Skip to content

Pull requests: sonos/tract

New pull request New
35 Open 2,401 Closed
35 Open 2,401 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

(perf) wgpu 8/8: a blocked matmul over repacked weights
#2821 opened Sep 9, 2026 by czoli1976 Contributor Loading…
(fix) wgpu: spread a large dispatch across two grid dimensions
#2820 opened Sep 9, 2026 by czoli1976 Contributor Loading…
1
Add missing sign checks on loader dims casts (onnx/tflite)
#2814 opened Sep 9, 2026 by li-jin-quan Contributor Loading…
2
wgpu 7/7: take a frame in, hand a mask back
#2810 opened Sep 8, 2026 by czoli1976 Contributor Loading…
wgpu 6/7: fusion — chains, epilogues and axis ops
#2809 opened Sep 8, 2026 by czoli1976 Contributor Loading…
wgpu 5/7: reduction, softmax and resize
#2808 opened Sep 8, 2026 by czoli1976 Contributor Loading…
1
wgpu 4/7: convolution, pooling and transposed convolution
#2807 opened Sep 8, 2026 by czoli1976 Contributor • Draft
1
wgpu 3/7: matrix multiplication
#2806 opened Sep 8, 2026 by czoli1976 Contributor Loading…
2
RFC: a WebGPU backend (tract-wgpu)
#2801 opened Sep 8, 2026 by czoli1976 Contributor • Draft
7
cli: glibc's allocator carries a lot of overhead on the graphs the be…
#2789 opened Sep 7, 2026 by kali Collaborator • Draft
6
(perf) core/cnn: NCHW spatial kernels (direct conv, maxpool, nearest, deconv)
#2771 opened Sep 3, 2026 by czoli1976 Contributor • Draft
34
(perf) core/einsum: do not pick AMX/SME when C's n axis is strided
#2733 opened Aug 27, 2026 by czoli1976 Contributor Loading…
1
(perf) core: fuse a squeezed unit matmul axis and a unicast Mul+Add pair
#2728 opened Aug 26, 2026 by czoli1976 Contributor Loading…
9
(fix)core/cnn: fuse ONNX QDQ Conv into int8 GEMM instead of running it in float
#2725 opened Aug 25, 2026 by czoli1976 Contributor Loading…
(perf) core,onnx: GruSeq: run a GRU as one op instead of a per-timestep scan
#2717 opened Aug 24, 2026 by czoli1976 Contributor Loading…
1
Feat/gelu f16 lut
#2687 opened Aug 20, 2026 by kali Collaborator Loading…
@kali
1
onnx: accept a MatMulNBits weight that is a constant expression
#2656 opened Aug 17, 2026 by czoli1976 Contributor Loading…
1
5
onnx: import MatMulNBits at bits=2
#2655 opened Aug 17, 2026 by czoli1976 Contributor Loading…
onnx: keep asymmetric MatMulNBits weights block-quantized
#2648 opened Aug 17, 2026 by czoli1976 Contributor Loading…
2
onnx: import GatherBlockQuantized without widening the table
#2647 opened Aug 17, 2026 by czoli1976 Contributor Loading…
4
metal: mlx's preferred implicit-GEMM conv kernel (1.2-75x)
#2551 opened Aug 1, 2026 by czoli1976 Contributor Loading…
@kali
7
nnef: lazy weight loading, so a pruned model never reads what it discards
#2524 opened Jul 26, 2026 by czoli1976 Contributor • Draft
10
Meet Dis-Tract 0.1.0
#2482 opened Jul 16, 2026 by czoli1976 Contributor • Draft
5
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
ProTip! Find all pull requests that aren't related to any open issues with -linked:issue.