Daily edition · AI Tools
TILens turns technical updates into a focused daily brief: official releases, trusted reporting, and practitioner analysis, deduplicated and organized by topic.
Changes since langchain-openai==1.5.2 release(openai): 1.6.0 (#39762) feat(core): add standard model exception types (#39538) fix(openai): raise clear error on unexpected response type in _create_chat_result (#39731)
Changes since langchain-anthropic==1.5.6 release(anthropic): 1.6.0 (#39763) feat(core): add standard model exception types (#39538) fix(anthropic): exclude sibling directories from grep search scope (#39681)…
Revert ideep for cpu to 2.13 (#194087)
Changes since langchain-core==1.5.6 release(core): 1.6.0 (#39760) fix(core): resolve postponed annotations in StructuredTool._injected_args_keys (#39602) feat(core): add standard model exception types (#39538)…
Fixes intel/torch-xpu-ops#3010 Summary DTensor random ops (bernoulli, dropout, normal_, rand_like, etc.) failed under torch.compile on XPU because the traceable, graph-safe RNG path was CUDA-only. XPU fell back to the…
[inductor] Fix overlap scheduling crash when both memory caps are Non…
Follow-up to #168097, to ensure extensions can see/use the semantically-correct version. Should have really been done in conjunction with that PR? Useful in cases where a PyTorch extension wants to compare the version…
Make empty MPS addbmm and baddbmm ignore NaN inputs when beta is zero. Reproducer: import torch for device in ("cpu", "mps"): x = torch.full((2, 2), torch.nan, device=device) a = torch.empty(1, 2, 0, device=device) b =…
[torchTLX] tri-state knob with a build-supplied default and a JustKno…
Teach cpp_extension to recognize the win-arm64 VS target and use lib/arm64 for CUDA and cuDNN library paths when building on Windows ARM64. This keeps Windows ARM64 extension builds from looking under x64 CUDA library…
Summary Enable torch._scaled_mm and torch._scaled_mm_v2 OpInfo tests on ROCm FP8-capable GPUs by replacing the blanket TEST_WITH_ROCM skip with PLATFORM_SUPPORTS_FP8. Sample inputs use platform-native FP8 dtypes…
The py3.13/3.14/3.14t clang21 docker images have been failing to build since 2026-08-17. scikit-image==0.22.0 has no cp313/cp314 wheel, so pip builds it from source and resolves an unpinned pythran into the…
[dynamo] Split DequeIteratorVariable off ListIteratorVariable to matc…
[dynamo] Split DequeIteratorVariable off ListIteratorVariable to matc…
[xpu][skip] update skip all platform to skip specific platform for te…
[inductor] AOTI: materialize scalar-in-Tensor-arg for proxy-executor …
Impact: docs only (functorch ensembling tutorial) Risk: low What Removes a duplicated word ("each of of" -> "each of") in a comment in the functorch ensembling tutorial. Why Small typo cleanup in the tutorial source; no…
Includes the following commits: Stop the disabled-IPC-fabric stub from colliding with the real FabricManager (pytorch/kineto#1541) 4f8caaa [ROCm][CI] Restrict PyTorch build to MI350 to reduce CI build time…
[pytorch][PR] Migrate fastAtomicAdd to headeronly (#193176) (#193604)…
LLVM BOLT is a post-link optimizer that optimize's application's code layout based on execution profile gathered by perf. It has been used to optimize NVIDIA's PyTorch containers since 26.04. This PR brings that…
LLVM BOLT is a post-link optimizer that optimize's application's code layout based on execution profile gathered by perf. It has been used to optimize NVIDIA's PyTorch containers since 26.04. This PR brings that…
