COMMITS
July 27, 2026
A
executorch: make CUDA backend and partitioner targets PUBLIC
Anthony Shoumikhin committed
A
tensor_parser_aten: tag planned tensor DataPtr with its real device (#21289)
Anthony Shoumikhin committed
J
migrate dirs whose runtime.python_binary kwargs are fbcode-only (#21388)
Jon Janzen committed
Y
Use const_data_ptr/mutable_data_ptr instead of deprecated data_ptr (#20958)
Yuanyuan Chen committed
M
Fix optimized binary op broadcast rank mismatch (#21052)
Maya Bennett committed
Y
Arm backend: Remove statically proven TOSA no-ops (#21141)
Yufeng Shi committed
J
Add timeout default in NeutronBackend (#21293)
Jacob Stevens committed
B
Arm backend: Legalize unsigned RESCALE from wide inputs (#21393)
Baris committed
Y
Reland "Arm backend: Rerun duplicate-user fusion after TOSA lowering" (#21380)
Yufeng Shi committed
R
Make agent instruction for attribution broader. (#21139)
Rob Elliott committed
R
Arm backend: Normalize grid-sampler custom inputs to NHWC (#20965)
Rob Elliott committed
G
Route Harmony reasoning separately from final output (#21378)
Gasoonjia committed
July 25, 2026
A
Cover caller-stream TLS semantics (#21291) (#21291)
Anthony Shoumikhin committed
J
Refresh backend README (#21385) (#21385)
Julian Ng-Thow-Hing committed
July 24, 2026
E
Arm backend: Correct links in vgf docs (#21381)
Elena Zhelezina committed
J
migrate chunk-6 skip-list dirs (existing targets.bzl + extra TARGETS rules)
Jon Janzen committed
Y
portable: add fast path for contiguous innermost-dim reduction in amax.out and amin.out (#21142)
Youngsik Yang committed
J
merge fbcode-only TARGETS rules into existing targets.bzl files (#21145)
Jon Janzen committed
P
A
prevent AOTI library collisions between delegates (#21287)
Anthony Shoumikhin committed
Y
Arm backend: Avoid unnecessary lowering pass retraces (#21138)
Yufeng Shi committed
E
Arm backend: Make sure Softmax doesn't use EXT-INT16 (#21286)
Elena Zhelezina committed
E
Arm backend: Fix div.Scalar (#21288)
Elena Zhelezina committed
S
FF (#21296)
Scott Roy committed
July 23, 2026
S
support MXFP8 (#21146)
Scott Roy committed
J
Reorder channelwise gated delta rule chunked hot loops for autovectorization (#21021) (#21021)
Jacob Stevens committed
J
Add parallel chunked prefill for channelwise gated delta rule (#21061) (#21061)
Jacob Stevens committed
A
Arm backend: Improve permute/view fusing (#21055)
Adrian Lundell committed
July 22, 2026
P
Arm backend: Add PReLU decomposition pass
Per Held committed