You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Until now, an EP could report only one weightless mode, and apps could only turn weightless mode on or off. Each mode has different runtime requirements: ALL needs the source model at runtime, EXTERNAL_ONLY doesn't. With this change, EPs report every mode they support and apps choose the one that fits, without breaking EPs and apps built for 1.29 and 1.30.
Motivation and Context
Address the following shortcomings from PR 29607:
After a system update, a plugin EP could switch to a weightless mode whose requirements the app wasn't prepared to meet.
An EP that supports both modes could only advertise one. If it advertised the more demanding one, apps that can't meet those requirements would skip the EP.
Changes
Modes: OrtWeightlessSupport stays a sequential enum (ABI-compatible).
New value OrtWeightlessSupport_ALL_OR_EXTERNAL_ONLY = 3 describes an EP capability only; apps can't request it.
Future versions may add values, and apps should ignore values they don't recognize.
EP reporting:
OrtEp::GetWeightlessSupport() keeps its 1.29 signature. EPs return a single mode or OrtWeightlessSupport_ALL_OR_EXTERNAL_ONLY. ORT versions before 1.32 only check that the value isn't NONE, so returning 3 is safe for them too.
New EP metadata key weightless_supported_modes, with values "none", "external_only", "all" or "all_or_external_only".
The existing weightless_support key keeps its 1.29 values. ORT and apps fall back to it when the new key is absent.
App request:
New ModelCompilationOptions_SetWeightlessMode(OrtWeightlessSupport) (C++: SetWeightlessMode()) and session option ep.enable_weightless_mode. The option accepts exactly "0", "1" or "2".
Deprecated: ModelCompilationOptions_SetWeightlessEnabled(bool) and ep.enable_weightless. They keep their previous behavior when the new option isn't set.
Validation:
InferenceSession::Initialize() rejects invalid values (INVALID_ARGUMENT) for every API (C, C++, Python), even without plugin EPs or compiled nodes. That includes "3", other spellings like "0x2", and an explicitly empty value.
Before creating each plugin EP, ORT checks the requested mode against the EP metadata (EP_FAIL if it isn't supported). The EP then receives a copy of the session options with ep.enable_weightless set to match ("0" or "1"), so EPs built for earlier versions behave correctly. The app's session options aren't modified.
During Compile(), ORT checks the mode against GetWeightlessSupport(). It logs a warning if that disagrees with the EP metadata, and includes the mismatch in the error when the mode is rejected.
Error codes from plugin EP creation are now preserved instead of becoming ORT_FAIL.
Recorded mode: the chosen mode is written to the compiled model's metadata (weightless_mode). ORT warns when a model compiled with ALL is loaded without a source model path or buffer.
Fix: the weightless source model buffer moved into onnxruntime::SessionOptions. Previously it was lost when session options were copied or cloned, and the session couldn't see it.
Example EP: reports both metadata keys and honors the selected mode, both when compiling and when running directly:
NONE: copies all initializers.
EXTERNAL_ONLY: copies only internal initializers.
ALL: copies none.
Docs: updated header docs and a new design doc, docs/design/Weightless_Mode.md. It covers how EPs report modes, how apps choose one, compatibility with older EPs and apps, and the runtime and packaging requirements for each mode.
Compatibility
GetWeightlessSupport() and its OrtWeightlessSupport* output are unchanged.
weightless_support keeps its values ("none", "external_only", "all").
SetWeightlessEnabled(bool) and ep.enable_weightless still work. When ep.enable_weightless_mode is set, it takes precedence, and ORT translates it for EPs that only read the deprecated option.
Testing
onnxruntime_autoep_test:
Compiling with each mode, including the mode recorded in the compiled model.
How the EP handles initializers for internal, external and mixed models, both compiled and run directly.
Invalid values and their error codes.
Rejection when the EP metadata doesn't support the mode.
ep.enable_weightless being set correctly for older EPs.
The deprecated APIs.
onnxruntime_test_all: PluginExecutionProviderTest.Compile_WeightlessSupportDoesNotMatchEpMetadata checks the mismatch warning.
Python (onnxruntime_test_python.py): invalid values are rejected and the ALL source-model warning is logged on the Python path.
Javier Martinez (javier-intel)
changed the title
Improve the app-EP weightless contract to allow apps to select the be…
Improve the app-EP weightless contract
Sep 29, 2026
Thanks for writing this up. I agree with separating the modes an EP supports from the mode the application selects. For compatibility, I suggest the following approach.
This means the EP supports both selectable modes and honors the application's explicit selection. It is a capability value, not a compilation request. Applications still select exactly NONE (0), EXTERNAL_ONLY (1), or ALL (2).
Keeping the signature avoids the function-pointer type change. Adding a return value still needs compatibility consideration, although the released v1.30.0 ORT implementation checks only for NONE, so it accepts 3 for a legacy boolean request. When no explicit mode is supplied, upgraded EPs must preserve their established legacy behavior.
2. Normalize legacy options and validate the requested mode
When the new mode option is present, ORT should consistently override the legacy enable flag before CreateEp(). This should cover both the compile API and direct session-option configuration.
For an older EP whose reported scope describes its behavior when weightless is enabled:
New request
Older EP's reported scope
Action
NONE
Any
Force ep.enable_weightless = "0"
EXTERNAL_ONLY
EXTERNAL_ONLY
Set legacy option to "1"
ALL
ALL
Set legacy option to "1"
Either enabled mode
Different or unknown scope
Reject the request clearly
For example, an older EP reporting ALL cannot be instructed to choose EXTERNAL_ONLY through its boolean interface alone, unless another documented selector exists. Without a valid mapping, ORT should reject the request rather than claim it was honored.
For an upgraded EP reporting ALL_OR_EXTERNAL_ONLY, either individual mode is accepted, and the EP must honor ep.enable_weightless_mode.
When the new option is absent, preserve the existing legacy setting and behavior. Invalid requests, including 3, should return an invalid-argument error.
3. Preserve the existing metadata key and add a supported-modes key
This adds capability information for newer applications without changing what older applications read. New applications can prefer the new key and fall back to the old key when it is absent. ORT should compare the full callback capability against the new key when present.
When the new mode is explicitly NONE and the deprecated key is absent, GetConfigOrDefault(..., "0") makes this branch skip the copy, so the EP observes the old option as absent rather than the promised explicit "0". This contradicts the compatibility contract and the hook can distinguish the two states. Compare the optional entry itself so an absent key is also populated; add the missing { "0", nullptr, 0 } test case.
This file still declares and accepts std::string_view at lines 36–40, so removing its direct header makes compilation depend on an unrelated transitive include. Retain <string_view> while adding <string>.
Add regression test for copied source-model buffer
The regression this move fixes—retaining the source-model buffer when session options are copied or cloned—is not covered. The existing test only validates setter inputs and never copies the options or reads the buffer afterward. Add a clone/copy regression test that verifies both pointer and size survive, preferably through the EP getter path that previously lost them.
Invalid mode validation is compiled out for minimal builds, and the minimal-build option loop below does not reject this key. A minimal ORT-format session can therefore initialize successfully with ep.enable_weightless_mode="3" or an empty value, contrary to the public option contract that only exact values 0–2 are accepted. Keep ValidateWeightlessMode and its call outside the non-minimal guard; only the source-model warning needs to remain guarded if necessary.
This new API documentation omits the compatibility fallback that the implementation and other public docs require. An application following only this method's contract would miss 1.29–1.31 EPs that expose only weightless_support. State that kOrtEpDevice_EpMetadataKey_WeightlessSupport must be inspected when weightless_supported_modes is absent.
The shared parser also accepts "all_or_external_only" for the legacy weightless_support key, although that key is contractually limited to the 1.29 values. An EP that puts the aggregate value only in the legacy key will therefore pass pre-creation validation and avoid the intended metadata-mismatch warning, while older applications cannot understand its metadata. Treat the aggregate as unrecognized unless it came from weightless_supported_modes.
I've rebased the branch to allow FF and added a couple of changes, one to fix the CI build by making CudaPluginArenaTest.CudaGraphRejectsEnvironmentAllocator expect ORT_INVALID_ARGUMENT, the other to reduce the Android minimal build size by making a cheaper error message and guard the new virtual.
Most of this PR isn't in that binary due to being inside !ORT_MINIMAL_BUILD, the dependency tracking, the warnings, the new metadata and the checks done before CreateEp() add nothing. The review show five small pieces compiled into the minimal build. The pieces and their estimates for arm64 with LTO are:
What
File
Est. bytes
Why it's in the minimal build
epctx::ParseWeightlessMode(), mostly its error message
ep_context_options.cc
~600–800
The message was built from seven arguments (const char*, std::string, five const char*), which likely adds a new MakeString/std::ostringstream instantiation, plus about 180 bytes of message text.
ValidateWeightlessMode() and its call in InferenceSession::Initialize()
inference_session.cc
~100–150
Kept outside the minimal-build guard on purpose, so minimal builds also reject invalid ep.enable_weightless_mode values.
New virtual IExecutionProvider::GetEpContextExternalDependencies()
execution_provider.h
~80–120
It wasn't guarded, so it added a vtable slot (8 bytes plus a 24-byte relocation) in IExecutionProvider and CPUExecutionProvider, plus the default implementation and its unwind entry.
OrtCompileAPI::ModelCompilationOptions_SetWeightlessMode stub and its table slot
compile_api.cc
~60–80
Every OrtCompileApi function needs a slot in all builds to keep the struct layout. Minimal builds get a "not supported" stub, and the slot needs a relocation.
Source-model buffer moved into onnxruntime::SessionOptions
session_options.h
~0–50
The implicit SessionOptions copy is slightly larger, mostly offset by removing the custom OrtSessionOptions copy constructor.
Updated
[Microsoft.AspNetCore.Mvc.Testing](https://github.com/dotnet/dotnet)
from 10.0.14 to 10.0.15.
<details>
<summary>Release notes</summary>
_Sourced from [Microsoft.AspNetCore.Mvc.Testing's
releases](https://github.com/dotnet/dotnet/releases)._
No release notes found for this version range.
Commits viewable in [compare
view](https://github.com/dotnet/dotnet/commits).
</details>
Updated [Microsoft.Data.Sqlite](https://github.com/dotnet/dotnet) from
10.0.14 to 10.0.15.
<details>
<summary>Release notes</summary>
_Sourced from [Microsoft.Data.Sqlite's
releases](https://github.com/dotnet/dotnet/releases)._
No release notes found for this version range.
Commits viewable in [compare
view](https://github.com/dotnet/dotnet/commits).
</details>
Updated
[Microsoft.ML.OnnxRuntime](https://github.com/Microsoft/onnxruntime)
from 1.31.1 to 1.32.0.
<details>
<summary>Release notes</summary>
_Sourced from [Microsoft.ML.OnnxRuntime's
releases](https://github.com/Microsoft/onnxruntime/releases)._
## 1.32.0
ONNX Runtime 1.32.0 reintegrates ONNX 1.23.2 with opset 28 support,
introduces CPU/GPU hybrid MoE offloading and shared-layout CPU QMoE
kernels, expands device-memory, weightless-compilation, and Python
model-package APIs, dramatically accelerates session initialization on
large graphs, and strengthens runtime and optimizer reliability. These
notes cover changes since ONNX Runtime 1.31.1. CUDA and WebGPU kernel
updates are summarized here; detailed provider notes are covered by
their separate plugin EP releases.
## Highlights
- Reintegrated **ONNX 1.23.2** with **opset 28** operator support
(`SpaceToDepth`, `Mod`, signed `BitShift`, and `BFloat16` extensions)
and updated **DLPack to v1.3** across all implemented ORT numeric and
boolean tensor types
([#33134](microsoft/onnxruntime#33134),
[#32362](microsoft/onnxruntime#32362),
[#33848](microsoft/onnxruntime#33848)).
- Added a session-scoped **CPU/GPU hybrid MoE offload** planner and
execution backend for quantized `QMoE` (NVFP4, INT4, INT3, and INT2) and
block-packed `MoEPacked` models, backed by shared-layout AVX-512/AVX2
CPU `QMoE` kernels and memory-mapped expert storage
([#33450](microsoft/onnxruntime#33450),
[#33451](microsoft/onnxruntime#33451),
[#33452](microsoft/onnxruntime#33452),
[#33973](microsoft/onnxruntime#33973)).
- Expanded core APIs with `OrtApi::EpDevice_GetMemoryCapacity`, explicit
multi-mode weightless compilation
(`ModelCompilationOptions_SetWeightlessMode` and
`ep.enable_weightless_mode`), Python bindings for the Model Package
APIs, and zero-copy CPU buffer borrowing in `AddExternalInitializers`
([#33447](microsoft/onnxruntime#33447),
[#32899](microsoft/onnxruntime#32899),
[#33640](microsoft/onnxruntime#33640),
[#33877](microsoft/onnxruntime#33877)).
- Eliminated superlinear bottlenecks in allocation planning, constant
sharing, graph I/O mapping, and topological sorting—reducing session
initialization on 100k-node constant-heavy graphs from minutes to
seconds—and added pipelined pinned-memory staging for external GPU
weight loading
([#33407](microsoft/onnxruntime#33407),
[#33939](microsoft/onnxruntime#33939),
[#34009](microsoft/onnxruntime#34009),
[#33839](microsoft/onnxruntime#33839)).
- Expanded shared generative AI and quantized contrib operators,
including Kimi Delta Attention (`GatedDeltaNet` per-key-dimension
decay), right-padded chunked prefill in `GroupQueryAttention`, native
GGUF/block-packed operators (`GatherPacked`, `MatMulPacked`,
`MoEPacked`), and sparse-attention index sharing
([#33817](microsoft/onnxruntime#33817),
[#33503](microsoft/onnxruntime#33503),
[#33953](microsoft/onnxruntime#33953),
[#33224](microsoft/onnxruntime#33224)).
## Announcements & Compatibility
- **CUDA Plugin EP v0.4.0** and **WebGPU Plugin EP v0.6.0** are matched
with ONNX Runtime 1.32.0
([#33214](microsoft/onnxruntime#33214)). For
**contrib-op compatibility**, core and the plugin must use the same
contributed-operator schemas; building both from the **same ONNX Runtime
revision** is recommended. Schema mismatches can cause incorrect
execution or crashes and may not be detected during plugin registration.
- **ACL EP removed:** Following its deprecation in 1.31.0, the Arm
Compute Library (`ACLExecutionProvider`) execution provider, its build
options, and its standalone C registration API have been removed.
Python's `get_all_providers()` no longer lists `ACLExecutionProvider`;
applications should remove it from explicit provider lists
([#33851](microsoft/onnxruntime#33851)).
- **Legacy provider-bridge `NvTensorRTRTXExecutionProvider` removed:**
In-tree provider-bridge `NvTensorRTRTXExecutionProvider` sources have
been removed in favor of the standalone NVIDIA TensorRT RTX plugin EP
registered via `OrtApi::RegisterExecutionProviderLibrary`.
`NvTensorRtRtxProviderOptions` is retained as a deprecated declaration
for source compatibility, and legacy `AppendExecutionProvider` calls now
report a deprecation error directing callers to the plugin EP
([#32276](microsoft/onnxruntime#32276)).
- **ONNX 1.23.2 integration:** ONNX has been upgraded to **1.23.2**,
raising supported opset ceilings to **28** and restoring
`ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT6E2M3` and
`ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT6E3M2` public element-type mappings
and `Cast` conversions
([#33134](microsoft/onnxruntime#33134)).
- **DLPack v1.3 boolean contract change:** DLPack has been upgraded to
**v1.3** with native `kDLBool` metadata and bidirectional support for
all implemented ORT numeric and boolean types, including all five
`float8` formats, packed `float4`, `int4`/`uint4`, and `int2`/`uint2`
(`float6` remains unsupported in DLPack). The low-level `is_bool_tensor`
override was removed, and `uint8` DLPack metadata now always imports as
`uint8`
([#33848](microsoft/onnxruntime#33848)).
- **Weightless compilation API migration:**
`OrtCompileApi::ModelCompilationOptions_SetWeightlessEnabled` and the
`ep.enable_weightless` / `ep.enable_weightless_ep_context_nodes` session
options are deprecated in favor of
`ModelCompilationOptions_SetWeightlessMode` and
`ep.enable_weightless_mode` (`"0"` = `NONE`, `"1"` = `EXTERNAL_ONLY`,
`"2"` = `ALL`). EPs can now advertise multi-mode capability via
`OrtWeightlessSupport_ALL_OR_EXTERNAL_ONLY` and
`weightless_supported_modes`
([#32899](microsoft/onnxruntime#32899)).
- **Training CI retirement:** Training build and test lanes were retired
from active GitHub Actions CI workflows while retaining inference build
coverage
([#33836](microsoft/onnxruntime#33836)).
- This is the last release for [CocoaPods](https://cocoapods.org/), and
we will switch to [Swift Package
Manager](https://docs.swift.org/latest/documentation/packagemanagerdocs/).
- We will use combo onnxruntime-gpu python package to reduce storage
size from next release. You can test it
[here](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ORT-Nightly/PyPI/onnxruntime-gpu/overview)
## New Features
### Core APIs & Runtime
- Added `OrtApi::EpDevice_GetMemoryCapacity` (C++
`EpDevice::GetMemoryCapacity()`) and
`OrtEpFactory::GetDeviceMemoryCapacity` to query device-global total
memory and estimated currently available memory (`DEFAULT` or
`HOST_ACCESSIBLE`), with built-in and plugin CUDA EP support
([#33447](microsoft/onnxruntime#33447)).
- Expanded the app–EP weightless contract so EPs advertise all supported
weightless modes (`weightless_supported_modes`), applications request an
explicit mode (`ModelCompilationOptions_SetWeightlessMode` /
`ep.enable_weightless_mode`), and compiled EPContext models record
`weightless_mode` and `external_dependencies` (`source_model`,
`external_weights`, `ep_context_binary`) in model metadata
([#32899](microsoft/onnxruntime#32899)).
- Extended `session.use_external_initializer_file_buffers_directly="1"`
to `AddExternalInitializers` for zero-copy borrowing of naturally
aligned CPU buffers larger than 127 bytes on little-endian hosts, and
documented the read-only guarantee for external initializers added via
`OrtModelEditorApi::AddInitializerToGraph` with `data_is_external=true`
([#33877](microsoft/onnxruntime#33877),
[#33717](microsoft/onnxruntime#33717)).
- Added lock-free multi-worker pinned-memory staging
(`session.external_data_loader.*`) to accelerate external initializer
loading into GPU memory across built-in and plugin CUDA EPs
([#33839](microsoft/onnxruntime#33839)).
- Added a session-scoped MoE/QMoE CPU/GPU offload planner and
memory-budget ledger (`session.moe.offload.*`), generic
`session.kernel_dispatch_policy` (`auto`, `latency`, `memory`, `safe`)
configuration plumbing, Level-1 workspace/prepack memory accounting
fallback for nodes absent from memory-profile CSVs, and internal
capability-planning protocol types
([#33451](microsoft/onnxruntime#33451),
[#33042](microsoft/onnxruntime#33042),
[#33311](microsoft/onnxruntime#33311),
[#33841](microsoft/onnxruntime#33841)).
- Enabled cross-thread reuse of idle device streams when graph capture
is disabled and providers register device-selection callbacks, and
enabled compile-session profiling (`<profile_file>_compile`) in
`onnxruntime_perf_test` when `-p` is used with `--compile_ep_context`
([#33053](microsoft/onnxruntime#33053),
[#33694](microsoft/onnxruntime#33694)).
- Made `onnxruntime_DISABLE_EXCEPTIONS` (`--disable_exceptions`) usable
without `onnxruntime_MINIMAL_BUILD` and added the
`onnxruntime_DISABLE_STREAM=ON` CMake option to disable device-stream
support in non-minimal builds
([#33120](microsoft/onnxruntime#33120),
[#33482](microsoft/onnxruntime#33482)).
### ONNX Opset 28 & Shared Contrib Operators
These updates affect core ONNX and contributed operator schemas and
cross-provider behavior, not only the separately released GPU plugins.
- Added ONNX opset 28 operator support: native `SpaceToDepth` (`DCR` and
`CRD`) on CPU and CUDA with version-split registrations across DML, JS,
WebGPU, and QNN; opset-28 floating floor-modulus `Mod` on CPU, CUDA, and
DML; opset-28 `BFloat16` support for `Compress`, `OneHot`,
`ReverseSequence`, `Unique`, `Celu`, and `Einsum` on CPU (plus
`Compress`, `OneHot`, and `ReverseSequence` on CUDA); and `BFloat16`
`Pad` (opsets 21–25) on CPU and CUDA
([#32362](microsoft/onnxruntime#32362),
[#32369](microsoft/onnxruntime#32369),
[#32378](microsoft/onnxruntime#32378),
[#34062](microsoft/onnxruntime#34062)).
- Extended `GatedDeltaNet` with Kimi Delta Attention (KDA)
per-key-dimension decay across CPU, CUDA, and WebGPU, and extended
`GatedDeltaNet` and `CausalConvWithState` state/gate handling for
multi-token speculative decoding
([#33817](microsoft/onnxruntime#33817),
[#33471](microsoft/onnxruntime#33471)).
- Added an optional `query_seqlens` input to `GroupQueryAttention` to
skip right-padded prefill tokens and idle batch entries
(`query_seqlens[b] = 0`) on CPU and CUDA, and added `q_norm_epsilon` /
`k_norm_epsilon` attributes to preserve distinct Q and K normalization
epsilons during GQA pre-norm fusion
([#33503](microsoft/onnxruntime#33503),
[#33505](microsoft/onnxruntime#33505),
[#33832](microsoft/onnxruntime#33832),
[#33363](microsoft/onnxruntime#33363)).
- Defined the INT3 `QMoE` v1 packed-code contract, added opt-in
post-SiLU clipping (`swiglu_limit_after_activation`) and tanh-`GEGLU`
(`activation_type="geglu"`) support to MoE/QMoE schemas, and added
`GatherPacked` (CPU/CUDA), `MatMulPacked` (CUDA, including native Bolt
`global_scale`), and `MoEPacked` (CUDA) to consume native
GGUF/block-packed quantized weights directly
([#33394](microsoft/onnxruntime#33394),
[#33954](microsoft/onnxruntime#33954),
[#33633](microsoft/onnxruntime#33633),
[#33953](microsoft/onnxruntime#33953),
[#33962](microsoft/onnxruntime#33962)).
- Added `PackedSparseAttentionIndexerMerge` for multi-token prediction
index sharing and optional `state_status` / mid-state outputs on
`PackedSparseAttentionIndexer` across CUDA and WebGPU; extended
`GatedAdd` with an optional projected `gate_weight` across CPU, CUDA,
and WebGPU; and supported empty unpacked weights in CPU and WebGPU
`MatMulNBits`
([#33224](microsoft/onnxruntime#33224),
[#33383](microsoft/onnxruntime#33383),
[#33472](microsoft/onnxruntime#33472),
[#33470](microsoft/onnxruntime#33470),
[#33559](microsoft/onnxruntime#33559)).
- Registered core contrib schemas for provider-targeted LLM and
speculative-decoding operators, including `SwiGLU`, `RMSNormRotary`,
`HyperConnectionLowRankMix`, `RouterGateUpProj`,
`SpeculativePrefixCommit`, `GatedDeltaNetStateReplay`,
`ConvStateReplay`, and `MatMulNBitsQLoRA`
([#32126](microsoft/onnxruntime#32126),
[#33469](microsoft/onnxruntime#33469),
[#33384](microsoft/onnxruntime#33384),
[#33650](microsoft/onnxruntime#33650),
[#33474](microsoft/onnxruntime#33474),
[#33213](microsoft/onnxruntime#33213),
[#33493](microsoft/onnxruntime#33493),
[#33370](microsoft/onnxruntime#33370)).
### Plugin Execution Provider Infrastructure
- Fixed an `OrtMemoryInfo` ABI mismatch across the ORT/plugin EP
boundary by converting `OrtMemoryInfo` instances exclusively through the
C API ([#33571](microsoft/onnxruntime#33571)).
- Zero-initialized `OrtEpFactory` structs in provider-bridge EP
factories so unimplemented optional callbacks are null, and prevented
double DLL initialization and handle leaks when probing and registering
plugin EP libraries
([#33117](microsoft/onnxruntime#33117),
[#33701](microsoft/onnxruntime#33701)).
- Hardened portable model-package EP recreation, custom-domain
preservation, UTF-8 path handling, and `shared_assets` path confinement
and symlink validation
([#32790](microsoft/onnxruntime#32790),
[#33826](microsoft/onnxruntime#33826)).
... (truncated)
Commits viewable in [compare
view](microsoft/onnxruntime@v1.31.1...v1.32.0).
</details>
Updated
[Microsoft.ML.OnnxRuntime.Managed](https://github.com/microsoft/onnxruntime)
from 1.31.1 to 1.32.0.
<details>
<summary>Release notes</summary>
_Sourced from [Microsoft.ML.OnnxRuntime.Managed's
releases](https://github.com/microsoft/onnxruntime/releases)._
## 1.32.0
ONNX Runtime 1.32.0 reintegrates ONNX 1.23.2 with opset 28 support,
introduces CPU/GPU hybrid MoE offloading and shared-layout CPU QMoE
kernels, expands device-memory, weightless-compilation, and Python
model-package APIs, dramatically accelerates session initialization on
large graphs, and strengthens runtime and optimizer reliability. These
notes cover changes since ONNX Runtime 1.31.1. CUDA and WebGPU kernel
updates are summarized here; detailed provider notes are covered by
their separate plugin EP releases.
## Highlights
- Reintegrated **ONNX 1.23.2** with **opset 28** operator support
(`SpaceToDepth`, `Mod`, signed `BitShift`, and `BFloat16` extensions)
and updated **DLPack to v1.3** across all implemented ORT numeric and
boolean tensor types
([#33134](microsoft/onnxruntime#33134),
[#32362](microsoft/onnxruntime#32362),
[#33848](microsoft/onnxruntime#33848)).
- Added a session-scoped **CPU/GPU hybrid MoE offload** planner and
execution backend for quantized `QMoE` (NVFP4, INT4, INT3, and INT2) and
block-packed `MoEPacked` models, backed by shared-layout AVX-512/AVX2
CPU `QMoE` kernels and memory-mapped expert storage
([#33450](microsoft/onnxruntime#33450),
[#33451](microsoft/onnxruntime#33451),
[#33452](microsoft/onnxruntime#33452),
[#33973](microsoft/onnxruntime#33973)).
- Expanded core APIs with `OrtApi::EpDevice_GetMemoryCapacity`, explicit
multi-mode weightless compilation
(`ModelCompilationOptions_SetWeightlessMode` and
`ep.enable_weightless_mode`), Python bindings for the Model Package
APIs, and zero-copy CPU buffer borrowing in `AddExternalInitializers`
([#33447](microsoft/onnxruntime#33447),
[#32899](microsoft/onnxruntime#32899),
[#33640](microsoft/onnxruntime#33640),
[#33877](microsoft/onnxruntime#33877)).
- Eliminated superlinear bottlenecks in allocation planning, constant
sharing, graph I/O mapping, and topological sorting—reducing session
initialization on 100k-node constant-heavy graphs from minutes to
seconds—and added pipelined pinned-memory staging for external GPU
weight loading
([#33407](microsoft/onnxruntime#33407),
[#33939](microsoft/onnxruntime#33939),
[#34009](microsoft/onnxruntime#34009),
[#33839](microsoft/onnxruntime#33839)).
- Expanded shared generative AI and quantized contrib operators,
including Kimi Delta Attention (`GatedDeltaNet` per-key-dimension
decay), right-padded chunked prefill in `GroupQueryAttention`, native
GGUF/block-packed operators (`GatherPacked`, `MatMulPacked`,
`MoEPacked`), and sparse-attention index sharing
([#33817](microsoft/onnxruntime#33817),
[#33503](microsoft/onnxruntime#33503),
[#33953](microsoft/onnxruntime#33953),
[#33224](microsoft/onnxruntime#33224)).
## Announcements & Compatibility
- **CUDA Plugin EP v0.4.0** and **WebGPU Plugin EP v0.6.0** are matched
with ONNX Runtime 1.32.0
([#33214](microsoft/onnxruntime#33214)). For
**contrib-op compatibility**, core and the plugin must use the same
contributed-operator schemas; building both from the **same ONNX Runtime
revision** is recommended. Schema mismatches can cause incorrect
execution or crashes and may not be detected during plugin registration.
- **ACL EP removed:** Following its deprecation in 1.31.0, the Arm
Compute Library (`ACLExecutionProvider`) execution provider, its build
options, and its standalone C registration API have been removed.
Python's `get_all_providers()` no longer lists `ACLExecutionProvider`;
applications should remove it from explicit provider lists
([#33851](microsoft/onnxruntime#33851)).
- **Legacy provider-bridge `NvTensorRTRTXExecutionProvider` removed:**
In-tree provider-bridge `NvTensorRTRTXExecutionProvider` sources have
been removed in favor of the standalone NVIDIA TensorRT RTX plugin EP
registered via `OrtApi::RegisterExecutionProviderLibrary`.
`NvTensorRtRtxProviderOptions` is retained as a deprecated declaration
for source compatibility, and legacy `AppendExecutionProvider` calls now
report a deprecation error directing callers to the plugin EP
([#32276](microsoft/onnxruntime#32276)).
- **ONNX 1.23.2 integration:** ONNX has been upgraded to **1.23.2**,
raising supported opset ceilings to **28** and restoring
`ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT6E2M3` and
`ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT6E3M2` public element-type mappings
and `Cast` conversions
([#33134](microsoft/onnxruntime#33134)).
- **DLPack v1.3 boolean contract change:** DLPack has been upgraded to
**v1.3** with native `kDLBool` metadata and bidirectional support for
all implemented ORT numeric and boolean types, including all five
`float8` formats, packed `float4`, `int4`/`uint4`, and `int2`/`uint2`
(`float6` remains unsupported in DLPack). The low-level `is_bool_tensor`
override was removed, and `uint8` DLPack metadata now always imports as
`uint8`
([#33848](microsoft/onnxruntime#33848)).
- **Weightless compilation API migration:**
`OrtCompileApi::ModelCompilationOptions_SetWeightlessEnabled` and the
`ep.enable_weightless` / `ep.enable_weightless_ep_context_nodes` session
options are deprecated in favor of
`ModelCompilationOptions_SetWeightlessMode` and
`ep.enable_weightless_mode` (`"0"` = `NONE`, `"1"` = `EXTERNAL_ONLY`,
`"2"` = `ALL`). EPs can now advertise multi-mode capability via
`OrtWeightlessSupport_ALL_OR_EXTERNAL_ONLY` and
`weightless_supported_modes`
([#32899](microsoft/onnxruntime#32899)).
- **Training CI retirement:** Training build and test lanes were retired
from active GitHub Actions CI workflows while retaining inference build
coverage
([#33836](microsoft/onnxruntime#33836)).
- This is the last release for [CocoaPods](https://cocoapods.org/), and
we will switch to [Swift Package
Manager](https://docs.swift.org/latest/documentation/packagemanagerdocs/).
- We will use combo onnxruntime-gpu python package to reduce storage
size from next release. You can test it
[here](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ORT-Nightly/PyPI/onnxruntime-gpu/overview)
## New Features
### Core APIs & Runtime
- Added `OrtApi::EpDevice_GetMemoryCapacity` (C++
`EpDevice::GetMemoryCapacity()`) and
`OrtEpFactory::GetDeviceMemoryCapacity` to query device-global total
memory and estimated currently available memory (`DEFAULT` or
`HOST_ACCESSIBLE`), with built-in and plugin CUDA EP support
([#33447](microsoft/onnxruntime#33447)).
- Expanded the app–EP weightless contract so EPs advertise all supported
weightless modes (`weightless_supported_modes`), applications request an
explicit mode (`ModelCompilationOptions_SetWeightlessMode` /
`ep.enable_weightless_mode`), and compiled EPContext models record
`weightless_mode` and `external_dependencies` (`source_model`,
`external_weights`, `ep_context_binary`) in model metadata
([#32899](microsoft/onnxruntime#32899)).
- Extended `session.use_external_initializer_file_buffers_directly="1"`
to `AddExternalInitializers` for zero-copy borrowing of naturally
aligned CPU buffers larger than 127 bytes on little-endian hosts, and
documented the read-only guarantee for external initializers added via
`OrtModelEditorApi::AddInitializerToGraph` with `data_is_external=true`
([#33877](microsoft/onnxruntime#33877),
[#33717](microsoft/onnxruntime#33717)).
- Added lock-free multi-worker pinned-memory staging
(`session.external_data_loader.*`) to accelerate external initializer
loading into GPU memory across built-in and plugin CUDA EPs
([#33839](microsoft/onnxruntime#33839)).
- Added a session-scoped MoE/QMoE CPU/GPU offload planner and
memory-budget ledger (`session.moe.offload.*`), generic
`session.kernel_dispatch_policy` (`auto`, `latency`, `memory`, `safe`)
configuration plumbing, Level-1 workspace/prepack memory accounting
fallback for nodes absent from memory-profile CSVs, and internal
capability-planning protocol types
([#33451](microsoft/onnxruntime#33451),
[#33042](microsoft/onnxruntime#33042),
[#33311](microsoft/onnxruntime#33311),
[#33841](microsoft/onnxruntime#33841)).
- Enabled cross-thread reuse of idle device streams when graph capture
is disabled and providers register device-selection callbacks, and
enabled compile-session profiling (`<profile_file>_compile`) in
`onnxruntime_perf_test` when `-p` is used with `--compile_ep_context`
([#33053](microsoft/onnxruntime#33053),
[#33694](microsoft/onnxruntime#33694)).
- Made `onnxruntime_DISABLE_EXCEPTIONS` (`--disable_exceptions`) usable
without `onnxruntime_MINIMAL_BUILD` and added the
`onnxruntime_DISABLE_STREAM=ON` CMake option to disable device-stream
support in non-minimal builds
([#33120](microsoft/onnxruntime#33120),
[#33482](microsoft/onnxruntime#33482)).
### ONNX Opset 28 & Shared Contrib Operators
These updates affect core ONNX and contributed operator schemas and
cross-provider behavior, not only the separately released GPU plugins.
- Added ONNX opset 28 operator support: native `SpaceToDepth` (`DCR` and
`CRD`) on CPU and CUDA with version-split registrations across DML, JS,
WebGPU, and QNN; opset-28 floating floor-modulus `Mod` on CPU, CUDA, and
DML; opset-28 `BFloat16` support for `Compress`, `OneHot`,
`ReverseSequence`, `Unique`, `Celu`, and `Einsum` on CPU (plus
`Compress`, `OneHot`, and `ReverseSequence` on CUDA); and `BFloat16`
`Pad` (opsets 21–25) on CPU and CUDA
([#32362](microsoft/onnxruntime#32362),
[#32369](microsoft/onnxruntime#32369),
[#32378](microsoft/onnxruntime#32378),
[#34062](microsoft/onnxruntime#34062)).
- Extended `GatedDeltaNet` with Kimi Delta Attention (KDA)
per-key-dimension decay across CPU, CUDA, and WebGPU, and extended
`GatedDeltaNet` and `CausalConvWithState` state/gate handling for
multi-token speculative decoding
([#33817](microsoft/onnxruntime#33817),
[#33471](microsoft/onnxruntime#33471)).
- Added an optional `query_seqlens` input to `GroupQueryAttention` to
skip right-padded prefill tokens and idle batch entries
(`query_seqlens[b] = 0`) on CPU and CUDA, and added `q_norm_epsilon` /
`k_norm_epsilon` attributes to preserve distinct Q and K normalization
epsilons during GQA pre-norm fusion
([#33503](microsoft/onnxruntime#33503),
[#33505](microsoft/onnxruntime#33505),
[#33832](microsoft/onnxruntime#33832),
[#33363](microsoft/onnxruntime#33363)).
- Defined the INT3 `QMoE` v1 packed-code contract, added opt-in
post-SiLU clipping (`swiglu_limit_after_activation`) and tanh-`GEGLU`
(`activation_type="geglu"`) support to MoE/QMoE schemas, and added
`GatherPacked` (CPU/CUDA), `MatMulPacked` (CUDA, including native Bolt
`global_scale`), and `MoEPacked` (CUDA) to consume native
GGUF/block-packed quantized weights directly
([#33394](microsoft/onnxruntime#33394),
[#33954](microsoft/onnxruntime#33954),
[#33633](microsoft/onnxruntime#33633),
[#33953](microsoft/onnxruntime#33953),
[#33962](microsoft/onnxruntime#33962)).
- Added `PackedSparseAttentionIndexerMerge` for multi-token prediction
index sharing and optional `state_status` / mid-state outputs on
`PackedSparseAttentionIndexer` across CUDA and WebGPU; extended
`GatedAdd` with an optional projected `gate_weight` across CPU, CUDA,
and WebGPU; and supported empty unpacked weights in CPU and WebGPU
`MatMulNBits`
([#33224](microsoft/onnxruntime#33224),
[#33383](microsoft/onnxruntime#33383),
[#33472](microsoft/onnxruntime#33472),
[#33470](microsoft/onnxruntime#33470),
[#33559](microsoft/onnxruntime#33559)).
- Registered core contrib schemas for provider-targeted LLM and
speculative-decoding operators, including `SwiGLU`, `RMSNormRotary`,
`HyperConnectionLowRankMix`, `RouterGateUpProj`,
`SpeculativePrefixCommit`, `GatedDeltaNetStateReplay`,
`ConvStateReplay`, and `MatMulNBitsQLoRA`
([#32126](microsoft/onnxruntime#32126),
[#33469](microsoft/onnxruntime#33469),
[#33384](microsoft/onnxruntime#33384),
[#33650](microsoft/onnxruntime#33650),
[#33474](microsoft/onnxruntime#33474),
[#33213](microsoft/onnxruntime#33213),
[#33493](microsoft/onnxruntime#33493),
[#33370](microsoft/onnxruntime#33370)).
### Plugin Execution Provider Infrastructure
- Fixed an `OrtMemoryInfo` ABI mismatch across the ORT/plugin EP
boundary by converting `OrtMemoryInfo` instances exclusively through the
C API ([#33571](microsoft/onnxruntime#33571)).
- Zero-initialized `OrtEpFactory` structs in provider-bridge EP
factories so unimplemented optional callbacks are null, and prevented
double DLL initialization and handle leaks when probing and registering
plugin EP libraries
([#33117](microsoft/onnxruntime#33117),
[#33701](microsoft/onnxruntime#33701)).
- Hardened portable model-package EP recreation, custom-domain
preservation, UTF-8 path handling, and `shared_assets` path confinement
and symlink validation
([#32790](microsoft/onnxruntime#32790),
[#33826](microsoft/onnxruntime#33826)).
... (truncated)
Commits viewable in [compare
view](microsoft/onnxruntime@v1.31.1...v1.32.0).
</details>
Updated [Microsoft.NET.Test.Sdk](https://github.com/microsoft/vstest)
from 18.11.0 to 18.12.0.
<details>
<summary>Release notes</summary>
_Sourced from [Microsoft.NET.Test.Sdk's
releases](https://github.com/microsoft/vstest/releases)._
## 18.12.0
## Security fix
This release includes the fix for CVE-2026-105544, an elevation of
privilege vulnerability in the VSTest artifact post-processor. The
post-processor could consume a forged `executionComplete.json` written
by another local user, and then run attacker-controlled output handling.
Affected: `Microsoft.TestPlatform.CLI` 18.11.0 and earlier. Every
version from 18.11.1 on is fixed, including 18.12.0.
`Microsoft.TestPlatform.CLI` is the VSTest build that we ship into the
.NET SDK. The recommended way to get this fix is to update your .NET SDK
to the latest patched version, see the advisory for the SDK versions.
The fix is also released in the `Microsoft.TestPlatform.CLI` package,
for completeness. That package is rarely used directly. It is not a
dependency of `Microsoft.NET.Test.Sdk` or `Microsoft.TestPlatform`, so
those packages do not bring it in.
Details: microsoft/vstest#16654 and
dotnet/announcements#517
Using custom Docker images or Nano Server? See
microsoft/vstest#16652
## What's Changed
* Fix TestCaseStart/Stop events dropped for data-driven tests with
shared TestCase.Id by @nohwnd in
microsoft/vstest#16033
* Stop the discovery manager factory from shadowing discoveryCriteria by
@nohwnd in microsoft/vstest#16391
* Gate the HTTP link checker on a fingerprint of the broken links by
@nohwnd in microsoft/vstest#16413
* Quote boolean conditions in shipped Microsoft.TestPlatform.targets by
@nohwnd in microsoft/vstest#16404
* Obsolete API hygiene: the compatibility-safe changes by @azat-msft in
microsoft/vstest#16420
* Add xxHash128 test ids, available but not default by @azat-msft in
microsoft/vstest#16378
* Isolate attachment manager test temporary files by @mmitche in
microsoft/vstest#16463
* Test ids report logger by @azat-msft in
microsoft/vstest#16443
* Isolate code coverage files in published output by @azat-msft in
microsoft/vstest#16483
* Preserve OutputType for Android test projects by @jonathanpeppers in
microsoft/vstest#16496
* Fix dotnet host resolution under proot (ARM64/Termux) by @azat-msft
in microsoft/vstest#16462
* Mark the xxHash128 test id API as experimental by @azat-msft in
microsoft/vstest#16465
* Ship Microsoft.Diagnostics.NETCore.Client in Microsoft.TestPlatform
and Portable packages by @nohwnd in
microsoft/vstest#16633
* Fix localized resource paths in CLI source-build/non-Windows packages
by @nohwnd in microsoft/vstest#16644
**Full Changelog**:
microsoft/vstest@v18.11.0...v18.12.0
## 18.11.1
## Security fix
This release fixes CVE-2026-105544, an elevation of privilege
vulnerability in the VSTest artifact post-processor. The post-processor
could consume a forged `executionComplete.json` written by another local
user, and then run attacker-controlled output handling.
Affected: `Microsoft.TestPlatform.CLI` 18.11.0.
Fixed in: `Microsoft.TestPlatform.CLI` 18.11.1.
`Microsoft.TestPlatform.CLI` is the VSTest build that we ship into the
.NET SDK. The recommended way to get this fix is to update your .NET SDK
to the latest patched version, see the advisory for the SDK versions.
The fix is also released in the `Microsoft.TestPlatform.CLI` package,
for completeness. That package is rarely used directly. It is not a
dependency of `Microsoft.NET.Test.Sdk` or `Microsoft.TestPlatform`, so
those packages do not bring it in.
Details: microsoft/vstest#16654 and
dotnet/announcements#517
Using custom Docker images or Nano Server? See
microsoft/vstest#16652
## What's Changed
* Security fix for CVE-2026-105544, see
microsoft/vstest#16654
* Ship Microsoft.Diagnostics.NETCore.Client in Microsoft.TestPlatform
and Portable packages by @nohwnd in
microsoft/vstest#16632
* Fix localized resource paths in CLI source-build packages by @nohwnd
in microsoft/vstest#16643
**Full Changelog**:
microsoft/vstest@v18.11.0...v18.11.1
Commits viewable in [compare
view](microsoft/vstest@v18.11.0...v18.12.0).
</details>
Dependabot will resolve any conflicts with this PR as long as you don't
alter it yourself. You can also trigger a rebase manually by commenting
`@dependabot rebase`.
[//]: # (dependabot-automerge-start)
[//]: # (dependabot-automerge-end)
---
<details>
<summary>Dependabot commands and options</summary>
<br />
You can trigger Dependabot actions by commenting on this PR:
- `@dependabot rebase` will rebase this PR
- `@dependabot recreate` will recreate this PR, overwriting any edits
that have been made to it
- `@dependabot show <dependency name> ignore conditions` will show all
of the ignore conditions of the specified dependency
- `@dependabot ignore <dependency name> major version` will close this
group update PR and stop Dependabot creating any more for the specific
dependency's major version (unless you unignore this specific
dependency's major version or upgrade to it yourself)
- `@dependabot ignore <dependency name> minor version` will close this
group update PR and stop Dependabot creating any more for the specific
dependency's minor version (unless you unignore this specific
dependency's minor version or upgrade to it yourself)
- `@dependabot ignore <dependency name>` will close this group update PR
and stop Dependabot creating any more for the specific dependency
(unless you unignore this specific dependency or upgrade to it yourself)
- `@dependabot unignore <dependency name>` will remove all of the ignore
conditions of the specified dependency
- `@dependabot unignore <dependency name> <ignore condition>` will
remove the ignore condition of the specified dependency and ignore
conditions
</details>
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Updated
[Microsoft.AspNetCore.Mvc.Testing](https://github.com/dotnet/dotnet)
from 10.0.14 to 10.0.15.
<details>
<summary>Release notes</summary>
_Sourced from [Microsoft.AspNetCore.Mvc.Testing's
releases](https://github.com/dotnet/dotnet/releases)._
No release notes found for this version range.
Commits viewable in [compare
view](https://github.com/dotnet/dotnet/commits).
</details>
Updated [Microsoft.Data.Sqlite](https://github.com/dotnet/dotnet) from
10.0.14 to 10.0.15.
<details>
<summary>Release notes</summary>
_Sourced from [Microsoft.Data.Sqlite's
releases](https://github.com/dotnet/dotnet/releases)._
No release notes found for this version range.
Commits viewable in [compare
view](https://github.com/dotnet/dotnet/commits).
</details>
Updated
[Microsoft.ML.OnnxRuntime](https://github.com/Microsoft/onnxruntime)
from 1.31.1 to 1.32.0.
<details>
<summary>Release notes</summary>
_Sourced from [Microsoft.ML.OnnxRuntime's
releases](https://github.com/Microsoft/onnxruntime/releases)._
## 1.32.0
ONNX Runtime 1.32.0 reintegrates ONNX 1.23.2 with opset 28 support,
introduces CPU/GPU hybrid MoE offloading and shared-layout CPU QMoE
kernels, expands device-memory, weightless-compilation, and Python
model-package APIs, dramatically accelerates session initialization on
large graphs, and strengthens runtime and optimizer reliability. These
notes cover changes since ONNX Runtime 1.31.1. CUDA and WebGPU kernel
updates are summarized here; detailed provider notes are covered by
their separate plugin EP releases.
## Highlights
- Reintegrated **ONNX 1.23.2** with **opset 28** operator support
(`SpaceToDepth`, `Mod`, signed `BitShift`, and `BFloat16` extensions)
and updated **DLPack to v1.3** across all implemented ORT numeric and
boolean tensor types
([#33134](microsoft/onnxruntime#33134),
[#32362](microsoft/onnxruntime#32362),
[#33848](microsoft/onnxruntime#33848)).
- Added a session-scoped **CPU/GPU hybrid MoE offload** planner and
execution backend for quantized `QMoE` (NVFP4, INT4, INT3, and INT2) and
block-packed `MoEPacked` models, backed by shared-layout AVX-512/AVX2
CPU `QMoE` kernels and memory-mapped expert storage
([#33450](microsoft/onnxruntime#33450),
[#33451](microsoft/onnxruntime#33451),
[#33452](microsoft/onnxruntime#33452),
[#33973](microsoft/onnxruntime#33973)).
- Expanded core APIs with `OrtApi::EpDevice_GetMemoryCapacity`, explicit
multi-mode weightless compilation
(`ModelCompilationOptions_SetWeightlessMode` and
`ep.enable_weightless_mode`), Python bindings for the Model Package
APIs, and zero-copy CPU buffer borrowing in `AddExternalInitializers`
([#33447](microsoft/onnxruntime#33447),
[#32899](microsoft/onnxruntime#32899),
[#33640](microsoft/onnxruntime#33640),
[#33877](microsoft/onnxruntime#33877)).
- Eliminated superlinear bottlenecks in allocation planning, constant
sharing, graph I/O mapping, and topological sorting—reducing session
initialization on 100k-node constant-heavy graphs from minutes to
seconds—and added pipelined pinned-memory staging for external GPU
weight loading
([#33407](microsoft/onnxruntime#33407),
[#33939](microsoft/onnxruntime#33939),
[#34009](microsoft/onnxruntime#34009),
[#33839](microsoft/onnxruntime#33839)).
- Expanded shared generative AI and quantized contrib operators,
including Kimi Delta Attention (`GatedDeltaNet` per-key-dimension
decay), right-padded chunked prefill in `GroupQueryAttention`, native
GGUF/block-packed operators (`GatherPacked`, `MatMulPacked`,
`MoEPacked`), and sparse-attention index sharing
([#33817](microsoft/onnxruntime#33817),
[#33503](microsoft/onnxruntime#33503),
[#33953](microsoft/onnxruntime#33953),
[#33224](microsoft/onnxruntime#33224)).
## Announcements & Compatibility
- **CUDA Plugin EP v0.4.0** and **WebGPU Plugin EP v0.6.0** are matched
with ONNX Runtime 1.32.0
([#33214](microsoft/onnxruntime#33214)). For
**contrib-op compatibility**, core and the plugin must use the same
contributed-operator schemas; building both from the **same ONNX Runtime
revision** is recommended. Schema mismatches can cause incorrect
execution or crashes and may not be detected during plugin registration.
- **ACL EP removed:** Following its deprecation in 1.31.0, the Arm
Compute Library (`ACLExecutionProvider`) execution provider, its build
options, and its standalone C registration API have been removed.
Python's `get_all_providers()` no longer lists `ACLExecutionProvider`;
applications should remove it from explicit provider lists
([#33851](microsoft/onnxruntime#33851)).
- **Legacy provider-bridge `NvTensorRTRTXExecutionProvider` removed:**
In-tree provider-bridge `NvTensorRTRTXExecutionProvider` sources have
been removed in favor of the standalone NVIDIA TensorRT RTX plugin EP
registered via `OrtApi::RegisterExecutionProviderLibrary`.
`NvTensorRtRtxProviderOptions` is retained as a deprecated declaration
for source compatibility, and legacy `AppendExecutionProvider` calls now
report a deprecation error directing callers to the plugin EP
([#32276](microsoft/onnxruntime#32276)).
- **ONNX 1.23.2 integration:** ONNX has been upgraded to **1.23.2**,
raising supported opset ceilings to **28** and restoring
`ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT6E2M3` and
`ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT6E3M2` public element-type mappings
and `Cast` conversions
([#33134](microsoft/onnxruntime#33134)).
- **DLPack v1.3 boolean contract change:** DLPack has been upgraded to
**v1.3** with native `kDLBool` metadata and bidirectional support for
all implemented ORT numeric and boolean types, including all five
`float8` formats, packed `float4`, `int4`/`uint4`, and `int2`/`uint2`
(`float6` remains unsupported in DLPack). The low-level `is_bool_tensor`
override was removed, and `uint8` DLPack metadata now always imports as
`uint8`
([#33848](microsoft/onnxruntime#33848)).
- **Weightless compilation API migration:**
`OrtCompileApi::ModelCompilationOptions_SetWeightlessEnabled` and the
`ep.enable_weightless` / `ep.enable_weightless_ep_context_nodes` session
options are deprecated in favor of
`ModelCompilationOptions_SetWeightlessMode` and
`ep.enable_weightless_mode` (`"0"` = `NONE`, `"1"` = `EXTERNAL_ONLY`,
`"2"` = `ALL`). EPs can now advertise multi-mode capability via
`OrtWeightlessSupport_ALL_OR_EXTERNAL_ONLY` and
`weightless_supported_modes`
([#32899](microsoft/onnxruntime#32899)).
- **Training CI retirement:** Training build and test lanes were retired
from active GitHub Actions CI workflows while retaining inference build
coverage
([#33836](microsoft/onnxruntime#33836)).
- This is the last release for [CocoaPods](https://cocoapods.org/), and
we will switch to [Swift Package
Manager](https://docs.swift.org/latest/documentation/packagemanagerdocs/).
- We will use combo onnxruntime-gpu python package to reduce storage
size from next release. You can test it
[here](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ORT-Nightly/PyPI/onnxruntime-gpu/overview)
## New Features
### Core APIs & Runtime
- Added `OrtApi::EpDevice_GetMemoryCapacity` (C++
`EpDevice::GetMemoryCapacity()`) and
`OrtEpFactory::GetDeviceMemoryCapacity` to query device-global total
memory and estimated currently available memory (`DEFAULT` or
`HOST_ACCESSIBLE`), with built-in and plugin CUDA EP support
([#33447](microsoft/onnxruntime#33447)).
- Expanded the app–EP weightless contract so EPs advertise all supported
weightless modes (`weightless_supported_modes`), applications request an
explicit mode (`ModelCompilationOptions_SetWeightlessMode` /
`ep.enable_weightless_mode`), and compiled EPContext models record
`weightless_mode` and `external_dependencies` (`source_model`,
`external_weights`, `ep_context_binary`) in model metadata
([#32899](microsoft/onnxruntime#32899)).
- Extended `session.use_external_initializer_file_buffers_directly="1"`
to `AddExternalInitializers` for zero-copy borrowing of naturally
aligned CPU buffers larger than 127 bytes on little-endian hosts, and
documented the read-only guarantee for external initializers added via
`OrtModelEditorApi::AddInitializerToGraph` with `data_is_external=true`
([#33877](microsoft/onnxruntime#33877),
[#33717](microsoft/onnxruntime#33717)).
- Added lock-free multi-worker pinned-memory staging
(`session.external_data_loader.*`) to accelerate external initializer
loading into GPU memory across built-in and plugin CUDA EPs
([#33839](microsoft/onnxruntime#33839)).
- Added a session-scoped MoE/QMoE CPU/GPU offload planner and
memory-budget ledger (`session.moe.offload.*`), generic
`session.kernel_dispatch_policy` (`auto`, `latency`, `memory`, `safe`)
configuration plumbing, Level-1 workspace/prepack memory accounting
fallback for nodes absent from memory-profile CSVs, and internal
capability-planning protocol types
([#33451](microsoft/onnxruntime#33451),
[#33042](microsoft/onnxruntime#33042),
[#33311](microsoft/onnxruntime#33311),
[#33841](microsoft/onnxruntime#33841)).
- Enabled cross-thread reuse of idle device streams when graph capture
is disabled and providers register device-selection callbacks, and
enabled compile-session profiling (`<profile_file>_compile`) in
`onnxruntime_perf_test` when `-p` is used with `--compile_ep_context`
([#33053](microsoft/onnxruntime#33053),
[#33694](microsoft/onnxruntime#33694)).
- Made `onnxruntime_DISABLE_EXCEPTIONS` (`--disable_exceptions`) usable
without `onnxruntime_MINIMAL_BUILD` and added the
`onnxruntime_DISABLE_STREAM=ON` CMake option to disable device-stream
support in non-minimal builds
([#33120](microsoft/onnxruntime#33120),
[#33482](microsoft/onnxruntime#33482)).
### ONNX Opset 28 & Shared Contrib Operators
These updates affect core ONNX and contributed operator schemas and
cross-provider behavior, not only the separately released GPU plugins.
- Added ONNX opset 28 operator support: native `SpaceToDepth` (`DCR` and
`CRD`) on CPU and CUDA with version-split registrations across DML, JS,
WebGPU, and QNN; opset-28 floating floor-modulus `Mod` on CPU, CUDA, and
DML; opset-28 `BFloat16` support for `Compress`, `OneHot`,
`ReverseSequence`, `Unique`, `Celu`, and `Einsum` on CPU (plus
`Compress`, `OneHot`, and `ReverseSequence` on CUDA); and `BFloat16`
`Pad` (opsets 21–25) on CPU and CUDA
([#32362](microsoft/onnxruntime#32362),
[#32369](microsoft/onnxruntime#32369),
[#32378](microsoft/onnxruntime#32378),
[#34062](microsoft/onnxruntime#34062)).
- Extended `GatedDeltaNet` with Kimi Delta Attention (KDA)
per-key-dimension decay across CPU, CUDA, and WebGPU, and extended
`GatedDeltaNet` and `CausalConvWithState` state/gate handling for
multi-token speculative decoding
([#33817](microsoft/onnxruntime#33817),
[#33471](microsoft/onnxruntime#33471)).
- Added an optional `query_seqlens` input to `GroupQueryAttention` to
skip right-padded prefill tokens and idle batch entries
(`query_seqlens[b] = 0`) on CPU and CUDA, and added `q_norm_epsilon` /
`k_norm_epsilon` attributes to preserve distinct Q and K normalization
epsilons during GQA pre-norm fusion
([#33503](microsoft/onnxruntime#33503),
[#33505](microsoft/onnxruntime#33505),
[#33832](microsoft/onnxruntime#33832),
[#33363](microsoft/onnxruntime#33363)).
- Defined the INT3 `QMoE` v1 packed-code contract, added opt-in
post-SiLU clipping (`swiglu_limit_after_activation`) and tanh-`GEGLU`
(`activation_type="geglu"`) support to MoE/QMoE schemas, and added
`GatherPacked` (CPU/CUDA), `MatMulPacked` (CUDA, including native Bolt
`global_scale`), and `MoEPacked` (CUDA) to consume native
GGUF/block-packed quantized weights directly
([#33394](microsoft/onnxruntime#33394),
[#33954](microsoft/onnxruntime#33954),
[#33633](microsoft/onnxruntime#33633),
[#33953](microsoft/onnxruntime#33953),
[#33962](microsoft/onnxruntime#33962)).
- Added `PackedSparseAttentionIndexerMerge` for multi-token prediction
index sharing and optional `state_status` / mid-state outputs on
`PackedSparseAttentionIndexer` across CUDA and WebGPU; extended
`GatedAdd` with an optional projected `gate_weight` across CPU, CUDA,
and WebGPU; and supported empty unpacked weights in CPU and WebGPU
`MatMulNBits`
([#33224](microsoft/onnxruntime#33224),
[#33383](microsoft/onnxruntime#33383),
[#33472](microsoft/onnxruntime#33472),
[#33470](microsoft/onnxruntime#33470),
[#33559](microsoft/onnxruntime#33559)).
- Registered core contrib schemas for provider-targeted LLM and
speculative-decoding operators, including `SwiGLU`, `RMSNormRotary`,
`HyperConnectionLowRankMix`, `RouterGateUpProj`,
`SpeculativePrefixCommit`, `GatedDeltaNetStateReplay`,
`ConvStateReplay`, and `MatMulNBitsQLoRA`
([#32126](microsoft/onnxruntime#32126),
[#33469](microsoft/onnxruntime#33469),
[#33384](microsoft/onnxruntime#33384),
[#33650](microsoft/onnxruntime#33650),
[#33474](microsoft/onnxruntime#33474),
[#33213](microsoft/onnxruntime#33213),
[#33493](microsoft/onnxruntime#33493),
[#33370](microsoft/onnxruntime#33370)).
### Plugin Execution Provider Infrastructure
- Fixed an `OrtMemoryInfo` ABI mismatch across the ORT/plugin EP
boundary by converting `OrtMemoryInfo` instances exclusively through the
C API ([#33571](microsoft/onnxruntime#33571)).
- Zero-initialized `OrtEpFactory` structs in provider-bridge EP
factories so unimplemented optional callbacks are null, and prevented
double DLL initialization and handle leaks when probing and registering
plugin EP libraries
([#33117](microsoft/onnxruntime#33117),
[#33701](microsoft/onnxruntime#33701)).
- Hardened portable model-package EP recreation, custom-domain
preservation, UTF-8 path handling, and `shared_assets` path confinement
and symlink validation
([#32790](microsoft/onnxruntime#32790),
[#33826](microsoft/onnxruntime#33826)).
... (truncated)
Commits viewable in [compare
view](microsoft/onnxruntime@v1.31.1...v1.32.0).
</details>
Updated
[Microsoft.ML.OnnxRuntime.Managed](https://github.com/microsoft/onnxruntime)
from 1.31.1 to 1.32.0.
<details>
<summary>Release notes</summary>
_Sourced from [Microsoft.ML.OnnxRuntime.Managed's
releases](https://github.com/microsoft/onnxruntime/releases)._
## 1.32.0
ONNX Runtime 1.32.0 reintegrates ONNX 1.23.2 with opset 28 support,
introduces CPU/GPU hybrid MoE offloading and shared-layout CPU QMoE
kernels, expands device-memory, weightless-compilation, and Python
model-package APIs, dramatically accelerates session initialization on
large graphs, and strengthens runtime and optimizer reliability. These
notes cover changes since ONNX Runtime 1.31.1. CUDA and WebGPU kernel
updates are summarized here; detailed provider notes are covered by
their separate plugin EP releases.
## Highlights
- Reintegrated **ONNX 1.23.2** with **opset 28** operator support
(`SpaceToDepth`, `Mod`, signed `BitShift`, and `BFloat16` extensions)
and updated **DLPack to v1.3** across all implemented ORT numeric and
boolean tensor types
([#33134](microsoft/onnxruntime#33134),
[#32362](microsoft/onnxruntime#32362),
[#33848](microsoft/onnxruntime#33848)).
- Added a session-scoped **CPU/GPU hybrid MoE offload** planner and
execution backend for quantized `QMoE` (NVFP4, INT4, INT3, and INT2) and
block-packed `MoEPacked` models, backed by shared-layout AVX-512/AVX2
CPU `QMoE` kernels and memory-mapped expert storage
([#33450](microsoft/onnxruntime#33450),
[#33451](microsoft/onnxruntime#33451),
[#33452](microsoft/onnxruntime#33452),
[#33973](microsoft/onnxruntime#33973)).
- Expanded core APIs with `OrtApi::EpDevice_GetMemoryCapacity`, explicit
multi-mode weightless compilation
(`ModelCompilationOptions_SetWeightlessMode` and
`ep.enable_weightless_mode`), Python bindings for the Model Package
APIs, and zero-copy CPU buffer borrowing in `AddExternalInitializers`
([#33447](microsoft/onnxruntime#33447),
[#32899](microsoft/onnxruntime#32899),
[#33640](microsoft/onnxruntime#33640),
[#33877](microsoft/onnxruntime#33877)).
- Eliminated superlinear bottlenecks in allocation planning, constant
sharing, graph I/O mapping, and topological sorting—reducing session
initialization on 100k-node constant-heavy graphs from minutes to
seconds—and added pipelined pinned-memory staging for external GPU
weight loading
([#33407](microsoft/onnxruntime#33407),
[#33939](microsoft/onnxruntime#33939),
[#34009](microsoft/onnxruntime#34009),
[#33839](microsoft/onnxruntime#33839)).
- Expanded shared generative AI and quantized contrib operators,
including Kimi Delta Attention (`GatedDeltaNet` per-key-dimension
decay), right-padded chunked prefill in `GroupQueryAttention`, native
GGUF/block-packed operators (`GatherPacked`, `MatMulPacked`,
`MoEPacked`), and sparse-attention index sharing
([#33817](microsoft/onnxruntime#33817),
[#33503](microsoft/onnxruntime#33503),
[#33953](microsoft/onnxruntime#33953),
[#33224](microsoft/onnxruntime#33224)).
## Announcements & Compatibility
- **CUDA Plugin EP v0.4.0** and **WebGPU Plugin EP v0.6.0** are matched
with ONNX Runtime 1.32.0
([#33214](microsoft/onnxruntime#33214)). For
**contrib-op compatibility**, core and the plugin must use the same
contributed-operator schemas; building both from the **same ONNX Runtime
revision** is recommended. Schema mismatches can cause incorrect
execution or crashes and may not be detected during plugin registration.
- **ACL EP removed:** Following its deprecation in 1.31.0, the Arm
Compute Library (`ACLExecutionProvider`) execution provider, its build
options, and its standalone C registration API have been removed.
Python's `get_all_providers()` no longer lists `ACLExecutionProvider`;
applications should remove it from explicit provider lists
([#33851](microsoft/onnxruntime#33851)).
- **Legacy provider-bridge `NvTensorRTRTXExecutionProvider` removed:**
In-tree provider-bridge `NvTensorRTRTXExecutionProvider` sources have
been removed in favor of the standalone NVIDIA TensorRT RTX plugin EP
registered via `OrtApi::RegisterExecutionProviderLibrary`.
`NvTensorRtRtxProviderOptions` is retained as a deprecated declaration
for source compatibility, and legacy `AppendExecutionProvider` calls now
report a deprecation error directing callers to the plugin EP
([#32276](microsoft/onnxruntime#32276)).
- **ONNX 1.23.2 integration:** ONNX has been upgraded to **1.23.2**,
raising supported opset ceilings to **28** and restoring
`ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT6E2M3` and
`ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT6E3M2` public element-type mappings
and `Cast` conversions
([#33134](microsoft/onnxruntime#33134)).
- **DLPack v1.3 boolean contract change:** DLPack has been upgraded to
**v1.3** with native `kDLBool` metadata and bidirectional support for
all implemented ORT numeric and boolean types, including all five
`float8` formats, packed `float4`, `int4`/`uint4`, and `int2`/`uint2`
(`float6` remains unsupported in DLPack). The low-level `is_bool_tensor`
override was removed, and `uint8` DLPack metadata now always imports as
`uint8`
([#33848](microsoft/onnxruntime#33848)).
- **Weightless compilation API migration:**
`OrtCompileApi::ModelCompilationOptions_SetWeightlessEnabled` and the
`ep.enable_weightless` / `ep.enable_weightless_ep_context_nodes` session
options are deprecated in favor of
`ModelCompilationOptions_SetWeightlessMode` and
`ep.enable_weightless_mode` (`"0"` = `NONE`, `"1"` = `EXTERNAL_ONLY`,
`"2"` = `ALL`). EPs can now advertise multi-mode capability via
`OrtWeightlessSupport_ALL_OR_EXTERNAL_ONLY` and
`weightless_supported_modes`
([#32899](microsoft/onnxruntime#32899)).
- **Training CI retirement:** Training build and test lanes were retired
from active GitHub Actions CI workflows while retaining inference build
coverage
([#33836](microsoft/onnxruntime#33836)).
- This is the last release for [CocoaPods](https://cocoapods.org/), and
we will switch to [Swift Package
Manager](https://docs.swift.org/latest/documentation/packagemanagerdocs/).
- We will use combo onnxruntime-gpu python package to reduce storage
size from next release. You can test it
[here](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ORT-Nightly/PyPI/onnxruntime-gpu/overview)
## New Features
### Core APIs & Runtime
- Added `OrtApi::EpDevice_GetMemoryCapacity` (C++
`EpDevice::GetMemoryCapacity()`) and
`OrtEpFactory::GetDeviceMemoryCapacity` to query device-global total
memory and estimated currently available memory (`DEFAULT` or
`HOST_ACCESSIBLE`), with built-in and plugin CUDA EP support
([#33447](microsoft/onnxruntime#33447)).
- Expanded the app–EP weightless contract so EPs advertise all supported
weightless modes (`weightless_supported_modes`), applications request an
explicit mode (`ModelCompilationOptions_SetWeightlessMode` /
`ep.enable_weightless_mode`), and compiled EPContext models record
`weightless_mode` and `external_dependencies` (`source_model`,
`external_weights`, `ep_context_binary`) in model metadata
([#32899](microsoft/onnxruntime#32899)).
- Extended `session.use_external_initializer_file_buffers_directly="1"`
to `AddExternalInitializers` for zero-copy borrowing of naturally
aligned CPU buffers larger than 127 bytes on little-endian hosts, and
documented the read-only guarantee for external initializers added via
`OrtModelEditorApi::AddInitializerToGraph` with `data_is_external=true`
([#33877](microsoft/onnxruntime#33877),
[#33717](microsoft/onnxruntime#33717)).
- Added lock-free multi-worker pinned-memory staging
(`session.external_data_loader.*`) to accelerate external initializer
loading into GPU memory across built-in and plugin CUDA EPs
([#33839](microsoft/onnxruntime#33839)).
- Added a session-scoped MoE/QMoE CPU/GPU offload planner and
memory-budget ledger (`session.moe.offload.*`), generic
`session.kernel_dispatch_policy` (`auto`, `latency`, `memory`, `safe`)
configuration plumbing, Level-1 workspace/prepack memory accounting
fallback for nodes absent from memory-profile CSVs, and internal
capability-planning protocol types
([#33451](microsoft/onnxruntime#33451),
[#33042](microsoft/onnxruntime#33042),
[#33311](microsoft/onnxruntime#33311),
[#33841](microsoft/onnxruntime#33841)).
- Enabled cross-thread reuse of idle device streams when graph capture
is disabled and providers register device-selection callbacks, and
enabled compile-session profiling (`<profile_file>_compile`) in
`onnxruntime_perf_test` when `-p` is used with `--compile_ep_context`
([#33053](microsoft/onnxruntime#33053),
[#33694](microsoft/onnxruntime#33694)).
- Made `onnxruntime_DISABLE_EXCEPTIONS` (`--disable_exceptions`) usable
without `onnxruntime_MINIMAL_BUILD` and added the
`onnxruntime_DISABLE_STREAM=ON` CMake option to disable device-stream
support in non-minimal builds
([#33120](microsoft/onnxruntime#33120),
[#33482](microsoft/onnxruntime#33482)).
### ONNX Opset 28 & Shared Contrib Operators
These updates affect core ONNX and contributed operator schemas and
cross-provider behavior, not only the separately released GPU plugins.
- Added ONNX opset 28 operator support: native `SpaceToDepth` (`DCR` and
`CRD`) on CPU and CUDA with version-split registrations across DML, JS,
WebGPU, and QNN; opset-28 floating floor-modulus `Mod` on CPU, CUDA, and
DML; opset-28 `BFloat16` support for `Compress`, `OneHot`,
`ReverseSequence`, `Unique`, `Celu`, and `Einsum` on CPU (plus
`Compress`, `OneHot`, and `ReverseSequence` on CUDA); and `BFloat16`
`Pad` (opsets 21–25) on CPU and CUDA
([#32362](microsoft/onnxruntime#32362),
[#32369](microsoft/onnxruntime#32369),
[#32378](microsoft/onnxruntime#32378),
[#34062](microsoft/onnxruntime#34062)).
- Extended `GatedDeltaNet` with Kimi Delta Attention (KDA)
per-key-dimension decay across CPU, CUDA, and WebGPU, and extended
`GatedDeltaNet` and `CausalConvWithState` state/gate handling for
multi-token speculative decoding
([#33817](microsoft/onnxruntime#33817),
[#33471](microsoft/onnxruntime#33471)).
- Added an optional `query_seqlens` input to `GroupQueryAttention` to
skip right-padded prefill tokens and idle batch entries
(`query_seqlens[b] = 0`) on CPU and CUDA, and added `q_norm_epsilon` /
`k_norm_epsilon` attributes to preserve distinct Q and K normalization
epsilons during GQA pre-norm fusion
([#33503](microsoft/onnxruntime#33503),
[#33505](microsoft/onnxruntime#33505),
[#33832](microsoft/onnxruntime#33832),
[#33363](microsoft/onnxruntime#33363)).
- Defined the INT3 `QMoE` v1 packed-code contract, added opt-in
post-SiLU clipping (`swiglu_limit_after_activation`) and tanh-`GEGLU`
(`activation_type="geglu"`) support to MoE/QMoE schemas, and added
`GatherPacked` (CPU/CUDA), `MatMulPacked` (CUDA, including native Bolt
`global_scale`), and `MoEPacked` (CUDA) to consume native
GGUF/block-packed quantized weights directly
([#33394](microsoft/onnxruntime#33394),
[#33954](microsoft/onnxruntime#33954),
[#33633](microsoft/onnxruntime#33633),
[#33953](microsoft/onnxruntime#33953),
[#33962](microsoft/onnxruntime#33962)).
- Added `PackedSparseAttentionIndexerMerge` for multi-token prediction
index sharing and optional `state_status` / mid-state outputs on
`PackedSparseAttentionIndexer` across CUDA and WebGPU; extended
`GatedAdd` with an optional projected `gate_weight` across CPU, CUDA,
and WebGPU; and supported empty unpacked weights in CPU and WebGPU
`MatMulNBits`
([#33224](microsoft/onnxruntime#33224),
[#33383](microsoft/onnxruntime#33383),
[#33472](microsoft/onnxruntime#33472),
[#33470](microsoft/onnxruntime#33470),
[#33559](microsoft/onnxruntime#33559)).
- Registered core contrib schemas for provider-targeted LLM and
speculative-decoding operators, including `SwiGLU`, `RMSNormRotary`,
`HyperConnectionLowRankMix`, `RouterGateUpProj`,
`SpeculativePrefixCommit`, `GatedDeltaNetStateReplay`,
`ConvStateReplay`, and `MatMulNBitsQLoRA`
([#32126](microsoft/onnxruntime#32126),
[#33469](microsoft/onnxruntime#33469),
[#33384](microsoft/onnxruntime#33384),
[#33650](microsoft/onnxruntime#33650),
[#33474](microsoft/onnxruntime#33474),
[#33213](microsoft/onnxruntime#33213),
[#33493](microsoft/onnxruntime#33493),
[#33370](microsoft/onnxruntime#33370)).
### Plugin Execution Provider Infrastructure
- Fixed an `OrtMemoryInfo` ABI mismatch across the ORT/plugin EP
boundary by converting `OrtMemoryInfo` instances exclusively through the
C API ([#33571](microsoft/onnxruntime#33571)).
- Zero-initialized `OrtEpFactory` structs in provider-bridge EP
factories so unimplemented optional callbacks are null, and prevented
double DLL initialization and handle leaks when probing and registering
plugin EP libraries
([#33117](microsoft/onnxruntime#33117),
[#33701](microsoft/onnxruntime#33701)).
- Hardened portable model-package EP recreation, custom-domain
preservation, UTF-8 path handling, and `shared_assets` path confinement
and symlink validation
([#32790](microsoft/onnxruntime#32790),
[#33826](microsoft/onnxruntime#33826)).
... (truncated)
Commits viewable in [compare
view](microsoft/onnxruntime@v1.31.1...v1.32.0).
</details>
Updated [Microsoft.NET.Test.Sdk](https://github.com/microsoft/vstest)
from 18.11.0 to 18.12.0.
<details>
<summary>Release notes</summary>
_Sourced from [Microsoft.NET.Test.Sdk's
releases](https://github.com/microsoft/vstest/releases)._
## 18.12.0
## Security fix
This release includes the fix for CVE-2026-105544, an elevation of
privilege vulnerability in the VSTest artifact post-processor. The
post-processor could consume a forged `executionComplete.json` written
by another local user, and then run attacker-controlled output handling.
Affected: `Microsoft.TestPlatform.CLI` 18.11.0 and earlier. Every
version from 18.11.1 on is fixed, including 18.12.0.
`Microsoft.TestPlatform.CLI` is the VSTest build that we ship into the
.NET SDK. The recommended way to get this fix is to update your .NET SDK
to the latest patched version, see the advisory for the SDK versions.
The fix is also released in the `Microsoft.TestPlatform.CLI` package,
for completeness. That package is rarely used directly. It is not a
dependency of `Microsoft.NET.Test.Sdk` or `Microsoft.TestPlatform`, so
those packages do not bring it in.
Details: microsoft/vstest#16654 and
dotnet/announcements#517
Using custom Docker images or Nano Server? See
microsoft/vstest#16652
## What's Changed
* Fix TestCaseStart/Stop events dropped for data-driven tests with
shared TestCase.Id by @nohwnd in
microsoft/vstest#16033
* Stop the discovery manager factory from shadowing discoveryCriteria by
@nohwnd in microsoft/vstest#16391
* Gate the HTTP link checker on a fingerprint of the broken links by
@nohwnd in microsoft/vstest#16413
* Quote boolean conditions in shipped Microsoft.TestPlatform.targets by
@nohwnd in microsoft/vstest#16404
* Obsolete API hygiene: the compatibility-safe changes by @azat-msft in
microsoft/vstest#16420
* Add xxHash128 test ids, available but not default by @azat-msft in
microsoft/vstest#16378
* Isolate attachment manager test temporary files by @mmitche in
microsoft/vstest#16463
* Test ids report logger by @azat-msft in
microsoft/vstest#16443
* Isolate code coverage files in published output by @azat-msft in
microsoft/vstest#16483
* Preserve OutputType for Android test projects by @jonathanpeppers in
microsoft/vstest#16496
* Fix dotnet host resolution under proot (ARM64/Termux) by @azat-msft
in microsoft/vstest#16462
* Mark the xxHash128 test id API as experimental by @azat-msft in
microsoft/vstest#16465
* Ship Microsoft.Diagnostics.NETCore.Client in Microsoft.TestPlatform
and Portable packages by @nohwnd in
microsoft/vstest#16633
* Fix localized resource paths in CLI source-build/non-Windows packages
by @nohwnd in microsoft/vstest#16644
**Full Changelog**:
microsoft/vstest@v18.11.0...v18.12.0
## 18.11.1
## Security fix
This release fixes CVE-2026-105544, an elevation of privilege
vulnerability in the VSTest artifact post-processor. The post-processor
could consume a forged `executionComplete.json` written by another local
user, and then run attacker-controlled output handling.
Affected: `Microsoft.TestPlatform.CLI` 18.11.0.
Fixed in: `Microsoft.TestPlatform.CLI` 18.11.1.
`Microsoft.TestPlatform.CLI` is the VSTest build that we ship into the
.NET SDK. The recommended way to get this fix is to update your .NET SDK
to the latest patched version, see the advisory for the SDK versions.
The fix is also released in the `Microsoft.TestPlatform.CLI` package,
for completeness. That package is rarely used directly. It is not a
dependency of `Microsoft.NET.Test.Sdk` or `Microsoft.TestPlatform`, so
those packages do not bring it in.
Details: microsoft/vstest#16654 and
dotnet/announcements#517
Using custom Docker images or Nano Server? See
microsoft/vstest#16652
## What's Changed
* Security fix for CVE-2026-105544, see
microsoft/vstest#16654
* Ship Microsoft.Diagnostics.NETCore.Client in Microsoft.TestPlatform
and Portable packages by @nohwnd in
microsoft/vstest#16632
* Fix localized resource paths in CLI source-build packages by @nohwnd
in microsoft/vstest#16643
**Full Changelog**:
microsoft/vstest@v18.11.0...v18.11.1
Commits viewable in [compare
view](microsoft/vstest@v18.11.0...v18.12.0).
</details>
Dependabot will resolve any conflicts with this PR as long as you don't
alter it yourself. You can also trigger a rebase manually by commenting
`@dependabot rebase`.
[//]: # (dependabot-automerge-start)
[//]: # (dependabot-automerge-end)
---
<details>
<summary>Dependabot commands and options</summary>
<br />
You can trigger Dependabot actions by commenting on this PR:
- `@dependabot rebase` will rebase this PR
- `@dependabot recreate` will recreate this PR, overwriting any edits
that have been made to it
- `@dependabot show <dependency name> ignore conditions` will show all
of the ignore conditions of the specified dependency
- `@dependabot ignore <dependency name> major version` will close this
group update PR and stop Dependabot creating any more for the specific
dependency's major version (unless you unignore this specific
dependency's major version or upgrade to it yourself)
- `@dependabot ignore <dependency name> minor version` will close this
group update PR and stop Dependabot creating any more for the specific
dependency's minor version (unless you unignore this specific
dependency's minor version or upgrade to it yourself)
- `@dependabot ignore <dependency name>` will close this group update PR
and stop Dependabot creating any more for the specific dependency
(unless you unignore this specific dependency or upgrade to it yourself)
- `@dependabot unignore <dependency name>` will remove all of the ignore
conditions of the specified dependency
- `@dependabot unignore <dependency name> <ignore condition>` will
remove the ignore condition of the specified dependency and ignore
conditions
</details>
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Javier Martinez (javier-intel)
changed the title
Description
Until now, an EP could report only one weightless mode, and apps could only turn weightless mode on or off. Each mode has different runtime requirements: ALL needs the source model at runtime, EXTERNAL_ONLY doesn't. With this change, EPs report every mode they support and apps choose the one that fits, without breaking EPs and apps built for 1.29 and 1.30.
Motivation and Context
Address the following shortcomings from PR 29607:
Changes
Compatibility
Testing