Markdown SVG upgrades
I started building my markdown-svg-renderer tool in May, but I've since added enough features to it that it's worth talking about here again. It's evolved into my ideal tool for sharing Markdown transcripts that include…
Daily edition · AI
The daily ledger
TILens turns technical updates into a focused daily brief: official releases, trusted reporting, and practitioner analysis, deduplicated and organized by topic.
Qwen 3.8 27B is excellent, but it defaults to wildly overthinking things
Friday's big release was Qwen 3.8 27B, an Apache 2 licensed 27B parameter vision-capable LLM from Alibaba's Qwen research lab. I've been looking forward to this one: 27B is an excellent size for running a model on a…
Quoting Dario Amodei
I do agree that the public has a negative view of AI (and that this is a big problem), but I don’t think it is primarily caused by me or any other AI leader warning about AI’s risks. I think it is fundamentally a crisis…
pytorch/pytorch: trunk/8f988c9c6b3586efbc00a981d9d8cac11f26bcdb: [vllm hash update] update the pinned vllm hash (#193699)
This PR is auto-generated nightly by this action. Update the pinned vllm hash. Pull Request resolved: #193699 Approved by: https://github.com/pytorchbot
pytorch/pytorch: viable/strict/1786868831: [vllm hash update] update the pinned vllm hash (#193699)
This PR is auto-generated nightly by this action. Update the pinned vllm hash. Pull Request resolved: #193699 Approved by: https://github.com/pytorchbot
pytorch/pytorch: trunk/35b14a9f2e0ea67680929967f505f6a723cb3381
Make is_symm_mem_enabled_for_group() not require deprecated enable_sy…
pytorch/pytorch: viable/strict/1786866943
Make is_symm_mem_enabled_for_group() not require deprecated enable_sy…
pytorch/pytorch: trunk/e376a2f789afb49286006eb6562c34eda2d08664
[CI][ROCm] Mark 3 nondeterministic gfx950 models flaky to unblock ind…
pytorch/pytorch: trunk/d09ac90eb44775245c96dd9668de96da9044fd5f: [inductor] Skip pointless cumsum optimization for scalars (#193422)
Fixes #193265. Summary pointless_cumsum_replacement assumes rank >= 1 because it indexes shape[dim]. Rank-0 constants are intentionally excluded from this optimization. Scalar cumsum is already handled by normal…
pytorch/pytorch: viable/strict/1786865240
[functionalization] Hand _make_wrapper_subclass the storage size it a…
pytorch/pytorch: trunk/cfe2e6a46552f5c5eb584b2b2732269a43aa8b08: [BugFix] release tracked fake tensors (#193015)
ShapeEnv.tracked_fakes can hold CUDA tensors after compile. this clears it and removes the leak-check skip from the currenet regression Testing: PYTORCH_TEST_CUDA_MEM_LEAK_CHECK=1 python test/dynamo/test_misc.py…
pytorch/pytorch: trunk/9c69995476c80ab3ef6f926facfed7dc7b5a6cfe
[TEST] Refactor test/test_import_stats.py to make it device-generic (…
pytorch/pytorch: viable/strict/1786863729
[Testcase Refactoring] Add hw_classification in test_dce_pass.py (#1…
pytorch/pytorch: trunk/0145684169b6427c63742ee8fd053df9695cb727: Canonicalize trunc of symbolic integer division by one (#186194)
Issue #138926 reported a torch.cond metadata mismatch where one branch returned a tensor sized with TruncToInt(IntTrueDiv(s, 1)) while the other returned the same tensor sized with s. The branches were semantically…
pytorch/pytorch: trunk/8bd9c2032e242bfc018e75015e93f29607a81289: Document the main build stages at the top of CMakeLists.txt (#193615)
Author Notes Just because I was wondering and couldn't find the info easily. Agent Notes The text below was generated by Claude Code (as was the diff, which I reviewed). The root CMakeLists.txt has no map of what…
pytorch/pytorch: viable/strict/1786861670: Document the main build stages at the top of CMakeLists.txt (#193615)
Author Notes Just because I was wondering and couldn't find the info easily. Agent Notes The text below was generated by Claude Code (as was the diff, which I reviewed). The root CMakeLists.txt has no map of what…
pytorch/pytorch: ciflow/torchtitan/193697: Update
[ghstack-poisoned]
pytorch/pytorch: ciflow/trunk/186194: [UPDATE] Update
[ghstack-poisoned]
pytorch/pytorch: ciflow/trunk/193699
update vllm commit hash