Require pytorch 2.12 for metal dlpack tests by @zcbenz in https://github.com/ml-explore/mlx/pull/4206
Remove std::optional<bool> in C++ public interface by @zcbenz in https://github.com/ml-explore/mlx/pull/4207
Fix SliceUpdate JVPs with one traced input by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4200
Raise instead of hanging when reflect/symmetric pad gets an empty axis by @codeAnqiang-ma in https://github.com/ml-explore/mlx/pull/4223
Reuse Stockham twiddles in large batched Bluestein FFTs by @PhysicistJohn in https://github.com/ml-explore/mlx/pull/4084
Validate decay_steps and step_size in the schedulers by @ayaangazali in https://github.com/ml-explore/mlx/pull/4217
Add runnable distributed examples by @erwinzhang7 in https://github.com/ml-explore/mlx/pull/4219
Break monolithic MTLResidencySet into smaller sets by @JamesMcCarthy44 in https://github.com/ml-explore/mlx/pull/4211
[CUDA][Improvement] RMSNorm backward by @nastya236 in https://github.com/ml-explore/mlx/pull/3881
Fix MultiOptimizer on models containing empty modules by @erenmenges in https://github.com/ml-explore/mlx/pull/4215
Return tuple in meshgrid by @JasonHonKL in https://github.com/ml-explore/mlx/pull/4229
Add endpoint parameter to linspace by @AKnassa in https://github.com/ml-explore/mlx/pull/4184
Fix vmap of partition/argpartition dropping the kth argument by @Adityaj0 in https://github.com/ml-explore/mlx/pull/4116
Fix nan_to_num replacing inf with 0 for float16 and bfloat16 by @codeAnqiang-ma in https://github.com/ml-explore/mlx/pull/4222
Fix einsum not broadcasting batch dimensions in batched tensordot by @Adityaj0 in https://github.com/ml-explore/mlx/pull/4125
Dequantize in float32 by @RohanGautam in https://github.com/ml-explore/mlx/pull/4241
Reject complex in erf and erfinv by @ayaangazali in https://github.com/ml-explore/mlx/pull/4243
Make abs a no-op for unsigned types in the simd helpers by @ayaangazali in https://github.com/ml-explore/mlx/pull/4240
Fix quantize matrix multiplication floor issue by @JasonHonKL in https://github.com/ml-explore/mlx/pull/4251
Only use MPI backend for world size > 1 by @erwinzhang7 in https://github.com/ml-explore/mlx/pull/4210
Reject complex in expm1, sigmoid and arctan2 by @ayaangazali in https://github.com/ml-explore/mlx/pull/4257
Decompose small kernel-depth 3D convs into 2D convs by @katlun-lgtm in https://github.com/ml-explore/mlx/pull/3785
Fix Metal sort of a view with a negative stride by @robertomeroni in https://github.com/ml-explore/mlx/pull/4252
Mirror the depth axis in the decomposed 3D conv when flipped by @ayaangazali in https://github.com/ml-explore/mlx/pull/4277
Fix Metal row reductions on negative-stride views by @FU-max-boop in https://github.com/ml-explore/mlx/pull/4267
[CUDA] Fix custom kernel cache collision for same name, different source by @HaoXuAI in https://github.com/ml-explore/mlx/pull/4273
Fix ops rejecting integers larger than INT32_MAX by @FeliGame in https://github.com/ml-explore/mlx/pull/4255
Fix var/std for complex numbers by @ayaangazali in https://github.com/ml-explore/mlx/pull/4260
Fix int32 overflow in conv padded input and pad shapes by @eyupcanakman in https://github.com/ml-explore/mlx/pull/4258
Reject complex in remainder by @ayaangazali in https://github.com/ml-explore/mlx/pull/4270
Compare the macOS SDK version as a version when gating JACCL by @erwinzhang7 in https://github.com/ml-explore/mlx/pull/4286
Clamp ring socket transfers so a payload of 2 GiB or more can be sent by @erwinzhang7 in https://github.com/ml-explore/mlx/pull/4281
Replace hand-rolled axis normalization with normalize_axis_index in split/unstack/partition/topk by @Adityaj0 in https://github.com/ml-explore/mlx/pull/4288
Remove grouped output in CI by @zcbenz in https://github.com/ml-explore/mlx/pull/4195
[CUDA] Fix finding cuda 13 headers in JIT compilation by @zcbenz in https://github.com/ml-explore/mlx/pull/3995
Refactor wheel building script by @zcbenz in https://github.com/ml-explore/mlx/pull/3818
Make mx.compile cache erasing thread safe by @zcbenz in https://github.com/ml-explore/mlx/pull/4248
Add builds for free-threaded python by @zcbenz in https://github.com/ml-explore/mlx/pull/3812
Fix int32 overflow in concatenate/repeat/kron by @ayaangazali in https://github.com/ml-explore/mlx/pull/4303
Widen list elements that do not fit in int32 to int64 by @ayaangazali in https://github.com/ml-explore/mlx/pull/4305
Propagate CPU errors to events by @zcbenz in https://github.com/ml-explore/mlx/pull/3742
Fix mx.arange dtype inference overflow regression by @shoutoutuoadi325 in https://github.com/ml-explore/mlx/pull/4324
Add workflow to update pull request limit bypass list by @zcbenz in https://github.com/ml-explore/mlx/pull/4320
Support head dimension 72 in Metal full attention by @dhiltgen in https://github.com/ml-explore/mlx/pull/4330