11 Pull requests merged by 2 people

• Fix grad with inplace updates

  #1961 merged Mar 14, 2025

• Only compile warnings as errors for circle

  #1957 merged Mar 12, 2025

• Fix vmap for flatten

  #1955 merged Mar 11, 2025

• Reduce binary size

  #1952 merged Mar 11, 2025

• Fix copy for large arrays with JIT

  #1953 merged Mar 10, 2025

• Support transposed head/seq for kv

  #1950 merged Mar 10, 2025

• Add type hint for _extra_repr

  #1948 merged Mar 10, 2025

• Remove unused modules

  #1949 merged Mar 10, 2025

• Add file info when rasing errors in save

  #1943 merged Mar 8, 2025

• Fix obsured warning

  #1944 merged Mar 8, 2025

• Fix unreachable warning

  #1939 merged Mar 8, 2025

2 Pull requests opened by 1 person

• Fences must exit

  #1940 opened Mar 7, 2025

• Use same accumulation precision in gemv as gemm

  #1962 opened Mar 14, 2025

7 Issues closed by 2 people

• [BUG] running_mean and running_var of nn.BatchNorm doesn't update in value_and_grad_fn() call

  #1960 closed Mar 14, 2025

• Bulk derivative (big Jacobian) in MLX

  #1927 closed Mar 11, 2025

• Question about Pool 1D/2D/3D

  #1947 closed Mar 11, 2025

• [BUG] JIT looks for copyfloat32float32 but generates copyfloatfloat

  #1951 closed Mar 10, 2025

• [BUG] Code snippet runs 4x slower since mlx 0.19.0

  #1918 closed Mar 10, 2025

• [BUG] Trying to use mlx gputrace capture, fails

  #1946 closed Mar 10, 2025

• [BUG] Error: rope_scaling 'type' currently only supports 'linear'

  #1942 closed Mar 8, 2025

2 Issues opened by 2 people

• [BUG] Using Matrix Multiplication on Batched Arrays Leads to Different Results

  #1958 opened Mar 13, 2025

• How to run in a (Docker) container environment?

  #1945 opened Mar 8, 2025

7 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Distributed layers

  #1270 commented on Mar 14, 2025 • 7 new comments

• [Feature] Multi-Machine Support for Distributed Inference

  #1046 commented on Mar 7, 2025 • 0 new comments

• Adding Functionality to Distributed Communication

  #1930 commented on Mar 11, 2025 • 0 new comments

• scaled_dot_product_attention: allow Q x Q mask instead of just Q x K mask

  #1842 commented on Mar 12, 2025 • 0 new comments

• [BUG] Distributed inference OOMs on machines with different RAM size

  #1804 commented on Mar 12, 2025 • 0 new comments

• Adding support for the Muon Optimizer

  #1914 commented on Mar 12, 2025 • 0 new comments

• More donation take 2

  #1923 commented on Mar 11, 2025 • 0 new comments