pytorch/pytorch: viable/strict/1787612152
[autograd] Expose queue_callback as torch.autograd.graph.queue_callba…
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.
[autograd] Expose queue_callback as torch.autograd.graph.queue_callba…
[Dynamo] derive registered Stream/Event classes from DeviceInterface …
[ROCm][CI] Bump rocm-preview wheels to August 5 and bring in amdsmi c…
Human Note Part 1 of doing the acutal change I want which is aligning both sdpa and varlen to lower-right the bettter choice of mask semantic in the age of llms. This fixes some edge cases that would have borken…
Fix CUDA illegal memory access in embedding_bag per_sample_weights ba…
Fix CUDA illegal memory access in embedding_bag per_sample_weights ba…
This reverts commit eac5354. Reverted #164560 on behalf of https://github.com/pytorch-auto-revert due to Reverted automatically by pytorch's autorevert, to avoid this behaviour add the tag autorevert: disable (comment)
[XPU][Aten] Fix memory format detection for conv backward overrideabl…
[Testcase Refactoring] Enable PrivateUse1 tests and add hw-classifica…
[Testcase Refactoring] Use is_backend_available for MultiProcContinuo…
Summary Add hw_classification = HardwareClassification.GENERIC to all 17 TestCase subclasses in test_store.py, following the convention that every TestCase subclass carries a hardware classification label. Changes…
This PR is auto-generated nightly by this action. Update the pinned vllm hash. Pull Request resolved: #194552 Approved by: https://github.com/pytorchbot
[Test][Refactor] Decouple test_dataloader according to the accelerato…
Corrects misspellings, malformed RST inline-literal markers (single closing backtick instead of double), and a few grammatical slips in comments and docstrings across torch/csrc and torch/distributed. No functional…
Corrects misspellings, malformed RST inline-literal markers (single closing backtick instead of double), and a few grammatical slips in comments and docstrings across torch/csrc and torch/distributed. No functional…
Follow-up to the recent typo-fix passes, covering another set of grammatical slips in comments and docstrings across inductor, ao quantization, distributed, jit, nn, onnx, and torch/utils. All changes are…
Corrects small grammatical errors and typos across a handful of comments and docstrings: missing articles and verbs ("this a" -> "this is a", "the number elements" -> "the number of elements"), subject-verb agreement in…
I reviewed the implementation and confirm that the following AI-assisted summary accurately describes the change: SymBool equality can produce a SymPy relation whose operands are themselves boolean relations, such as…
I reviewed the implementation and confirm that the following AI-assisted summary accurately describes the change: SymBool equality can produce a SymPy relation whose operands are themselves boolean relations, such as…
My agent generated the quoted summary below. It is included because it explains the root cause, implementation choice, CI baseline updates, and validation for #149909. Dynamo previously graph-broke on RNN, GRU, and LSTM…
Showing 1 day · 32 items available