Skip to content

Pull requests: sonos/tract

New pull request New
36 Open 2,153 Closed
36 Open 2,153 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

metal: max and average pooling kernels
#2553 opened Aug 2, 2026 by czoli1976 Contributor Loading…
core: write the sum when SumPool does not normalize
#2552 opened Aug 2, 2026 by czoli1976 Contributor Loading…
metal: mlx's preferred implicit-GEMM conv kernel, plus grouped convolutions (1.2-75x)
#2551 opened Aug 1, 2026 by czoli1976 Contributor Loading…
1
metal: depthwise convolution via a ported MLX kernel (2-14x)
#2550 opened Aug 1, 2026 by czoli1976 Contributor Loading…
metal: implicit-GEMM convolution via a ported MLX kernel (9-66x)
#2549 opened Aug 1, 2026 by czoli1976 Contributor Loading…
2
metal: route skinny f16 matmuls through mlx gemv_wide (~2.9x on M4)
#2547 opened Aug 1, 2026 by czoli1976 Contributor Loading…
1
metal: drop the MFA triangular-constant probe test- fails on M1 Pro (macOS 26.5.2) and an M4 (macOS 26.6).
#2546 opened Aug 1, 2026 by czoli1976 Contributor Loading…
metal: fused Sdpa via ported MLX attention kernels (native GQA, decode kernel)
#2545 opened Aug 1, 2026 by czoli1976 Contributor Loading…
2
metal: split multi-axis Reduce into single-axis reduces
#2544 opened Aug 1, 2026 by czoli1976 Contributor Loading…
8
linalg/pack: transpose k-contiguous blocks instead of one store per element
#2537 opened Jul 30, 2026 by simon0191 Contributor Loading…
9
linalg/mmm: decouple parallel chunking from cache blocking
#2534 opened Jul 29, 2026 by simon0191 Contributor Loading…
10
nnef: lazy weight loading, so a pruned model never reads what it discards
#2524 opened Jul 26, 2026 by czoli1976 Contributor Loading…
6
fix(api): keep runnable state live across run() calls
#2522 opened Jul 24, 2026 by cverrier Contributor • Draft
1
cuda: add recurrent Gated DeltaNet decode kernels
#2503 opened Jul 22, 2026 by Reza2kn Contributor • Draft
10
Meet Dis-Tract 0.1.0
#2482 opened Jul 16, 2026 by czoli1976 Contributor • Draft
5
cpu decode: route f16 SDPA through FlashSdpa, fuse GQA broadcast through cache-read cast/rope
#2453 opened Jul 6, 2026 by czoli1976 Contributor Loading…
3
bench: add Pocket TTS, DeepFilterNet3 and DTLN to the CI bench suite
#2449 opened Jul 6, 2026 by czoli1976 Contributor • Draft
4
block_quant: role-aware weight quantization policy
#2433 opened Jun 30, 2026 by czoli1976 Contributor • Draft
WIP: W4A8 (Q4_0 int8-dot) path for block_quant — lane-mapped SDOT kernel + wiring
#2432 opened Jun 30, 2026 by czoli1976 Contributor • Draft
1
autobatch: add a cuda-batched runtime that micro-batches concurrent run() …
#2421 opened Jun 29, 2026 by kali Collaborator Loading…
1
RFC: fuse bias/activation epilogue through the einsum unmerge reshape (workload-dependent)
#2415 opened Jun 26, 2026 by czoli1976 Contributor • Draft
RFC: future-looking emulated ACE (AI Compute Extensions) matrix backend
#2397 opened Jun 21, 2026 by czoli1976 Contributor • Draft
metal: pack element-wise kernels into full-width threadgroups
#2365 opened Jun 13, 2026 by czoli1976 Contributor • Draft
5
linalg/arm64/sme: experimental f16 SME kernels for Apple M5/A19 Need tester with hardware
#2273 opened May 22, 2026 by czoli1976 Contributor • Draft
feat: add tract_moe_ffn operator for Mixture-of-Experts FFN
#2084 opened Mar 30, 2026 by JulienBalianSonos Collaborator Loading…
2
ProTip! Find all pull requests that aren't related to any open issues with -linked:issue.