Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1243 commits
Select commit Hold shift + click to select a range
34ac200
Arm backend: Add docstrings to compile specs (#15886)
Sebastian-Larsson Nov 19, 2025
12a7e13
Arm backend: Make PassPipeline test valid inputs (#15892)
oscarandersson8218 Nov 19, 2025
43fcc00
Update ReplaceSqueezeAndUnsqueezeWithViewPass to use new pass interface
DrJessop Nov 19, 2025
c0e664d
Security: Update glob to 10.5.0 to fix CVE-2025-64756 (#15904)
abhinaykukkadapu Nov 20, 2025
3591f4f
Fix circular dep in flat tensor import (#15902)
lucylq Nov 20, 2025
4ace07f
Arm backend: Support quantized cond and while (#15849)
Erik-Lundell Nov 20, 2025
b9d86bc
Arm backend: Rename passes for consistency (#15888)
martinlsm Nov 20, 2025
15d83b3
Arm backend: Add support for ubsan in executor_runner (#15894)
perheld Nov 20, 2025
546cda2
Cortex_M backend: Add conv op (#15896)
AdrianLundell Nov 20, 2025
3299a3d
Arm backend: Update pass name in arm_aot_compiler (#15916)
AdrianLundell Nov 20, 2025
854d6c5
Remove no-op clones in xnnpack (#15884)
GregoryComer Nov 20, 2025
ff9c9af
Fix UnboundLocalError: local variable 'qnn_compile_spec_buffer' refer…
cccclai Nov 20, 2025
6595657
Fix extract model script (#15924)
metascroy Nov 20, 2025
c2774f9
Arm backend: Fix get_module_name_filter (#15910)
oscarandersson8218 Nov 20, 2025
8006198
Arm backend: Add docstrings to vgf/backend.py and ethosu/backend.py (…
Sebastian-Larsson Nov 20, 2025
169bde8
Arm backend: Move support_extension to base class (#15909)
oscarandersson8218 Nov 20, 2025
a2e14d9
Arm backend: test without output re-order workaround (#15826)
mansnils Nov 20, 2025
80fea88
Fix Neutron backend API compatibility and multiprocessing fallback
apullin Nov 20, 2025
4c022df
Refactor download function for robustness and retries (#15933)
cccclai Nov 21, 2025
a07c0af
[aoti-backend-consolidation 2/3] backend.py (#15528)
Gasoonjia Nov 21, 2025
817f128
add cuda benchmark ci (#15883)
Gasoonjia Nov 21, 2025
54d7f37
Add bundled IO support to portable executor runner (#15368)
mansnils Nov 21, 2025
d3c3603
Rename the qnn demo backend (#15930)
cccclai Nov 21, 2025
76e0b29
Quantize lora linears (#15935)
lucylq Nov 21, 2025
2974798
Add mypy ignore for snakeviz (#15806)
perheld Nov 21, 2025
001dcc9
Tag scales for external data
lucylq Nov 21, 2025
41db544
Save external constant tensors to custom filename
lucylq Nov 21, 2025
3c4207f
make partial export failure not stop cuda benchmark (#15939)
Gasoonjia Nov 21, 2025
1730e63
Remove download progress reporting (#15946)
cccclai Nov 21, 2025
ef1b6ca
Fix build isolation and editable install issues (#15941)
larryliu0820 Nov 21, 2025
0963146
Create a button to manually trigger all cuda benchmarks (#15938)
Gasoonjia Nov 21, 2025
3e94868
only download benchmark data for further uploading (#15945)
Gasoonjia Nov 21, 2025
08569c0
Remove AWS Device Farm benchmark jobs (#15433)
GregoryComer Nov 21, 2025
4414580
Disable QNN build option for the default x86 build (#15949)
cccclai Nov 21, 2025
25dddba
Expose get_num_threads via pybind
kimishpatel Nov 22, 2025
e58587d
Move user-defined passes after SpecPropPass
lucylq Nov 22, 2025
bbe944c
Arm backend: Add missing default case in executor_runner (#15895)
perheld Nov 24, 2025
6a7cba4
Arm backend: Minimal example of pruning (#15851)
gggekov Nov 24, 2025
7ef000d
Arm backend: Update bias quantization range (#15918)
oscarandersson8218 Nov 24, 2025
3b59312
Added more users to exclude (#15781)
nil-is-all Nov 24, 2025
ed909ca
forward fix PR 15368 (#15964)
cccclai Nov 24, 2025
455841d
fix eval_llama_qnn custom annotation (#15953)
DannyYuyang-quic Nov 24, 2025
e976bdf
add 12.9 cuda support (#15818)
JacobSzwejbka Nov 24, 2025
7b983a0
Adding compile option warmup_execute_after_compile to optionally run …
trivedivivek Nov 25, 2025
1804952
Fix cpp/c compilation warnings.
hsharma35 Nov 25, 2025
9e7627a
Add ETDump event tracer support to LLaMa runner
navsud Nov 25, 2025
5ed8ad4
Qualcomm AI Engine Direct - GLM1.5B (#15691)
winskuo-quic Nov 25, 2025
72a27c5
Fix relu test + use xt macros.
hsharma35 Nov 25, 2025
ef3e198
Cortex-M backend: Fuse Relu, Hardtanh and Hardsigmoid (#15917)
AdrianLundell Nov 25, 2025
fe4b003
Switch to pnpm for dependency management (#16019)
mergennachin Dec 1, 2025
9bf1423
Update ReplaceAddMMWithLinearPass to use new pass interface
DrJessop Dec 1, 2025
d16e0bf
Update QCOM llama hardware support (#15965)
cccclai Dec 1, 2025
9d9d937
Add 16a4w_block QAT config
sxu Dec 1, 2025
cbf2419
Do not include cortex_m_lib for unit tests. Include only specific ops…
psiddh Dec 1, 2025
3a77741
custom comparator support in calculate_numeric_gap
Gasoonjia Dec 2, 2025
e1a2fb1
Fix index out of bounds error for 1d convs
DrJessop Dec 2, 2025
196d5e2
Use 4x3 tiled shader for linear mat mul which performs slightly better.
trivedivivek Dec 2, 2025
6211e5c
Add CentOS to the supported host OS (#16021)
cccclai Dec 2, 2025
78b8670
QNN Backend: Enable QualComm SM8850 targets (#15987)
yinrun Dec 2, 2025
8918f62
Qualcomm AI Engine Direct - Remove legacy code related to the shared …
shewu-quic Dec 2, 2025
c13d7af
Qualcomm AI Engine Direct - Improve CLI tools (#16012)
chenweng-quic Dec 2, 2025
e671dd9
More Passes updated to use new interface
DrJessop Dec 2, 2025
2a12b5e
Qualcomm AI Engine Direct - Support triu op and linear op with non-co…
shewu-quic Dec 2, 2025
1a58550
Fix -Wimplicit-const-int-float-conversion error.
hsharma35 Dec 2, 2025
4e78349
Add generic operator implementations.
hsharma35 Dec 2, 2025
36a0205
Rename generic operator cpp/headers.
hsharma35 Dec 2, 2025
1523836
use *args and **kwargs as signature of monkey patched function
Gasoonjia Dec 2, 2025
d27b772
Arm backend: Add test for BatchNorm1D QAT folding (#16001)
oscarandersson8218 Dec 2, 2025
1dbe38a
Arm backend: Add int16x8 LayerNorm test cases (#16015)
SaoirseARM Dec 2, 2025
416df9d
Arm backend: Update output reorder workaround (#15981)
mansnils Dec 2, 2025
8bba84b
Call prepare_traced_pt2 instead of prepare_pt2 in get_fake_quant_model
ethansfng Dec 2, 2025
900e101
Arm backend: Align docstrings in arm_quantizer.py with backend standa…
Sebastian-Larsson Dec 2, 2025
ffe7490
Some minor performance improvements to buffer 4b mat mul.
trivedivivek Dec 2, 2025
689301a
Adding a payload version to become compatible with extra output tenso…
jirioc Oct 24, 2025
db36037
Arm backend: Add missing bool case in prepare_input_tensors
perheld Nov 25, 2025
ac2305c
Arm backend: Pass location as string instead of TosaOpLocation (#15994)
tom-arm Dec 2, 2025
a4375a3
Cortex_M backend: Fuse clamp + hardswish decompostion (#16016)
AdrianLundell Dec 2, 2025
4e09fae
Arm backend: Add support for int32 clamp (#15977)
ArmRyan Dec 2, 2025
9ba6b00
docs: Add logging guidance for debug partition (#15979)
yujiaoliang Dec 2, 2025
975621f
make triton kernel usage user controlable
Gasoonjia Dec 2, 2025
c0a2d85
Remove sharded ckpt from export_llama (#15968)
lucylq Dec 2, 2025
8b3eb1c
Make ExecuTorch CoreML asset/trash/databse folders configurable.
metascroy Dec 3, 2025
1bdf121
Using buffer for weight tensors for quantized mat mul op.
trivedivivek Dec 3, 2025
0df9c69
Remove fairseq from export_llama
lucylq Dec 3, 2025
4008d66
More passes updated to be more efficient and correctly set their modi…
DrJessop Dec 3, 2025
a4d553a
Arm Backend: improve non-persistent placeholder and bool handling (#1…
robell Dec 3, 2025
c734c8b
Cortex_M backend: Remove int8 requirement for zero_points (#16061)
AdrianLundell Dec 3, 2025
34b17f6
Arm backend: Fixes and ignores for bandit
perheld Dec 1, 2025
7193ae9
NXP Backend: Add pass to remove unnecessary Quantize/Dequantize nodes…
MartinPavella Dec 3, 2025
ec786c9
Fix fake mode detection
angelayi Dec 3, 2025
12d2b2f
[OpenVINO] Fix Quantizer for PTQ (#15891)
anzr299 Dec 3, 2025
d4edd3f
Using generic implementation for 16-bit activations and 8 bit weights…
RahulC7 Dec 3, 2025
815113e
[etrecord] Implement generic fallback for `GraphModuleSerializer.han…
pytorchbot Dec 4, 2025
9e6876a
Add Run Android LlamaDemo with QNN backend (#16011)
luffy-yu Dec 4, 2025
95ee6fd
Using generic implemntation for 16-bit activations and 8 bit weights …
RahulC7 Dec 4, 2025
0353522
Arm backend: Add support for quant. decomposition (#15993)
oscarandersson8218 Dec 4, 2025
2b1a887
Arm backend: Add support for asan in executor_runner (#16009)
perheld Dec 4, 2025
58a082a
Arm backend: Use %zu instead of %lu (#15974)
gggekov Dec 4, 2025
fa02964
Arm backend: Make SupportedTOSAOperatorChecks work for INT+FP (#16072)
martinlsm Dec 4, 2025
61829be
Arm backend: Improve dtype validation (#15871)
oscarandersson8218 Dec 4, 2025
f5f17f3
Arm backend: Make sure DW-conv weights are reshaped once (#16071)
oscarandersson8218 Dec 4, 2025
410cc0f
Arm backend: Improve quantizer configuration in tests (#16073)
Erik-Lundell Dec 4, 2025
cc87c4d
Arm Backend: Add support for select_scatter.default (#15972)
agrima1304 Dec 4, 2025
e28f7b6
Arm backend: Move down a few passes to after fold Q/DQ pass (#16035)
martinlsm Dec 4, 2025
0c63ae7
Cortex_M backend: Remove scalar-scalar tests (#16078)
oscarandersson8218 Dec 4, 2025
432907a
Add view_copy/static_reshape support to XNNPACK delegate (#7959)
GregoryComer Dec 4, 2025
ac0717b
Arm backend: Move FuseConstantArgsPass (#16076)
oscarandersson8218 Dec 4, 2025
89353c0
[ET-VK][ez] Address regressed conv2d perf numbers on main (#16084)
pytorchbot Dec 4, 2025
303a9c8
grid_sampler_2d_out Portable Kernel Implementation (#16051)
jgibson2 Dec 4, 2025
c261334
Using generic implementation for 16-bit activations and 8-bit weights…
RahulC7 Dec 4, 2025
bd5df22
Fix OSS import for ref_implementations
DrJessop Dec 4, 2025
d58c682
[ET-VK] Enable and test texture IO for quantized convolution ops (#16…
SS-JIA Dec 5, 2025
59259e2
Support eliminate_quant_dequant_pairs flag.
eigen-k Dec 5, 2025
c46afd9
Update more replace ops passes to new interface
DrJessop Dec 5, 2025
f457bfd
Fix duplication bug in serialize_for_executorch
lucylq Dec 5, 2025
746fcfd
Arm backend: Replace ±inf and FP limit values with ±255.0 (#15976)
YufengShi-dudu Dec 5, 2025
2c69dac
Re-land cve security fix (#16103)
lucylq Dec 5, 2025
1a08915
Fix data pointer const-correctness in executorch kernels
maxjing Dec 5, 2025
7ff18ca
Using push constants for softmax op.
trivedivivek Dec 6, 2025
a918adf
Arm backend: Update epsilon to correct default of 2**-12 (#16096)
SaoirseARM Dec 6, 2025
d0bc874
Arm backend: Increase tolerance on Tanh int16 test (#16094)
SaoirseARM Dec 6, 2025
7d2f6c9
Arm backend: Rename build_reshape_tosa_1_0 to build_reshape_tosa (#16…
Sebastian-Larsson Dec 6, 2025
6ea9a4b
Arm backend: Handle all types in prepare_input_tensors() (#16092)
perheld Dec 6, 2025
74cfd16
Reduce allocation overhead in quantized sdpa (#16119)
pytorchbot Dec 7, 2025
ce2ee4b
[Executorch] Introduce caching cpu memory allocator (#16120)
pytorchbot Dec 7, 2025
d77e3eb
Using push constants for clone op parameters.
trivedivivek Dec 7, 2025
5ab7348
[Executorch] Use temp allocator for allocating scratch memory (#16121)
pytorchbot Dec 7, 2025
c9aeadb
[Executorch] Make module constructors uniform across (#16122)
pytorchbot Dec 7, 2025
9826a02
NXP backend: Improve `view_copy` delegation (#15270)
MartinPavella Dec 8, 2025
f92f898
Support negative dimensions on "aten.split_with_sizes_copy.default" (…
jgibson2 Dec 8, 2025
f5fae43
[Executorch] make slice_copy parallel (#16125)
pytorchbot Dec 8, 2025
d00a5c1
Ensure modified flag is returned correctly for RemovePermutesAroundEl…
eigen-k Dec 8, 2025
3df8b2b
Metal backend: SDPA metal implementation (#16086)
manuelcandales Dec 8, 2025
45fdf98
Add bitwise left/right (#15893)
mergennachin Dec 8, 2025
49e4675
bump arm cortex size test (#16134)
lucylq Dec 8, 2025
8d80647
reorganize VOLK_IMPLEMENTATION usage
SamGondelman Dec 8, 2025
35fb4e5
Using push constants for binary scalar op parameter.
trivedivivek Dec 8, 2025
f638b16
support skip atten in export
jackzhxng Dec 8, 2025
bb2d4f5
add ET_BUNDLE_IO_ENABLED in executor runner
zonglinpeng Dec 8, 2025
e4a32d7
Fix double-tracing in SpecPropPass (#15485)
GregoryComer Dec 8, 2025
79c89fb
Qualcomm AI Engine Direct - support SW6100 (#16003)
jethroqti Dec 9, 2025
8bd61ba
Remove op_linear_scratch_example
lucylq Dec 9, 2025
98364d8
Adding Test for CadenceWith16BitMatmulActivationsQuantizer
RahulC7 Dec 9, 2025
2572f5f
Enforce src and dst do not overlap in memcpy
lucylq Dec 9, 2025
7b3bdb3
Add RmsNormNopQuantizer and Pattern
ethansfng Dec 9, 2025
ba18a2a
Using push constants for add mm native op.
trivedivivek Dec 9, 2025
b91c4c7
Handle same tensor appearing multiple times in the cat input
ethansfng Dec 9, 2025
e07db02
Migrate more generic cadence operators to oss.
hsharma35 Dec 9, 2025
ad2b49e
Decompose after export in export_llama
lucylq Dec 9, 2025
f41f8ff
NXP backend: added aten.mul support (#15971)
novak-vaclav Dec 9, 2025
f295f43
Arm backend: Add U55 and U85 tests for deit_tiny (#16145)
oscarandersson8218 Dec 9, 2025
93cbe40
NXP backend: added aten.slice support (#15889)
novak-vaclav Dec 9, 2025
330b6c7
Fix ReplaceConvolutionOptionalArgsWithConcreteArgsPass (#16143)
DrJessop Dec 9, 2025
9ccdae2
[ET-VK][ez] Improve quantized convolution test case naming (#16141)
SS-JIA Dec 9, 2025
b6482ba
Annotate oncall for frequent signals
Ben0mega Dec 9, 2025
717c944
Prevent signed integer overflow in pixel_shuffle size calculation (#1…
GregoryComer Dec 9, 2025
60c3f1c
Rm platform args in executorch
jaejunku Dec 9, 2025
dd78bd4
Qualcomm AI Engine Direct - gpu support part1 (#12165)
haowhsu-quic Dec 9, 2025
9c56b91
Fix CoreML iOS26 numerics in static attention (#16144)
metascroy Dec 10, 2025
b6227b8
Fix rope ref implementation.
hsharma35 Dec 10, 2025
0213c62
Modify depthwise int8 conv2d to reduce register/memory pressure
GregoryComer Dec 10, 2025
ea2b7e0
Implement guardrail to check for existing branches (#16132)
shoumikhin Dec 10, 2025
9bcf4d0
Cortex_m backend: Support channels-broadcasting for ADD/MUL (#16131)
AdrianLundell Dec 10, 2025
1060fbe
Arm backend: Make INT+FP default for vgf-backend (#16176)
oscarandersson8218 Dec 10, 2025
1e66f5e
Arm backend: Add Quantization test pipeline (#16151)
oscarandersson8218 Dec 10, 2025
4a685b1
Qualcomm AI Engine Direct - Add MHA2SHA pass (#15438)
shewu-quic Dec 10, 2025
b35cde1
Qualcomm AI Engine Direct - enable operator max_pool3d by decompositi…
jethroqti Dec 10, 2025
051a4a2
Adding Test To Ensure All Future Quantizers Are Tested
RahulC7 Dec 10, 2025
9c83725
Fix off-by-on error
lucylq Dec 10, 2025
4eb75f9
Add lora test using qwen (#16161)
lucylq Dec 10, 2025
d3866b7
Add test for CadenceRmsNormNopQuantizer
ethansfng Dec 10, 2025
8559db2
Expose static llama in OSS
metascroy Dec 11, 2025
ad29f22
Migrate generic cadence operators to oss.
hsharma35 Dec 11, 2025
486d227
NXP backend: Add QAT support for NeutronQuantizer (#15692)
StrycekSimon Dec 11, 2025
6de746a
Arm backend: Add function to help linker force include backend (#16177)
zingo Dec 11, 2025
1e4c1c7
Update replace ops to correctly set modified bit
DrJessop Dec 11, 2025
4304c2d
Qwen lora quantize test (#16188)
lucylq Dec 11, 2025
1d19628
Scan support (#16028)
JacobSzwejbka Dec 11, 2025
199fc89
Add null pointer check for evalues (#14749)
rascani Dec 11, 2025
31de15a
Minor Refactor of ExecutorchRuntimeException (#16193)
psiddh Dec 11, 2025
4c444fd
Summary: Add context when LLMModule load fails to aid in debugging (#…
psiddh Dec 11, 2025
fac7814
[ez][ET-VK] Small fix for choose_qparams_affine_impl (#16186)
SS-JIA Dec 11, 2025
3f79e16
[ez][ET-VK] Update Vulkan runtime application name (#16187)
SS-JIA Dec 11, 2025
673fa67
Update torch pin to 11-20 (#16198)
mergennachin Dec 11, 2025
84303f1
Fix tests after torch pin PR (#16209)
mergennachin Dec 11, 2025
cea3ad3
Fix test-torchao-huggingface-checkpoints (#16211)
mergennachin Dec 11, 2025
47ec691
Add resume for download function (#15948)
cccclai Dec 11, 2025
ce252bb
Increase docker time limit (#16212)
mergennachin Dec 11, 2025
baea056
StaticAttetnion runtime support for generate_full_logits=False
sxu Dec 11, 2025
801f255
Remove internal dependencies for cadence ops.
hsharma35 Dec 12, 2025
0c927b7
optimum pin bump
JacobSzwejbka Dec 12, 2025
bd3283c
Revert "Add resume for download function" (#16215)
cccclai Dec 12, 2025
a3324a3
Update trunk.yml (#16216)
metascroy Dec 12, 2025
44cfa0c
Update examples/models/qwen2_5/README.md (#16150)
DamonFool Dec 12, 2025
ac3ece8
Update more ops to use the new pass interface, update some ref implem…
DrJessop Dec 12, 2025
e4d6060
Portable aoti (#15967)
JacobSzwejbka Dec 12, 2025
06908d9
NXP Backend: Update documentation to the new scheme (#15219)
robert-kalmar Dec 12, 2025
46b5c17
Arm backend: Fix broken index_tensor tests (#16220)
oscarandersson8218 Dec 12, 2025
a41fe64
Arm backend: Relax pre-push header checks (#16195)
Sebastian-Larsson Dec 12, 2025
67ab58b
Update ReplaceConvWithChannelLastConvPass and MakeSliceAndCatDimOuter…
DrJessop Dec 12, 2025
7deab35
Arm backend: Add conv3d support to Tosa/Vgf backends (#16093)
ArmRyan Dec 12, 2025
8050bbc
Arm backend: Refactor pass skipping mechanisms (#16153)
ArmRyan Dec 12, 2025
ac6e78e
Arm backend: Update vgf model tests (#16199)
oscarandersson8218 Dec 12, 2025
5a66efa
Add Python 3.13 support (#16004)
PetarTerziev-UL Dec 12, 2025
5612952
Replace duplicated RemoveCloneOpPass with RemoveCloneOpsTransformImpo…
eigen-k Dec 12, 2025
0461eb1
Temporarily disable view-copy partitioning in XNNPACK (#16226)
GregoryComer Dec 12, 2025
9b08ee7
Add ability to specify CoreML pipeline passes (#16118)
jgibson2 Dec 12, 2025
fc5a0f8
Upgrade Docker container from gcc9 to gcc11 and add gcc9-nopytorch va…
mergennachin Dec 15, 2025
3557654
Arm backend: Update vgf ops tests [Part 1] (#16245)
oscarandersson8218 Dec 15, 2025
8acfd67
NXP backend: Add dim order support to NeutronBackend. (#16146)
MartinPavella Dec 15, 2025
2c05892
Arm backend: Fix broken SD3Transformer2DModel test (#16247)
oscarandersson8218 Dec 15, 2025
7209f23
Fix typo in tutorial for exporting llm (#15906)
liu-mengyang Dec 15, 2025
fd392a5
Merging gh/abhinaykukkadapu/15/base to main (#16252)
manuelcandales Dec 15, 2025
bf821fc
Remove unused exception parameter from executorch/backends/vulkan/tes…
manuelcandales Dec 15, 2025
f086955
Fix fbcode/oss sync (#16254)
manuelcandales Dec 15, 2025
8360d5a
Qualcomm AI Engine Direct - Merge the two pybind libraries into a sin…
shewu-quic Dec 15, 2025
a0d7819
Arm backend: Update vgf ops tests [Part 2] (#16251)
oscarandersson8218 Dec 15, 2025
910378c
Append to cflag and honor CMAKE_POSITION_INDEPENDENT_CODE OFF (#16074)
zingo Dec 15, 2025
1462df3
Update supported Python versions in windows build (#16232)
kirklandsign Dec 15, 2025
ccabe80
Fix slice quant pattern match bug (#15624)
jackzhxng Dec 15, 2025
19c0e00
More robust PTAX finding (#16229)
JacobSzwejbka Dec 15, 2025
9b34302
Support cosine operator on XNNPACK (#15431)
GregoryComer Dec 16, 2025
38acb62
Generalize quantized input wrapper
ethansfng Dec 16, 2025
9583420
Update staging buffer allocation flags by transfer direction
GregoryComer Dec 16, 2025
a5145f9
Arm backend: Test partial quantization of models (#16249)
martinlsm Dec 16, 2025
92b5846
Arm backend: Match arg ranks for min/max ops (#16181)
YufengShi-dudu Dec 16, 2025
0242596
Arm backend: Add support for masked_fill_.Scalar (#16272)
oscarandersson8218 Dec 16, 2025
8d3d8c3
Arm backend: Fix two control flow bugs (#16274)
Erik-Lundell Dec 16, 2025
f0a50a8
Rename ethous-u-scratch to arm-scratch (#16224)
mansnils Dec 16, 2025
a11f08d
Cover tokenizer build when running install_executorch.sh --clean (#16…
cccclai Dec 16, 2025
ce236df
[ET-VK] Pass detailed op information to event tracer (#16266)
SS-JIA Dec 16, 2025
16ef01d
[ET-VK][test-utils] Implement submodule extraction utilities (#16267)
SS-JIA Dec 16, 2025
74dd5ba
Add Python 3.13 to windows wheel build CI (#16284)
larryliu0820 Dec 17, 2025
11627cf
Qualcomm AI Engine Direct - QNN ExecuTorch Intermediate Output Debugg…
winskuo-quic Dec 17, 2025
7bec5c7
Qualcomm AI Engine Direct - Refactor QNN Manager (#15903)
shewu-quic Dec 17, 2025
73c243e
Arm backend: Add processed module to global scope in aot_arm_compiler…
Erik-Lundell Dec 17, 2025
b264880
Arm backend: Add T5 model tests (#16278)
YufengShi-dudu Dec 17, 2025
784a08c
Lint fix: qnnpreprocess (#16295)
manuelcandales Dec 17, 2025
d4c96a5
Fix openvino torchao build (#16296)
manuelcandales Dec 17, 2025
0dc7a5e
Arm backend: Add support for clamp.Tensor (#16273)
YufengShi-dudu Dec 17, 2025
523542a
Arm backend: Improve while support (#16287)
Erik-Lundell Dec 17, 2025
88e72f8
Arm backend: Add TOSA-dtype validation to tester (#16293)
oscarandersson8218 Dec 17, 2025
39e8852
Arm backend: Disable relu-fusing for sym. quant (#16286)
oscarandersson8218 Dec 17, 2025
f51fab8
Add broadcast semantics to quantized ops.
hsharma35 Dec 17, 2025
969a686
Fix a typo on build-wheels-windows.yml (#16300)
larryliu0820 Dec 17, 2025
08af1e3
Clean up various leftover platform arg usage around the repo that nev…
8Keep Dec 17, 2025
3faab31
Add 'docker-builds' to viable/strict requirement (#16282)
kirklandsign Dec 17, 2025
b60af73
forward fix
manuelcandales Dec 18, 2025
4c1768d
triton sdpa kernel with better perf (#16167)
Gasoonjia Dec 18, 2025
e5deb89
NXP Backend: Adding the loader for eIQ Neutron SDK converter module.
jirioc Dec 18, 2025
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
12 changes: 9 additions & 3 deletions .ci/docker/build.sh
Original file line number Diff line number Diff line change
Expand Up @@ -23,9 +23,14 @@ MINICONDA_VERSION=23.10.0-1
BUCK2_VERSION=$(cat ci_commit_pins/buck2.txt)

case "${IMAGE_NAME}" in
executorch-ubuntu-22.04-gcc9)
executorch-ubuntu-22.04-gcc11)
LINTRUNNER=""
GCC_VERSION=11
;;
executorch-ubuntu-22.04-gcc9-nopytorch)
LINTRUNNER=""
GCC_VERSION=9
SKIP_PYTORCH=yes
;;
executorch-ubuntu-22.04-clang12)
LINTRUNNER=""
Expand Down Expand Up @@ -54,13 +59,13 @@ case "${IMAGE_NAME}" in
executorch-ubuntu-22.04-mediatek-sdk)
MEDIATEK_SDK=yes
CLANG_VERSION=12
ANDROID_NDK_VERSION=r27b
ANDROID_NDK_VERSION=r28c
;;
executorch-ubuntu-22.04-clang12-android)
LINTRUNNER=""
CLANG_VERSION=12
# From https://developer.android.com/ndk/downloads
ANDROID_NDK_VERSION=r27b
ANDROID_NDK_VERSION=r28c
;;
*)
echo "Invalid image name ${IMAGE_NAME}"
Expand Down Expand Up @@ -95,6 +100,7 @@ docker build \
--build-arg "QNN_SDK=${QNN_SDK:-}" \
--build-arg "MEDIATEK_SDK=${MEDIATEK_SDK:-}" \
--build-arg "ANDROID_NDK_VERSION=${ANDROID_NDK_VERSION:-}" \
--build-arg "SKIP_PYTORCH=${SKIP_PYTORCH:-}" \
-f "${OS}"/Dockerfile \
"$@" \
.
2 changes: 1 addition & 1 deletion .ci/docker/ci_commit_pins/optimum-executorch.txt
Original file line number Diff line number Diff line change
@@ -1 +1 @@
40b02a2dc61bbf901a2df91719f47c98d65368ec
0123293118efb08ac4ffc4fefe9d330201465c93
2 changes: 1 addition & 1 deletion .ci/docker/ci_commit_pins/pytorch.txt
Original file line number Diff line number Diff line change
@@ -1 +1 @@
4d4abec80f03cd8fdefe1d9cb3a60d3690cd777e
7a064ed3eafa43f17412d434b395240c727b3000
18 changes: 18 additions & 0 deletions .ci/docker/common/install_arm.sh
Original file line number Diff line number Diff line change
@@ -0,0 +1,18 @@
#!/bin/bash
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
# Copyright 2025 Arm Limited and/or its affiliates.
#
# This source code is licensed under the BSD-style license found in the
# LICENSE file in the root directory of this source tree.

set -ex

install_arm_prerequiresites() {
apt-get update -y
apt-get install -y --no-install-recommends \
mesa-vulkan-drivers libvulkan1
rm -rf /var/lib/apt/lists/*
}

install_arm_prerequiresites
4 changes: 2 additions & 2 deletions .ci/docker/common/install_pytorch.sh
Original file line number Diff line number Diff line change
Expand Up @@ -12,8 +12,8 @@ source "$(dirname "${BASH_SOURCE[0]}")/utils.sh"

install_domains() {
echo "Install torchvision and torchaudio"
pip_install --no-use-pep517 --user "git+https://github.com/pytorch/audio.git@${TORCHAUDIO_VERSION}"
pip_install --no-use-pep517 --user "git+https://github.com/pytorch/vision.git@${TORCHVISION_VERSION}"
pip_install --no-build-isolation --user "git+https://github.com/pytorch/audio.git@${TORCHAUDIO_VERSION}"
pip_install --no-build-isolation --user "git+https://github.com/pytorch/vision.git@${TORCHVISION_VERSION}"
}

install_pytorch_and_domains() {
Expand Down
28 changes: 15 additions & 13 deletions .ci/docker/requirements-ci.txt
Original file line number Diff line number Diff line change
@@ -1,12 +1,12 @@
mpmath==1.3.0
numpy>=2.0.0; python_version >= '3.10'
PyYAML==6.0.1
ruamel.yaml==0.17.32
sympy==1.12
ruamel.yaml==0.18.15
sympy>=1.13.3
timm==0.6.13
tomli==2.0.1
torchsr==1.0.4
transformers==4.47.1
transformers==4.56.1
zstd==1.5.5.1
pandas>=2.2.2; python_version >= '3.10'
pytest==7.2.0
Expand All @@ -16,18 +16,20 @@ hypothesis==6.84.2
parameterized==0.9.0

# Doc build requirements, same as https://github.com/pytorch/pytorch/blob/main/.ci/docker/requirements-docs.txt
sphinx==5.3.0
sphinx==7.2.6
sphinxcontrib.katex==0.9.10
breathe==4.36.0 # only if generating C++
exhale==0.3.7 # only if generating C++ docs
docutils==0.18.1,<0.21
sphinx-design==0.6.1
sphinxcontrib-mermaid==1.0.0
myst-parser==3.0.1 # if want to contribute in markdown
sphinx-gallery==0.14.0 # only if hosting interactive tutorials
sphinx-sitemap==2.7.1
sphinx-reredirects==0.1.4
sphinx-gallery==0.14.0
breathe==4.34.0
exhale==0.2.3
docutils==0.16
matplotlib>=3.9.4
sphinx-copybutton==0.5.2
# PyTorch Theme
-e git+https://github.com/pytorch/pytorch_sphinx_theme.git#egg=pytorch_sphinx_theme
myst-parser==0.18.1
sphinx_design==0.4.1
sphinx-copybutton==0.5.0

pytorch_sphinx_theme2==0.2.0
# script unit test requirements
yaspin==3.1.0
6 changes: 5 additions & 1 deletion .ci/docker/ubuntu/Dockerfile
Original file line number Diff line number Diff line change
Expand Up @@ -64,9 +64,10 @@ ENV SCCACHE_S3_KEY_PREFIX executorch
ENV SCCACHE_REGION us-east-1

ARG TORCH_VERSION
ARG SKIP_PYTORCH
COPY ./common/install_pytorch.sh install_pytorch.sh
COPY ./common/utils.sh utils.sh
RUN bash ./install_pytorch.sh && rm install_pytorch.sh utils.sh
RUN if [ -z "${SKIP_PYTORCH}" ]; then bash ./install_pytorch.sh; fi && rm install_pytorch.sh utils.sh

ARG LINTRUNNER
# Install lintrunner if needed
Expand All @@ -83,6 +84,9 @@ RUN if [ -n "${ANDROID_NDK_VERSION}" ]; then bash ./install_android.sh; fi
RUN rm install_android.sh

ARG ARM_SDK
COPY ./common/install_arm.sh install_arm.sh
RUN if [ -n "${ARM_SDK}" ]; then bash ./install_arm.sh; fi
RUN rm install_arm.sh

ARG ZEPHYR_SDK
COPY ./common/install_zephyr.sh install_zephyr.sh
Expand Down
6 changes: 3 additions & 3 deletions .ci/scripts/build-qnn-sdk.sh
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@ build_qnn_backend() {
export EXECUTORCH_ROOT="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")/../.." && pwd)"

parallelism=$(( $(nproc) - 1 ))
bash backends/qualcomm/scripts/build.sh --skip_aarch64 --job_number ${parallelism} --release
bash backends/qualcomm/scripts/build.sh --skip_linux_android --skip_linux_embedded --job_number ${parallelism} --release
}

set_up_aot() {
Expand All @@ -38,14 +38,14 @@ set_up_aot() {
-DEXECUTORCH_BUILD_EXTENSION_EXTENSION_LLM=ON \
-DEXECUTORCH_BUILD_EXTENSION_EXTENSION_LLM_RUNNER=ON \
-DEXECUTORCH_BUILD_EXTENSION_FLAT_TENSOR=ON \
-DEXECUTORCH_BUILD_EXTENSION_NAMED_DATA_MAP=ON \
-DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \
-DEXECUTORCH_ENABLE_EVENT_TRACER=ON \
-DPYTHON_EXECUTABLE=python3
cmake --build $PWD --target "PyQnnManagerAdaptor" "PyQnnWrapperAdaptor" -j$(nproc)
cmake --build $PWD --target "PyQnnManagerAdaptor" -j$(nproc)
# install Python APIs to correct import path
# The filename might vary depending on your Python and host version.
cp -f backends/qualcomm/PyQnnManagerAdaptor.cpython-310-x86_64-linux-gnu.so $EXECUTORCH_ROOT/backends/qualcomm/python
cp -f backends/qualcomm/PyQnnWrapperAdaptor.cpython-310-x86_64-linux-gnu.so $EXECUTORCH_ROOT/backends/qualcomm/python
popd

# Workaround for fbs files in exir/_serialize
Expand Down
Loading