chore(deps): update dependency torch to v2.13.0 [security]

[Coldaine]

ℹ️ Note

This PR body was truncated due to platform limits.

Agent review expected: verify the dependency impact, summarize CI status, and merge only when the update is coherent for this repo.

This PR contains the following updates:

Package	Change	Age	Confidence
torch	2.9.1 → 2.13.0

---

PyTorch is vulnerable to memory corruption through its torch.lstm_cell function

CVE-2025-3001 / GHSA-qfhq-4f3w-5fph

More information

Details

A vulnerability classified as critical was found in PyTorch 2.6.0. This vulnerability affects the function torch.lstm_cell. The manipulation leads to memory corruption. The attack needs to be approached locally. The exploit has been disclosed to the public and may be used.

A patch is available through commit 999d94b.

Severity

• CVSS Score: 1.9 / 10 (Low)
• Vector String: CVSS:4.0/AV:L/AC:L/AT:N/PR:L/UI:N/VC:L/VI:L/VA:L/SC:N/SI:N/SA:N/E:P

References

• https://nvd.nist.gov/vuln/detail/CVE-2025-3001
• https://github.com/pytorch/pytorch/issues/149626
• https://github.com/pytorch/pytorch/issues/149626#issue-2935860995
• https://vuldb.com/?ctiid.302050
• https://vuldb.com/?id.302050
• https://vuldb.com/?submit.524212
• https://github.com/pytorch/pytorch/commit/999d94b5ede5f4ec111ba7dd144129e2c2725b03
• https://github.com/pypa/advisory-database/tree/main/vulns/torch/PYSEC-2025-195.yaml
• https://github.com/advisories/GHSA-qfhq-4f3w-5fph

This data is provided by the GitHub Advisory Database (CC-BY 4.0).

---

PyTorch is vulnerable to memory corruption through its torch.jit.script function

CVE-2025-3000 / GHSA-rrmf-rvhw-rf47

More information

Details

A vulnerability classified as critical has been found in PyTorch 2.6.0. This affects the function torch.jit.script. The manipulation leads to memory corruption. It is possible to launch the attack on the local host. The exploit has been disclosed to the public and may be used.

Severity

• CVSS Score: 1.9 / 10 (Low)
• Vector String: CVSS:4.0/AV:L/AC:L/AT:N/PR:L/UI:N/VC:L/VI:L/VA:L/SC:N/SI:N/SA:N/E:P

References

• https://nvd.nist.gov/vuln/detail/CVE-2025-3000
• https://github.com/pytorch/pytorch/issues/149623
• https://github.com/pytorch/pytorch/issues/149623#issue-2935703015
• https://vuldb.com/?ctiid.302049
• https://vuldb.com/?id.302049
• https://vuldb.com/?submit.524197
• https://github.com/pypa/advisory-database/tree/main/vulns/torch/PYSEC-2025-194.yaml
• https://github.com/pytorch/pytorch/commit/b90c94991cdf8b87c8f7439f79518e0ef2c4ca4f
• https://github.com/advisories/GHSA-rrmf-rvhw-rf47

This data is provided by the GitHub Advisory Database (CC-BY 4.0).

---

Release Notes

pytorch/pytorch (torch)

v2.13.0: PyTorch 2.13.0 Release

Compare Source

PyTorch 2.13.0 Release Notes

• Highlights
• Backwards Incompatible Changes
• Deprecations
• New Features
• Improvements
• Bug fixes
• Performance
• Documentation
• Developers

Highlights

FlexAttention lands on Apple Silicon (MPS), with up to ~12x speedup over SDPA on sparse patterns, and gains a deterministic backward path on CUDA for reproducible gradient computation.

CuTeDSL "Native DSL" backend gives Inductor a second high-performance code path (alongside Triton) for key GPU operations, with faster compilation. [Prototype]

nn.LinearCrossEntropyLoss combines the final prediction and loss computation to cut peak GPU memory by up to 4x for large-vocabulary language model training.

torchcomms, a new communications backend for PyTorch Distributed, improves fault tolerance, scalability, and debuggability for large-cluster training.

FSDP2 now overlaps reduce-scatter and all-gather communications via a dedicated process group (opt-in), increasing distributed training throughput.

Python 3.15 wheel support for PyTorch on Linux via the pytorch repository index, including builds compatible with free-threaded 3.15t.

Broader platform support: ROCm gains AOTriton 0.12b with native HIP CMake, Arm adds Armv9-A torch.compile targeting, and Intel XPU exposes new device telemetry APIs.

For more details about these highlighted features, you can look at the release blogpost. Below are the full release notes for this release.

Tracked Regressions

ROCm wheels break torch.compile on CPU in environments without a GPU

Running a torch==2.13.0+rocm7.2 wheel in an environment where no GPU is available (torch.cuda.is_available() is False) breaks torch.compile on the CPU path: the first compile raises RuntimeError: Can't detect vectorized ISA for CPU (#​189194). This is a regression from torch==2.12.1+rocm7.2, which compiles CPU code fine (detecting e.g. VecAVX2) in the same setup. The 2.13 ROCm wheel appears to rely on something present in the ROCm builder image to detect the CPU vectorized ISA, so it works when run on a ROCm image but fails on a plain CPU-only image.

Workaround: run the +rocm wheel on a ROCm image, or install a standard CPU/CUDA build for GPU-less environments.

Backwards Incompatible Changes

• Stop building CPython 3.13t (free-threaded) binaries (#​182951)

  Upstream pypa/manylinux removed CPython 3.13t (free-threaded) on 2026-05-07, because 3.13t
  was experimental and has been superseded by the now-non-experimental CPython 3.14t. As a result,
  PyTorch 2.13 no longer ships cp313t wheels (Linux, Triton, and related artifacts). Users on the
  free-threaded interpreter should move to Python 3.14t.

  PyTorch 2.12:

  # cp313t (free-threaded 3.13) wheels were available
  python3.13t -m pip install torch

  PyTorch 2.13:

  # Use free-threaded Python 3.14t instead
  python3.14t -m pip install torch

• Bare PyObject is no longer allowed in operator schemas (#​184209)

  Bare PyObject was accidentally accepted in operator schema strings in
  PyTorch 2.12. This was undocumented and is now rejected, since torch.compile
  does not support arbitrary PyObject inputs to custom ops. If
  you parse or register a schema with a bare PyObject argument or return type,
  you will now get a schema parse error.

  PyTorch 2.12:

  >>> from torch._C import parse_schema
  >>> parse_schema("foo(PyObject x) -> ()")  # accepted

  PyTorch 2.13:

  >>> from torch._C import parse_schema
  >>> parse_schema("foo(PyObject x) -> ()")  # raises a schema parse error

• Remove Bazel build support (#​180883)

  The Bazel build was never broadly adopted and still depended on the antiquated Bazel 6,
  while the wider ecosystem has since moved to Bazel 9. All Bazel build files and CI jobs have
  been removed. Users building PyTorch with Bazel should migrate to the supported CMake/pip install
  build flow.

  PyTorch 2.12:

  # Build PyTorch with Bazel
  bazel build //:torch

  PyTorch 2.13:

  # Bazel build files have been removed; build from source with pip instead
  pip install --no-build-isolation -e .

• StorageImpl's built-in copy-on-write (COW) materialization is replaced by a pluggable materializer hook (#​179063)

  StorageImpl no longer knows about COW directly. Its internal COW entry points
  StorageImpl::is_cow(), StorageImpl::maybe_materialize_cow(), and the friend
  cow::materialize_cow_storage() have been removed in favor of a single pluggable
  MaterializeFn hook (void(*)(StorageImpl*)) that a backend registers to run once,
  on the first mutable data-pointer access. COW is now just one consumer of this hook
  (c10::impl::cow::materialize_cow), and all COW behavior (lazy clone, refcounted
  shared data, copy-on-write) is unchanged. This also gives accelerator backends and
  eager-mode graph compilers a zero-fast-path-cost place to commit deferred allocations
  or materialize symbolic buffers on first mutation.

  This is a C++-only change. It affects out-of-tree backends/extensions that called the
  removed StorageImpl COW symbols directly; they will fail to compile against 2.13
  with errors such as no member named 'is_cow' in 'c10::StorageImpl'. Migrate to the
  new hook API (set_materializer() / has_materializer() / clear_materializer()).

  PyTorch 2.12:

  // Detect a COW storage and force it to materialize.
  if (storage.is_cow()) {
    storage.maybe_materialize_cow();
  }

  PyTorch 2.13:

  // Register a one-shot materializer; it runs on the next mutable-data access
  // and then clears itself. COW registers c10::impl::cow::materialize_cow this way.
  storage.set_materializer(&my_backend_materialize);  // void(StorageImpl*)
  
  // `has_materializer()` replaces `is_cow()` for "is a deferred materialization pending?"
  if (storage.has_materializer()) { /* ... */ }

• Convert shared_ptr<Node> to intrusive_ptr<Node> (#​181139). This changes the signature of Tensor::grad_fn. Accesses to Tensor.grad_fn() should change from std::shared_ptr<Node> to c10::intrusive_ptr<Node>. Similarly, construction of a C++ autograd function should change:

  PyTorch 2.12:

  std::shared_ptr<CustomCppNode> node(new CustomCppNode(), torch::autograd::deleteNode);

  PyTorch 2.13:

  auto node = c10::make_intrusive<CustomCppNode>();

• The minimum supported NCCL version when building from source is now 2.23 (#​186292)

  PyTorch now requires NCCL >= 2.23 at compile time, and the preprocessor/runtime gates that guarded NCCL features introduced in 2.23 or earlier have been removed. Users who build PyTorch from source against a system NCCL older than 2.23 will hit compile errors against the dropped gates. Upgrade the NCCL installation to >= 2.23 to build. The prebuilt PyTorch wheels already bundle a compatible NCCL, so pip/conda users are unaffected.

• Remove named tensors (#​173895)

  The named tensor feature (a long-deprecated prototype) has been fully removed to reduce overhead and code bloat. All associated Python and C++ APIs are gone, including Tensor.names, Tensor.rename(), Tensor.refine_names(), Tensor.align_to(), Tensor.align_as(), torch.align_tensors(), the names= keyword on factory functions (e.g. torch.zeros, torch.empty, torch.ones), and the C++ Dimname / DimnameList APIs. Code that previously relied on named dimensions must track dimension order positionally and avoid usage of any of these now-removed APIs or op overloads.

• The onednn::qconv2d_pointwise.binary and .binary_tensor operators no longer alias their input but rather return fresh tensors. Previously these ops mutated the qaccum input buffer and returned it directly, violating the PyTorch invariant that custom operator outputs must not alias inputs. This silently bypassed aliasing checks via the old -> Tensor(a!) schema and would become a hard error in a future PyTorch version (as mentioned in #​182063), so the schema and implementation were corrected to return a fresh output. Most users are unaffected, only code that calls these ops directly and relies on the in-place mutation of qaccum must now read the returned tensor instead. (#​177171)

Deprecations

• Custom operators that return an output aliasing one of their inputs are deprecated (#​182063)

  When a custom operator returns an output that is the same tensor as (or otherwise aliases) one of its inputs under torch.compile, PyTorch now emits a UserWarning stating that this is deprecated and will become an error in a future version of PyTorch. Previously the warning stated the change would land in PyTorch 2.12; that timeline has been pushed back. To update your code, return a clone of the offending output instead of the input, or refactor the operator so it does not return the aliased tensor.

  Deprecated:

  @torch.library.custom_op("mylib::foo", mutates_args=())
  def foo(x: torch.Tensor) -> torch.Tensor:
      return x  # output aliases the input -- deprecated

  Updated:

  @torch.library.custom_op("mylib::foo", mutates_args=())
  def foo(x: torch.Tensor) -> torch.Tensor:
      return x.clone()  # return a clone instead

• Creating tensors with the quantized dtypes quint8, qint8, and qint32 is now deprecated and emits a warning. This covers both Python and C++ call sites; see #​184982 for migration guidance (#​184984)

  PyTorch 2.12:

  >>> x = torch.quantize_per_tensor(torch.randn(3), 0.1, 0, torch.quint8)

  PyTorch 2.13:

  >>> x = torch.quantize_per_tensor(torch.randn(3), 0.1, 0, torch.quint8)
  UserWarning: Creating tensors with quantized dtypes (quint8, qint8, qint32) is deprecated

• Rename distributed collective ops to the _single naming scheme and deprecate the old names (#​186123, #​186124, #​186125, #​186134, #​186135, #​186144)

  To align the public torch.distributed collective APIs with the naming used by torchcomms' TorchCommBackend, all_gather_into_tensor is renamed to all_gather_single and reduce_scatter_tensor to reduce_scatter_single. The previous names continue to work as thin wrappers that delegate to the new functions, but now emit a FutureWarning.

  PyTorch 2.12:

  dist.all_gather_into_tensor(output, input)
  dist.reduce_scatter_tensor(output, input)

  PyTorch 2.13:

  dist.all_gather_single(output, input)
  dist.reduce_scatter_single(output, input)

New Features

Python Frontend

• Add two new operator tags, torch.Tag.inplace and torch.Tag.out, that let an operator declare how it writes its result: inplace means it mutates a tensor in place, and out means it writes into a caller-provided output tensor. Native PyTorch operators are tagged automatically, and custom operators defined with torch.library can opt in by adding the tag. To be tagged inplace, an operator must take the tensor it mutates as its first positional argument (declared as Tensor(a!), and the only mutable argument) and return that same tensor. Tagging a custom operator this way improves its behavior under torch.compile: inplace ops now go through auto_functionalize, so the reinplacing pass can analyze clones and skip unnecessary copies, and both inplace and out ops get their fake/meta kernels generated for free. See the Python custom operators tutorial for how to author and tag custom operators. (#​181100, #​181099, #​184199, #​184200, #​184201, #​184202, #​184203, #​180851, #​180852)
• Add const_data_ptr() Python binding to torch.Tensor for read-only data pointer access (#​180382)
• Add an abbr property to torch.dtype that returns a dtype's short string abbreviation (e.g. torch.float32.abbr returns "f32") (#​177296)
• Allow positional arguments to be passed as keyword arguments to autograd custom Functions (#​182206)
• Expose rearrange in the torch.func namespace for einops-style tensor reshaping (#​173183)

torch.nn

• Add nn.LinearCrossEntropyLoss, a fused linear-projection plus cross-entropy loss module that avoids materializing the full logits tensor (#​181573, #​185852, #​172286, #​186113)

Autograd

• Add torch.autograd.graph.region_activation_memory_budget (#​185979)
• Support passing gradient inputs as a dict to torch.autograd.grad and torch.autograd.backward (#​178140)

Distributed

• Add a registration API for symmetric memory arguments (lib.register_symm_mem_args()), letting operators (including out-of-tree ops) declare which arguments require symmetric-memory allocation (#​173513)

• Remove NCCLSymmetricMemory's explicit dependency on ProcessGroupNCCL, enabling symmetric memory to work with out-of-tree backends such as torchcomms (#​184260)

• Support accessing the ReduceOp.PREMUL_SUM factor from Python when implementing process group backends in Python (#​185863)

• Expose the NCCL 2.30 maxP2pPeers config binding (#​181686)

• Add rocSHMEM Triton integration for symmetric memory on ROCm (#​178658)

• Support passing extra keyword arguments to the loss function in pipeline schedules via a new loss_kwargs parameter to step(), enabling loss functions that require arguments beyond (output, target) (such as chunked cross-entropy needing token counts for scaling) (#​181057)

Distributed FSDP2

• Add FSDPModule.set_separate_reduce_scatter_group to give reduce-scatter its own NCCL communicator, enabling opt-in overlap of all-gather and reduce-scatter (#​186335)
• Add set_reduce_scatter_max_input_buffers to keep multiple reduce-scatter input buffers in flight, so backward compute no longer stalls waiting to recycle a single reduce-scatter buffer (#​186000)

Profiler

• Profiler/Kineto now emits channel metadata on CUDA backends (#​185968)

Dynamo

• Add torch.compiler.set_default_backend to override the default torch.compile backend globally, so out-of-tree backend authors don't need to pass backend= at every call site (following the pattern of torch.set_default_dtype/torch.set_default_device). Explicit backend= arguments still take precedence (#​178944)
• Add torch.compile(f, isolate_recompiles=True) to give each torch.compile call its own isolated cache bucket, preventing cross-compile interference in cache lookups and recompile-limit checks when multiple torch.compile calls target the same function (#​178351)
• Add register_multi_grad_hook support to @leaf_function, allowing a backward hook to fire once per backward pass when all requires_grad inputs have their gradients computed (#​179609)

Inductor

• Add flash-decoding support to the CPU FlexAttention template (chosen when query length is 1) with a new configurable PARTITION_SIZE kernel option (#​159835)
• Add Triton convolution backward kernels (input and weight gradients) as an autotuning backend in place of the ATen-only fallback (#​178945)
• Add an Inductor FX pass (decomp_comms) that eliminates all_gather for Gram-matrix optimizer patterns (Muon/Shampoo) under FSDP, gated by config.aten_distributed_optimizations.allow_comms_decompositions, yielding 1.25-1.95x training speedups (#​184370)

Ahead-Of-Time Inductor (AOTI)

• Generated C shims for the AOTI stable ABI are now versioned and gated by TORCH_TARGET_VERSION, so shims introduced in newer releases are only exposed when the target version supports them (#​181916)
• Triton CPU AOTI models now work end-to-end through the public torch._inductor.aoti_compile_and_package / aoti_load_package API, including packaging and loading of the multiple .so files emitted per kernel (#​182251)
• Added stable C shim functions (torch_exception_get_what, torch_exception_get_what_without_backtrace, and STABLE_TORCH_ERROR_CODE_CHECK) so extensions built against the stable ABI can retrieve the original error message across the C API boundary (target version 2.13+) (#​180135)
• Added a stable AOTI stream shim aoti_torch_stream_native_handle and torch::stable::accelerator::Stream::nativeHandle(), gated behind TORCH_FEATURE_VERSION >= 2.13, for retrieving a native stream handle from the stable ABI (#​183930)

Release Engineering

• Add Python 3.15 wheel builds across Linux (CPU/CUDA), Triton, ROCm, and XPU (#​182954, #​184600, #​185409, #​184829, #​184891, #​184906, #​185094)

CUDA

• Add CUDAGraph.get_graph_data() for graph topology introspection (#​183165)
• Lightweight API to get private pool reserved memory bytes (#​178240)

MPS

• Add FlexAttention support for MPS (#​182552, #​186215)
• Add support for torch.distributions.Dirichlet on MPS by adding _sample_dirichlet and _dirichlet_grad Metal implementations (#​185458, #​185854)
• Add grid_sampler_2d backward support on MPS (#​179756)
• Add grid_sampler_3d backward support on MPS (#​179388)
• Add lcm support on MPS via a new Metal kernel (#​186279)
• Add complex support to c10/metal/reduction_utils.h (#​180708) and a complex->bool specialization (#​185938)

ROCm

• Enable external events in CUDA graphs (#​178264)
• Enable GPU Address Sanitizer build (#​183792, #​176461)
• Improve Inductor GEMM search space performance using the Origami project (#​172512)
• Use CMake native HIP language support, enable_language(HIP) (#​180485)
• New Inductor benchmarker based on Torch Profiler (#​175097)

XPU

• Add XPU device telemetry APIs for temperature, frequency, power draw, engine utilization, memory bandwidth usage, and used device memory through torch.xpu.* (#​181082, #​183427, #​183428, #​183429, #​183430, #​183431)
• Add FP8 blockwise scaling support for scaled_mm on XPU (#​173630, #​176043)

Improvements

Python Frontend

• Make it possible to load safetensors with torch.load (#​170592)
• Make Storage.pin_memory / Storage.is_pinned device-agnostic (#​186223)
• Add op_overloads to OpOverloadPacket to enumerate an operator's overloads (#​182993)

torch.nn

• Expose num_splits in FlashAttention-2 and bump the flash-attention submodule (#​179760)
• Support linear_bias in linear_cross_entropy on the reference and chunked paths (#​185129, #​185276)

Optimizer

• Fix SequentialLR wrong learning rate initialization when milestones contain 0 (#​185986)

Autograd

• Implement autograd derivatives for torch.nextafter (#​148820)
• Add torch.autograd.enforce_grad_layout_policy to control the memory layout policy for accumulated gradients (#​180552)

Distributed

• When TorchComms is enabled, route new_group through split_group for subgroup creation, raising NotImplementedError for arguments split_group cannot honor (e.g. use_local_synchronization=True, sort_ranks=False) instead of silently falling back (#​185416)

• Delegate dist.new_group to custom process group subclasses (#​184262)

• Surface started-work metadata in NCCL watchdog timeouts (#​183656)

• Add a health check endpoint to the distributed debug server (#​179326)

• Make the DeviceMesh non-overlapping check stricter (#​172343)

• Allow elastic_launch/launch_agent to accept a pre-created torchelastic health check server, so it can be started before rendezvous (#​180543)

• Add an overlap_pp_comm flag to pipeline schedules (default True) that, when set to False, defers each pipeline RECV op to immediately before the compute op that consumes it, using rank-parity P2P ordering to avoid deadlock (helps platforms such as AMD ROCm where a pending RECV blocks unrelated compute) (#​178815)

DTensor

• Migrate embedding and random ops to single-dim sharding strategies and increase op coverage (#​180281, #​180503)
• Add auto-infrastructure that derives single-dim sharding strategies for autogenerated op variants (.out, inplace, functional, and foreach), expanding strategy coverage to hundreds of additional ops (#​185386)
• Register sharding strategies for additional ops: scatter, upsample/interpolation backward, anti-aliased upsample, batch norm backward, and aten.detach_.default (#​186149, #​180311, #​184626, #​182743, #​181876)

Distributed FSDP2

• Support forward-mode automatic differentiation (torch.func.jvp) on models wrapped with fully_shard or replicate, including with mixed precision (#​182732)

Linear Algebra Frontend

• Add Half and BFloat16 dispatch support for torch.trace on CPU (#​184874)
• Improve heuristics for the cuSOLVER vs cuBLAS backend switch in torch.linalg.lu (#​185344)

Profiler

• The memory viz tool now more accurately represents GPU footprint when impacted by fragmentation (#​180515)
• The memory viz tool now aggregates stripes per-pool to improve visualization for large snapshots (#​180613)
• Profiler now also exposes CUDA occupancy metadata as a nested dictionary in the .events() output (#​180275)

FX

• split_module now supports torch.Size crossing graph split boundaries by decomposing size() calls into per-dimension sym_size nodes, and builds submodules lazily for faster inference graph splitting (#​179839)

• CapabilityBasedPartitioner can now opt out of horizontal fusion via skip_horizontal_fusion=True, partitioning only through direct data dependencies (#​184904)

• Enable rewriting of FX traces containing complex tensors during compilation (#​169832)

Dynamo

• Implement additional Python operators in Dynamo: bitwise and (#​184788), bitwise xor (#​184789), left/right shift (#​183462), floor division (#​185652), true division (#​185653), remainder (#​185654), and divmod (#​185655)
• Support tracing more constructs in Dynamo: einops 0.8.2 (#​185619), record_function as a decorator (#​184703), inference_mode retracing helpers (#​185066), mark_dirty in the autograd Function HOP (#​184267), warn_only deterministic toggles (#​180373), and the _maybe_view_chunk_cat functional collective (#​180389)
• Support item assignment and deletion (__setitem__/__delitem__) on more container types in Dynamo via sq_ass_item/mp_ass_subscript slots (#​182862, #​182996)
• Support torch.accelerator.device_index and torch.xpu.device in the device context manager (#​181846, #​181847)
• Improve Triton support under torch.compile: accept tl.constexpr values as kernel arguments (#​181783) and handle capture_triton as a no-op during tracing (#​183555)
• Improve dynamic shape specification: reduce verbosity in shape specs for the common case (#​184271), add SeqSpec for list/tuple specs with better walk-spec errors (#​185327), add ObjectSpec (#​182764), pipe dynamic spec through torch.compile (#​184501), and revisit guarding in mark_dynamic APIs (#​181469)
• Improve torch.compile device mismatch errors with a dedicated FakeTensorDeviceMismatchError and actionable guidance to place inputs, parameters, and buffers on the same device (#​185412)
• Improve error messages and diagnostics: clearer data-dependent errors for .any()/.all() (#​180406), clearer torch._check tensor predicate errors (#​185777), user-friendly reasons for skipped frames (#​183596), carets in stack traces (#​182393), and reporting why a symbol was created dynamically in symbolic_shapes logs (#​168331)
• Make Dynamo exceptions pickleable (#​185725)
• Inline decomposed quantization helpers in Dynamo (#​185628)
• Make Dynamo debug/repro utilities device-agnostic (#​184851)

Inductor

• Support pin_memory for torch.ones, torch.zeros and torch.full (#​174595)
• Add a ROCm config flag to disable the pointer_range_32 optimization (#​179604)
• Add a peak memory threshold config for combo kernels (#​180578)
• Enable autotuning and a fast compensation path for CPU static/dynamic smooth-quant qlinear, with correct handling of 0D x_scale/x_zp and ReinterpretView strides (#​181090)
• Add aot_inductor.autotune_per_kernel_alloc config to allocate-run-delete tensors per kernel during AOTI autotuning, avoiding OOM on large models (#​181176)
• Bound AsyncCompile future waits with the compile_worker_wait_timeout setting (#​181293)
• Add a100_default_flex_config entries for head_dim=192 (#​181835)
• Add an explicit lowering (fallback) for aten.multinomial to avoid graph breaks (#​182423)
• Add an Inductor lowering for _scaled_mm_v2 (#​182527)
• Enable combo_kernel_autotune_grouping by default (#​182567)
• Add cudagraph_partition_memory_budget config for partition reordering (#​183569)
• Add ATen fallbacks for bincount, unique variants, and AMP scale ops so they compile without graph breaks (#​183590)
• Unfuse the bias add from addmm when the bias is a narrowing dtype cast (fp32->bf16/fp16) to preserve bias precision in XPU AMP training (#​183680)
• Emit a clearer diagnostic when a backward CUDAGraph output installed as a .grad buffer is invalidated on a later run (gradient accumulation) (#​184003)
• Support split online softmax reductions in Inductor (#​184069)
• Support fusing index_add-style atomic scatter mutations into Triton template epilogues behind a config flag (#​184179)
• Add a cpp.march Inductor config knob so AOTInductor cpp-only builds can override or suppress the default CPU architecture flag (#​184297)
• Improve the Triton cache directory guidance when loading shared objects from a noexec filesystem fails (#​184362)
• Enable the Bert SDPA pattern rewrite on CUDA while keeping the original matmul/softmax math path (#​184417)
• Improve the Triton launcher argument-mismatch error with a clearer message and a cached preflight check (#​184522)
• Add broader CuteDSL op overrides (rsqrt, exp2, log2, log10, tan, acos, asin, atan, atan2, floor, logical_xor) (#​184538)
• Add an optional fake_mode argument to standalone_compile (with dynamic_shapes="from_example_inputs") so it can reuse the caller's FakeTensorMode/ShapeEnv instead of always creating a fresh one (#​184776)
• Support signbit on unsigned integer dtypes (#​185985)
• Add a keep_static_cubin_raw config to retain cubin bytes in cached kernels so caches restored on another machine avoid recompilation (#​186404)
• Extend BatchLinearLHSFusion's matcher to also match the inlined torch._C._nn.linear form so the (opt-in) fusion can fire on Dynamo-inlined linear (#​186632)
• Add a clearer error message with install instructions when a compatible Flash Attention package is unavailable for flex attention (#​186827)
• Add another anchor node to the batch-linear fusion pass so more small torch._C._nn.linear operations are grouped into a single batched kernel (#​180477)
• Extend the batch fusion pass to support detach() method calls (#​180513)
• Add a deterministic backward for the FlexAttention flash kernel (#​174813)
• Use rand4x for Inductor Triton random number generation (#​184377)
• Add a Quack-based CuTeDSL RMSNorm kernel (#​182108)

Ahead-Of-Time Inductor (AOTI)

• Use fatbinary for multi-arch CUDA kernels (#​184456)
• Support mixed-device constants in update_constant_buffer (#​181114)
• Add FP8 header files in the AOTI shim.h (#​178120)
• Add throttled cudaMemcpy for AOTI constant loading to reduce peak memory usage (#​184823)
• Preserve AOTI proxy_executor error messages (#​180884)
• Enable Triton kernels in AOTI C++ wrapper on CPU (#​181068)
• Skip CPU vec ISA setup for device-only cpp_wrapper (#​182089)
• Expose torchbind constants from AOTIModelPackageLoader (#​182149)
• Improve AOTI error for Python custom ops (#​186305)

Export

• Support serialization of opaque type constants in torch.export save/load (#​181676)
• Make functorch JVP operator torch.export-able (#​179686)
• Add UpdateConstantBufferFromCpu for host-to-device copy (#​181637)

AOTAutograd

• Support CPU activation offloading in the rematerialization pass, including marking recomputed nodes for backward and adding a resize-to-0 deallocation op so offloaded tensors are freed after their host-to-device copy (#​181937, #​181938)
• Use FX node names in merge_view_inputs error messages, so non-differentiable view input mutation errors identify the specific offending inputs (#​180424)

Composability

• Add fake tensor support for _transformer_encoder_layer_fwd so it traces under torch.compile (#​183916)
• Enable Armv9-A target support for torch.compile on AArch64 (#​184555)
• Functionalize in-place c10d collectives in standalone compile (#​181836)

Foreach

• Fix/add empty check for _foreach_max (#​173483)

ONNX

• Add adaptive_max_pool2d and adaptive_max_pool3d decompositions for ONNX export (#​184396)

C++ Frontend

• Add stable ABI for set_python_module on torch::Library (#​182720)
• Add == overloads for HeaderOnlyArrayRef (#​185379)
• Add torch::stable::Generator (#​186423)
• Add c10::layout typecaster for torch.layout (#​179607)
• Add default-args support to def_static (#​175644)
• Add support for controlling scientific notation in C++-side tensor printing (#​173321)

Release Engineering

• Move the NCCL pin to 2.30 (#​181313)
• Advance the Triton pin to 3.7.1 (#​181001, #​186792)
• Upgrade the XPU support package to 2026.0 (#​182003)
• Add a configurable threshold to avoid power-of-two rounding for large pinned memory allocations (#​171662)
• Move some pre-build steps from setup.py to CMake (#​177641)

CUDA

• Debugging tool to verify that external inputs to a CUDA graph are alive before replay (#​174649)
• Add get/set/reset functions for BLAS workspace sizes (#​177912)
• Cleanup double import in BinaryDivFloorKernel.cu (#​179260)
• Return supported CUDA arch list when no GPU is present but GPU is compiled (#​180356)
• Detect and fix stale stream references in autograd during CUDA graph capture (#​180090)
• Use opmath_t in i1 and i1e CUDA kernels (#​183778)
• Support resize_ with address hint (#​178215)
• Support bfloat16 in _embedding_bag_per_sample_weights_backward on CUDA (#​185889)
• Align parsePerProcessMemoryFraction's return type with other parsers (#​185139)
• Improve error message when cuda-bindings version is too old (#​185990)
• Expose torch.cuda.current_solver_handle for cuSOLVER handle sharing (#​176705)

cuDNN

• Add flag to select depthwise convolution backend (#​176500)
• Upgrade cudnn_frontend submodule to 1.24 (#​185554)

MPS

• Migrate many ops from MPSGraph to native Metal kernels for bernoulli (#​182210), native_dropout (#​182232), uniform/normal/randint (#​182386), randperm (#​182528), comparison ops eq/ne/lt/le/gt/ge (#​183019), bitwise ops (#​182839), scatter/gather (#​184028), copy-cast (#​184740), gelu/gelu_backward (#​181451), replication pad (#​183065), embedding backward (#​185119), trace (#​183627), count_nonzero (#​180725), amax/amin/aminmax/all/any (#​180752), and cumsum/cumprod (#​185609), native_group_norm (#​183830, native_group_norm_backward #​184437), topk and kthvalue Metal kernels (#​184106), single-block and multi-block sort, including a stable sort path (#​180714, #​182242, #​181736)
• Support integer inputs to histc (#​178624)
• Support out variants of unary ops (#​184743)
• Improvements to SDPA Metal kernels: prefill attention kernels (#​181575), GQA support (#​183280), is_causal support (#​181855), head dim 256 (#​181852), float mask support (#​183458), and a clearer error for causal + attn mask (#​181856)
• Add an ILP variant for binary tensor iterators and replace unary VEC4 with a generic ILP-per-thread dense kernel (#​182155, #​181509, #​183055)
• Make several Metal kernels stride-aware to avoid unnecessary .contiguous() calls: Col2Im (#​181949), Im2Col (#​182709), Repeat (#​182718), LossOps (#​182714), and HistogramKernel (#​181951)
• Enable NDHWC+DHWIO fast path for Conv3d on channels_last_3d (#​184612)
• Clear the MPSGraph cache in torch.mps.empty_cache() (#​181485)
• Add a private API to get the host alias of Metal storage (#​180961)
• Add input validation: stride > 0 in pool ops (#​184875), F.fold (#​182067), im2col (#​183593), and bernoulli probabilities (#​185065)
• Alert on non-deterministic algorithms on MPS (#​185061)
• Raise a clear not-implemented error for dropout_p (#​184126)
• Use proper precise log functions in unary kernels (#​185381)

ROCm

• Additional cub::DeviceHistogram hipify mappings (#​180433)
• SDPA improvements via AOTriton 0.12b: head_dim != head_dim_v, use_deterministic_algorithms, gfx1100 and gfx1151 promoted out of experimental, partial FAv3 support on gfx950 (#​184288)

XPU

• Add last_level_cache_size and is_integrated_gpu to XPU device properties ([#​184499

✂ Note

PR body was truncated to here.

---

Configuration

📅 Schedule: (in timezone America/Chicago)

• Branch creation
  • At any time (no schedule defined)
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

This PR has been generated by Mend Renovate CLI.

[Copilot]

Copilot wasn't able to review any files in this pull request.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[codeant-ai]

🤖 CodeAnt AI — Review Status

Status	Commit	Started (UTC)	Finished (UTC)
✅ Incremental review completed	0a00f4e	Sep 03, 2026 · 14:10	14:10
✅ Incremental review completed	53f0a3a	Sep 02, 2026 · 14:07	14:07
✅ Incremental review completed	0605942	Aug 28, 2026 · 21:12	21:12
✅ Incremental review completed	07ab784	Aug 27, 2026 · 20:14	20:14
✅ Incremental review completed	665d912	Aug 24, 2026 · 10:39	10:40

[codeant-ai]

---

Thanks for using CodeAnt! 🎉

We're free for open-source projects. if you're enjoying it, help us grow by sharing.

Share on X ·
Reddit ·
LinkedIn

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_7d0f2978-1016-4c13-a48d-1d6e9d92c7d5)

[qodo-code-review]

PR Summary by Qodo

Update torch to 2.13.0 (security) and refresh CUDA/triton lock deps

⚙️ Configuration changes 🕐 20-40 Minutes

AI Description

• Bump PyTorch to 2.13.0 to address reported torch.lstm_cell memory-corruption advisory.
• Refresh lockfile with new Linux CUDA 13 runtime dependencies and updated triton wheels.
• Normalize a few transitive dependency markers as part of lock re-resolution.

Diagram

graph TD
  A(["Repo / Python env"]) --> B["torch 2.13.0"] --> E["triton 3.7.1 (Linux)"]
  B --> C["cuda-toolkit 13.x (Linux)"] --> F["NVIDIA CUDA libs"]
  B --> D["cuda-bindings 13.3.1"] --> G["cuda-pathfinder 1.6.0"]

Loading

High-Level Assessment

The following are alternative approaches to this PR:

1. Upgrade to the minimum patched torch release

• ➕ Smaller behavioral jump vs 2.13.0 (lower risk of API/ABI/runtime changes).
• ➕ May avoid the CUDA 13 dependency stack shift if earlier releases still use cu12-era deps.
• ➖ May not be available/maintained depending on the advisory’s fixed versions.
• ➖ Renovate/security policies may prefer latest available stable release.

2. Pin to a specific CUDA line (e.g., cu12) via an alternate index/source

• ➕ Keeps compatibility with existing CUDA drivers/tooling if the repo’s runtime expects CUDA 12.
• ➕ Can reduce surprise transitive dependency changes in the lockfile.
• ➖ Requires configuring package source/index and potentially platform-specific constraints.
• ➖ More maintenance burden than using the default PyPI resolution.

3. Use CPU-only torch where GPUs are not required

• ➕ Avoids large CUDA binary dependency chain and related runtime failures.
• ➕ Simplifies CI and improves portability.
• ➖ Not viable if the project requires GPU acceleration.
• ➖ May require code/config to ensure the CPU build is selected consistently.

Recommendation: Given this is a security-driven Renovate update, the PR’s approach (move to 2.13.0) is reasonable, but the CUDA stack shift is the key risk. If this repo runs on GPU-enabled Linux, validate that target environments have driver/toolkit compatibility for CUDA 13-era wheels and that CI includes at least an import/smoke test. If the runtime is pinned to CUDA 12, consider the alternate-index/pin approach to stay on cu12 while still taking the security fix.

Files changed (1) +173 / -88

Other (1) +173 / -88

uv.lockBump torch to 2.13.0 and refresh CUDA/triton transitive pins +173/-88

Bump torch to 2.13.0 and refresh CUDA/triton transitive pins

• Updates torch from 2.9.1 to 2.13.0 and re-resolves the lockfile to include new Linux-only CUDA packaging (cuda-bindings/cuda-toolkit) and CUDA 13-era nvidia-* components (e.g., cudnn-cu13, nccl-cu13). Also updates triton from 3.5.1 to 3.7.1 and adjusts a few dependency markers (e.g., typing-extensions and numpy markers) as part of the lock resolution.

uv.lock

[qodo-code-review]

Code Review by Qodo

🐞 Bugs (1) 📘 Rule violations (0) 📜 Skill insights (0)

1. CUDA version docs drift 🐞 Bug ⚙ Maintainability

Description

uv.lock now resolves torch’s Linux CUDA stack to CUDA 13-era packages (e.g., nvidia-*-cu13 +
cuda-toolkit 13.x), while repo documentation still describes CUDA 12.2 as the target; this PR
introduces that mismatch and can mislead future GPU setup/debugging.

Code

uv.lock[R1241-1244]

+    { name = "nvidia-cudnn-cu13", marker = "sys_platform == 'linux'" },
+    { name = "nvidia-cusparselt-cu13", marker = "sys_platform == 'linux'" },
+    { name = "nvidia-nccl-cu13", marker = "sys_platform == 'linux'" },
+    { name = "nvidia-nvshmem-cu13", marker = "sys_platform == 'linux'" },

Evidence

The lockfile shows torch now depends on Linux-only CUDA 13-era packages (e.g., nvidia-cudnn-cu13)
and the repo docs still state CUDA 12.2 as the target; this PR changes the lock, thereby introducing
the mismatch.

uv.lock[1229-1248]
uv.lock[193-199]
docs/research/dependency-audit-report-2026-03-24.md[7-21]
pyproject.toml[6-11]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`uv.lock` now pins PyTorch’s Linux CUDA dependency set to CUDA 13-era packages, but documentation still states CUDA 12.2 is the target version. This creates an inconsistency between the committed dependency graph and the repo’s documented CUDA expectations.

## Issue Context
- The PR updates `torch` from 2.9.1 to 2.13.0.
- The updated lock introduces CUDA 13-namespaced packages (e.g., `nvidia-cudnn-cu13`) and `cuda-toolkit 13.0.3.0`.
- Documentation in `docs/research/dependency-audit-report-2026-03-24.md` still references CUDA 12.2 as the target.

## Fix Focus Areas
- docs/research/dependency-audit-report-2026-03-24.md[7-21]
- uv.lock[1229-1248]
- uv.lock[194-199]

## Suggested fix
Choose one (based on the intended supported CUDA baseline):
1) **If CUDA 13 is acceptable/desired for Linux PyTorch**: update the referenced docs to reflect CUDA 13 for the PyTorch/Linux path (and clearly separate it from any Windows/system-toolkit guidance).
2) **If CUDA 12.x remains required**: constrain/select a CUDA-12-compatible PyTorch distribution (e.g., pin torch/triton to versions that resolve to cu12, or document/use a specific index/build) and re-lock so `uv.lock` matches that policy.

Also add a short validation note in docs (or a script) describing how to confirm the resolved torch CUDA version after `uv sync` (e.g., printing `torch.version.cuda`) so future upgrades are less ambiguous.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

Context sources

✅ Compliance rules (platform): 28 rules

Review mode: ⚖️ Balanced: This security-motivated torch upgrade changes the runtime dependency graph substantially, including CUDA, NVIDIA, and Triton packages, so it carries meaningful compatibility and behavioral risk despite touching only a lockfile.

Tip of the day

💡 Did you know, you can reply 'qodo' on any finding to push back, ask questions, or dig deeper

More tips ↗ | Customize Qodo ↗ | Qodo docs ↗

[qodo-code-review]

1. Cuda version docs drift 🐞 Bug ⚙ Maintainability

uv.lock now resolves torch’s Linux CUDA stack to CUDA 13-era packages (e.g., nvidia-*-cu13 +
cuda-toolkit 13.x), while repo documentation still describes CUDA 12.2 as the target; this PR
introduces that mismatch and can mislead future GPU setup/debugging.

Agent Prompt

## Issue description
`uv.lock` now pins PyTorch’s Linux CUDA dependency set to CUDA 13-era packages, but documentation still states CUDA 12.2 is the target version. This creates an inconsistency between the committed dependency graph and the repo’s documented CUDA expectations.

## Issue Context
- The PR updates `torch` from 2.9.1 to 2.13.0.
- The updated lock introduces CUDA 13-namespaced packages (e.g., `nvidia-cudnn-cu13`) and `cuda-toolkit 13.0.3.0`.
- Documentation in `docs/research/dependency-audit-report-2026-03-24.md` still references CUDA 12.2 as the target.

## Fix Focus Areas
- docs/research/dependency-audit-report-2026-03-24.md[7-21]
- uv.lock[1229-1248]
- uv.lock[194-199]

## Suggested fix
Choose one (based on the intended supported CUDA baseline):
1) **If CUDA 13 is acceptable/desired for Linux PyTorch**: update the referenced docs to reflect CUDA 13 for the PyTorch/Linux path (and clearly separate it from any Windows/system-toolkit guidance).
2) **If CUDA 12.x remains required**: constrain/select a CUDA-12-compatible PyTorch distribution (e.g., pin torch/triton to versions that resolve to cu12, or document/use a specific index/build) and re-lock so `uv.lock` matches that policy.

Also add a short validation note in docs (or a script) describing how to confirm the resolved torch CUDA version after `uv sync` (e.g., printing `torch.version.cuda`) so future upgrades are less ambiguous.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

[qodo-code-review]

Qodo Fixer

No findings are within the configured fix scope. To change which findings are fixed, adjust the setting on your Qodo configuration page.

[kilo-code-bot]

WARNING: macOS minimum version increased from 11 to 14

torch 2.13.0 drops support for macOS 11 (Big Sur) through macOS 13 (Ventura); wheels now require macOS 14 (Sonoma) or later. Any macOS-based contributors or CI runners on older OS versions will fail to install this dependency.

---

Reply with @kilocode-bot fix it to have Kilo Code address this issue.

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Files Reviewed (1 file)

• uv.lock

Incremental Review Notes

Model: kilo/poolside/laguna-s-2.1:free (read-only, non-interactive)

The incremental diff since commit 53f0a3a contains a single change to uv.lock: cuda-pathfinder 1.8.0 → 1.8.1 (lock-file hash/URL/size bump). No new issues were identified. The prior review at this commit already concluded the torch 2.14.0 uplift, CUDA 13 package renames, and macOS 11→14 wheel minimum are acceptable for this security-driven lock refresh.

Pre-existing external findings (not duplicated, not on changed lines):

• Other review bots flagged the macOS wheel minimum rising 11→14 (lines ~1216–1244, outside this incremental hunk) and a title/lock discrepancy (2.13.0 in PR title vs 2.14.0 in the resolved lock). Neither is on the incremental cuda-pathfinder change, so they are surfaced here for human triage only.

Documentation Observations (advisory)

• docs/northstar.md is marked freshness: stale, last_reviewed: 2026-02-09, review_due: 2026-05-10 — overdue by ~4 months. Its goals (CUDA-first STT, Moonshine fallback, streaming partials) should be reconciled against current code/Parakeet plans.
• docs/plans/cleanup-plan.md describes a 6-phase doc cleanup. Much of it appears executed (CLAUDE.md/GEMINI.md deleted, docs archived, current-status.md active), yet the plan itself remains status: active — it should be archived or marked complete.
• Stale references to the non-existent windows-multi-agent-recovery.md linger in docs/reviews/portable_standard_critique.md and in cleanup-plan.md itself (Phase 3 cleanup was not fully closed).
• AGENTS.md is a 17-line pointer, contradicting cleanup-plan Phase 5 which targets a full-onboarding AGENTS.md. The full version lives only in .kilocode/rules/agents.md.

Previous Review Summaries (6 snapshots, latest commit 53f0a3a)

Current summary above is authoritative. Previous snapshots are kept for context only.

Previous review (commit 53f0a3a)

Status: No Issues Found | Recommendation: Merge

Files Reviewed (1 file)

• uv.lock

Previous review (commit 0605942)

Status: 3 Issues Found | Recommendation: Address before merge

Overview

Severity	Count
WARNING	3

Issue Details (click to expand)

WARNING

File	Line	Issue
uv.lock	1244	CUDA version docs drift
uv.lock	1251	macOS minimum version increased from 11 to 14
uv.lock	1259	Retain support for pre-macOS 14 environments

Files Reviewed (1 file)

• uv.lock - 3 issues

Fix these issues in Kilo Cloud

Previous review (commit 07ab784)

Status: No Issues Found | Recommendation: Merge

Files Reviewed (1 file)

• uv.lock

Previous review (commit 665d912)

Status: No Issues Found | Recommendation: Merge

Files Reviewed (1 file)

• uv.lock

Previous review (commit 9bd8ed6)

Status: No Issues Found | Recommendation: Merge

Files Reviewed (1 file)

• uv.lock

Previous review (commit 25c5a41)

Status: 1 Issue Found | Recommendation: Address before merge

Overview

Severity	Count
WARNING	1

Issue Details (click to expand)

WARNING

File	Line	Issue
uv.lock	1251	macOS minimum version increased from 11 to 14

Files Reviewed (1 file)

• uv.lock - 1 issue

Additional Notes

• Existing Qodo comment on uv.lock:1244 flags CUDA 12.2 vs CUDA 13 documentation drift (not duplicated here).
• torch upgraded from 2.9.1 to 2.13.0 (major version jump); validate the updated dependency graph in the target Windows 11 environment before merging.

Fix these issues in Kilo Cloud

---

_{Reviewed by laguna-s-2.1:free · Input: 271.6K · Output: 24K · Cached: 437.9K}

[codeant-ai]

---

Thanks for using CodeAnt! 🎉

We're free for open-source projects. if you're enjoying it, help us grow by sharing.

Share on X ·
Reddit ·
LinkedIn

[codeant-ai]

---

Thanks for using CodeAnt! 🎉

We're free for open-source projects. if you're enjoying it, help us grow by sharing.

Share on X ·
Reddit ·
LinkedIn

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 665d912f34

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Retain support for pre-macOS 14 environments

The updated lock contains only macosx_14_0_arm64 artifacts for torch, whereas the previous 2.9.1 lock supplied macosx_11_0_arm64 wheels. Because torch is unconditional and the project declares no macOS 14 minimum, uv sync --dry-run --locked --python-platform aarch64-apple-darwin --python 3.12 now fails on uv's macOS 13 target with “doesn't have a source distribution or wheel,” while the parent lock resolves successfully. (uv sync --help defines this option as selecting “the platform for which requirements should be installed.”) Constrain torch to a release that still publishes compatible wheels or explicitly preserve a platform-specific older version.

Useful? React with 👍 / 👎.

[Coldaine]

Diff confirmed: torch 2.9.1 -> 2.13.0, a major version bump of a production ML runtime dependency. Needs a real test run which is impossible with CI down org-wide today. Holding without merge.

[codeant-ai]

---

Thanks for using CodeAnt! 🎉

We're free for open-source projects. if you're enjoying it, help us grow by sharing.

Share on X ·
Reddit ·
LinkedIn

[qodo-code-review]

ⓘ Your Qodo trial ends soon. Ask your workspace admin to set up billing to keep reviews running after the trial. Manage billing

[codeant-ai]

---

Thanks for using CodeAnt! 🎉

We're free for open-source projects. if you're enjoying it, help us grow by sharing.

Share on X ·
Reddit ·
LinkedIn

[codeant-ai]

---

Thanks for using CodeAnt! 🎉

We're free for open-source projects. if you're enjoying it, help us grow by sharing.

Share on X ·
Reddit ·
LinkedIn

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-03T14:14:48.789278Z	0a00f4e	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 53f0a3a8cf

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Pin the lock to the advertised Torch 2.13 release

This PR is scoped and documented as a 2.9.1 → 2.13.0 security update, but the current diff now locks 2.14.0; this is fresh evidence compared with the earlier review's 2.13-based state. Consequently every uv sync --locked installs an unreviewed release with a different CUDA 13 transitive stack, while the PR's release notes and compatibility assessment cover 2.13. Pin the resolution to 2.13, or explicitly rescope and validate the change as a 2.14 upgrade.

Useful? React with 👍 / 👎.