COMMITS
July 31, 2026
J
migrate backends/apple/coreml (TARGETS+BUCK both with rules) (#21429)
Jon Janzen committed
P
[executorch][cuda] Optimize short-query INT5 matvec kernels (#21505)
pytorchbot committed
R
NXP backend: Enable Hardswish with new Neutron flow (#21500)
roman-janik-nxp committed
J
Move targets.bzl to proper location (#21495)
Jon Janzen committed
V
NXP backend: added support for `aten.rsqrt` using new Neutron MLIR flow (#21009)
Vaclav Novak committed
M
Add pass to replace some ops with channels last variants. (#20788)
Martin Pavella committed
Y
Arm backend: Give SiLU its own output quantization params (#21437)
Youngsik Yang committed
J
Support bf16 delegation for fully-connected (#21493)
Jacob Stevens committed
J
Support grouped state heads in GDN attention (#21402)
Jacob Stevens committed
S
Fix XNNPACK pooling crashing on default stride and single-element kernel_size (#21489)
shresth rana committed
July 30, 2026
M
fix(mlx): forward boundary-partitioned getitem instead of indexing a Slot (#21441)
Mateusz Sluszniak committed
P
[executorch][cuda] Optimize short-query INT6 matvec kernels (#21504)
pytorchbot committed
P
[executorch][cuda] Optimize short-query INT4 matvec kernels (#21503)
pytorchbot committed
J
Fix DS-CNN Cortex-M input memory format (#21485)
Jacob Szwejbka committed
E
Arm backend: Simplify linking in example executor_runner (#21491)
Erik Lundell committed
B
Arm backend: Add serialized xlarge TOSA model suite (#21492)
Baris committed
S
Arm backend: Remove stale pre-commit hook setup (#21419)
Sangwon Ha committed
I
NXP backend: update eIQ Neutron SDK to 3.2.0 (#21438)
Irina Korchakova committed
E
Add quantized_div op (#21294)
Eli Amesefe committed
July 27, 2026
S
[executorch][devtools] Reset profiler state in ETDump tests
Stephen Jia committed
S
[executorch][llm] Make Gemma Apple test deterministic
Stephen Jia committed
July 29, 2026
J
migrate examples/models/llama/config (fbcode-only python_unittest)
Jon Janzen committed
J
migrate backends/cuda/runtime/shims/tests (fbcode-only CUDA tests)
Jon Janzen committed
J
migrate backends/aoti/slim/factory/test (fbcode-only GPU tests)
Jon Janzen committed
S
Fix linear_dq8ca_q4gsw wrong output when M > K
Sicheng Stephen Jia committed
V
NXP backend: fixed docs in `nxp-quantization.md` (#21464)
Vaclav Novak committed
R
NXP backend: Update docs for ET 1.4 release (#21466)
roman-janik-nxp committed
P
[executorch][vulkan] Add payload-bounded constant sharding (#21467)
pytorchbot committed
Y
Add Module.load overload with per-backend options (#21413)
yuweix94 committed
J
gate 5 more fbcode-only TARGETS dirs (no BUCK, no targets.bzl)
Jon Janzen committed