Publish PyTorch backlog intelligence public demo

.gitattributes

@@ -1,35 +1,5 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+*.json text
+*.csv text
+*.md text
+data/dry_run_proof_actions.csv filter=lfs diff=lfs merge=lfs -text
+data/issues.csv filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,10 +1,52 @@
----
-title: Pytorch Backlog Intelligence
-emoji: 🏃
-colorFrom: red
-colorTo: red
-sdk: static
-pinned: false
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+---
+title: PyTorch Backlog Intelligence
+sdk: static
+app_file: index.html
+license: other
+tags:
+  - github
+  - maintainer-tools
+  - backlog
+  - github-backlog
+  - triage
+---
+
+# PyTorch Backlog Intelligence
+
+This is a public, static, read-only Space for an offline maintainer backlog intelligence run over `pytorch/pytorch`.
+
+It represents `15437` open issues and `2770` open pull requests collected from GitHub at `2026-07-09T03:47:02+00:00`.
+
+The issue backlog is collapsed into `2704` public-safe maintainer review threads with `True` full open-issue coverage.
+
+The top workflow compresses those review groups into `9` recommended maintainer moves, grouped by the same next action.
+
+The local board assigns everything to `1579` maintainer action bundles, including `413` patch-plan bundles and `863` existing PR decision packages.
+
+It attaches actionable diagnostics to `1579` bundles, with `505` high-confidence diagnoses from local evidence.
+
+It also includes `15803` no-write audit rows covering all open issues and stale open PRs, with GitHub writes disabled.
+
+## What This Is
+
+- A maintainer attention report for open issues and open pull requests.
+- A reduced public-safe bundle with links, labels, scores, short excerpts, and artifact hashes.
+- A recommended maintainer workflow that starts with maintainer-owned unblocks, then repair-ready PRs, PR lineages, batch issue sweeps, and volunteer bundles.
+- A way to see collapsed work threads, linked issue/PR groups, review guidance, CI gates, and stale/low-urgency items.
+- A way to see how issue rows collapse into PR-backed threads, duplicate candidates, volunteer slices, and batchable component queues.
+- A local board showing what would be reviewed, merged, closed, linked, parked, or converted into a local patch plan first.
+- An actionable diagnostics view that explains likely blockers and the next maintainer step.
+- A no-write audit packet that demonstrates proposed stale PR and issue handling without posting to GitHub.
+
+## What This Is Not
+
+- Not an automated maintainer.
+- Not an instruction to close, label, merge, or comment on anything.
+- Not a replacement for project judgment or contributor context.
+- Not a public dump of raw API payloads.
+
+## Source Repository
+
+- Repository: [pytorch/pytorch](https://github.com/pytorch/pytorch)
+- Full collection: `True`
+- Packet hash: `bb2ad53aa6630f3381c2a30f2e4225d98ab5e28172ceb254137002bd27dd863c`

data/attention_queue.html

@@ -0,0 +1,75 @@
+
+<style>
+:root { color-scheme: light; }
+.bi-wrap { font-family: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; color: #1f2328 !important; }
+.bi-wrap h1, .bi-wrap h2, .bi-wrap h3, .bi-wrap p, .bi-wrap strong, .bi-wrap td, .bi-wrap th, .bi-wrap div { color: #1f2328 !important; }
+.bi-band { border: 1px solid #d8dee4; border-radius: 8px; background: #ffffff !important; padding: 16px; margin: 0 0 12px; }
+.bi-title { display: flex; gap: 12px; align-items: flex-start; justify-content: space-between; margin-bottom: 12px; }
+.bi-title h1, .bi-title h2 { margin: 0; line-height: 1.15; letter-spacing: 0; }
+.bi-title h1 { font-size: 1.55rem; }
+.bi-title h2 { font-size: 1.12rem; }
+.bi-title p { margin: 4px 0 0; color: #57606a; max-width: 820px; }
+.status-pill { border-radius: 999px; padding: 5px 9px; font-size: .78rem; font-weight: 700; background: #dafbe1; color: #116329 !important; white-space: nowrap; }
+.metric-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(138px, 1fr)); gap: 8px; }
+.metric { border: 1px solid #d0d7de; border-radius: 8px; padding: 10px; background: #f6f8fa !important; min-height: 86px; }
+.metric-value { font-size: 1.45rem; font-weight: 800; line-height: 1.1; color: #0969da !important; }
+.metric-label { margin-top: 4px; font-size: .84rem; font-weight: 700; }
+.metric-note { margin-top: 2px; font-size: .76rem; color: #57606a !important; }
+.callout-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 8px; }
+.callout { border-left: 4px solid #0969da; background: #f6f8fa !important; border-radius: 8px; padding: 10px 12px; }
+.callout:nth-child(2) { border-left-color: #1a7f37; }
+.callout:nth-child(3) { border-left-color: #bf8700; }
+.callout:nth-child(4) { border-left-color: #cf222e; }
+.callout strong { display: block; margin-bottom: 4px; }
+.group-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(310px, 1fr)); gap: 10px; }
+.group-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.group-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.meta-row { display: flex; flex-wrap: wrap; gap: 6px; margin: 6px 0; }
+.chip-row { display: inline-flex; flex-wrap: wrap; gap: 5px; align-items: center; vertical-align: middle; }
+.chip { display: inline-flex; align-items: center; border: 1px solid #d0d7de; border-radius: 999px; padding: 2px 7px; font-size: .78rem; background: #f6f8fa !important; color: #24292f !important; }
+.chip.bucket { border-color: #9a6700; background: #fff8c5 !important; }
+.chip.kind { border-color: #0969da; background: #ddf4ff !important; }
+.workflow-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(260px, 1fr)); gap: 10px; }
+.workflow-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.workflow-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.workflow-card .move-count { font-weight: 800; color: #0969da !important; }
+.move-list { display: grid; gap: 10px; }
+.move-row { border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+.move-row summary { list-style: none; cursor: pointer; padding: 12px; }
+.move-row summary::-webkit-details-marker { display: none; }
+.move-summary { display: grid; grid-template-columns: minmax(190px, 1fr) minmax(260px, 1.45fr) minmax(260px, 1.45fr) minmax(150px, .8fr) minmax(250px, 1.35fr) auto; gap: 16px; align-items: start; }
+.move-title, .move-field { min-width: 0; }
+.move-title h3 { margin: 0 0 4px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.move-field span { display: block; color: #57606a !important; font-size: .72rem; font-weight: 800; text-transform: uppercase; letter-spacing: 0; }
+.move-field p { margin: 3px 0 0; font-size: .84rem; line-height: 1.35; overflow-wrap: anywhere; }
+.move-field .chip, .move-title { overflow-wrap: anywhere; }
+.view-bundles { align-self: center; justify-self: end; border: 1px solid #000000; border-radius: 999px; color: #000000 !important; background: #ffffff !important; padding: 4px 9px; font-size: .78rem; font-weight: 800; white-space: nowrap; }
+.move-row[open] .view-bundles { background: #f6f8fa !important; }
+.move-expanded { border-top: 1px solid #d8dee4; padding: 0 12px 12px; }
+.bundle-row { border-top: 1px solid #eef1f4; padding: 12px 0; }
+.bundle-row:first-child { border-top: none; }
+.bundle-row h4 { margin: 0 0 6px; font-size: .94rem; line-height: 1.25; letter-spacing: 0; }
+.bundle-scope { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 7px; margin-top: 8px; }
+.evidence-list { margin: 8px 0 0 18px; padding: 0; color: #57606a !important; font-size: .82rem; line-height: 1.35; }
+.evidence-list li { margin: 3px 0; }
+.muted { color: #57606a !important; }
+.evidence { font-size: .84rem; color: #57606a !important; }
+.scroll-table { overflow-x: auto; border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+table.bi-table { width: 100%; border-collapse: collapse; font-size: .86rem; }
+.bi-table th { text-align: left; background: #f6f8fa !important; color: #57606a !important; font-size: .78rem; border-bottom: 1px solid #d8dee4; padding: 8px; white-space: nowrap; }
+.bi-table td { border-bottom: 1px solid #eef1f4; padding: 8px; vertical-align: top; min-width: 86px; }
+.bi-table tr:last-child td { border-bottom: none; }
+.wide { min-width: 280px; }
+.tight { max-width: 180px; }
+a { color: #0969da !important; text-decoration: none; }
+a:hover { text-decoration: underline; }
+.bi-wrap, .bi-wrap * { color: #000000 !important; }
+@media (max-width: 720px) {
+  .bi-title { display: block; }
+  .status-pill { display: inline-block; margin-top: 8px; }
+  .group-grid { grid-template-columns: 1fr; }
+  .move-summary { grid-template-columns: 1fr; }
+  .view-bundles { justify-self: start; }
+}
+</style>
+<div class='bi-wrap'><section class='bi-band'><div class='bi-title'><div><h2>Attention Queue</h2><p>Sorted by maintainer attention signal, still read-only and evidence-bound.</p></div></div><div class='scroll-table'><table class='bi-table'><thead><tr><th>Item</th><th>Score</th><th>Bucket</th><th>Next Action</th><th>Stale Days</th></tr></thead><tbody><tr><td><a href='https://github.com/pytorch/pytorch/issues/188812' target='_blank' rel='noopener noreferrer'>issue #188812: [MPS] MetalShaderLibrary::getBundledLibrary singleton can crash at process exit</a></td><td>151</td><td>needs design decision</td><td><div class='wide'>confirm whether existing evidence is enough; otherwise wait for a reproducer</div></td><td>4</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188900' target='_blank' rel='noopener noreferrer'>issue #188900: `sparse.mul`: broadcasting a size-1 dimension of the sparse operand silently drops data</a></td><td>139</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>2</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188323' target='_blank' rel='noopener noreferrer'>issue #188323: [Inductor][CPU] dynamic=True convolution lowering crashes with ValueError: Exponent must be non-negative</a></td><td>139</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>8</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/186535' target='_blank' rel='noopener noreferrer'>issue #186535: Windows, gloo: Access violation (0xC0000005) in ProcessGroupGloo::enqueue when calling allreduce on CUDA tensors — GlooAllreduceRegistry has no kCUDA creator</a></td><td>139</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>31</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/187912' target='_blank' rel='noopener noreferrer'>issue #187912: [CPU] Concurrent `cpublas::brgemm` calls can crash in the AMX path when the underlying oneDNN ukernel is shared</a></td><td>127</td><td>needs design decision</td><td><div class='wide'>confirm whether existing evidence is enough; otherwise wait for a reproducer</div></td><td>2</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/157668' target='_blank' rel='noopener noreferrer'>issue #157668: NCCL error caused due to use of NVLS in torch 2.7.1-cu128 on aarch64 gb200 cluster</a></td><td>127</td><td>needs design decision</td><td><div class='wide'>confirm whether existing evidence is enough; otherwise wait for a reproducer</div></td><td>29</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/150022' target='_blank' rel='noopener noreferrer'>issue #150022: Dynamic Shapes with **kwargs</a></td><td>121</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>216</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/162422' target='_blank' rel='noopener noreferrer'>issue #162422: Runtime failure when running torch.compile() &amp; using GCC 11.5.0 on Neoverse V1</a></td><td>121</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>220</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/162476' target='_blank' rel='noopener noreferrer'>issue #162476: heap-buffer-overflow in torch.quantized_max_pool2d via Python API</a></td><td>121</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>300</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189194' target='_blank' rel='noopener noreferrer'>issue #189194: [ROCm] torch 2.13 wheel: &quot;Can&#x27;t detect vectorized ISA for CPU&quot; in torch.compile smoke test on non-ROCm image (regression vs 2.12.1)</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>0</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189150' target='_blank' rel='noopener noreferrer'>issue #189150: SIGSEGV / cudaErrorIllegalAddress when replaying a CUDA graph that captured multiple training iterations of two trainers over deeply unrolled recurrent modules</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>0</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189282' target='_blank' rel='noopener noreferrer'>issue #189282: [CPU] operator_benchmark: backward-pass (add, + batchnorm on aarch64) ~200-500x slower since ~Nov 2025 (x86 + aarch64; possible measurement artifact)</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>0</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189281' target='_blank' rel='noopener noreferrer'>issue #189281: [CPU] operator_benchmark: embedding / embeddingbag ~70-130x slower since ~Nov 2025 (x86 + aarch64; possible measurement issue)</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>0</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189239' target='_blank' rel='noopener noreferrer'>issue #189239: [Inductor]Unable to apply layout optimization on convolution backward causes training performance regression.</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>0</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189144' target='_blank' rel='noopener noreferrer'>issue #189144: [XPU][B580] `flex-attn-causal` performance drop with `last_level_cache_size` cache clear between runs</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>1</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/144965' target='_blank' rel='noopener noreferrer'>issue #144965: RuntimeError &quot;global alloc not supported yet&quot; when using TorchScript optimization.</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>1</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189126' target='_blank' rel='noopener noreferrer'>issue #189126: Report a bug, as requested by the error message</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>1</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189065' target='_blank' rel='noopener noreferrer'>issue #189065: CI for distributed tests on B200 fully red since ~2026-07-01: rank 0 &quot;CUDA-capable device(s) busy or unavailable&quot; -&gt; NCCL id timeout -&gt; 1170-min job timeout</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>2</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/170003' target='_blank' rel='noopener noreferrer'>issue #170003: batch_isend_irecv with nccl causes illegal memory access depending on P2P ordering</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>23</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/174288' target='_blank' rel='noopener noreferrer'>issue #174288: [distributed] Batched isend/irecv with NCCL backend hangs on high load</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>26</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/162731' target='_blank' rel='noopener noreferrer'>issue #162731: DTensor cached op propagation results can be mutated when propagating other ops</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>27</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/153517' target='_blank' rel='noopener noreferrer'>issue #153517: [CI][CUDA][Distributed] test_non_blocking_with_eager_init timeout</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>28</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/167693' target='_blank' rel='noopener noreferrer'>issue #167693: DISABLED test_side_stream_backward_overlap_cuda (__main__.TestAutogradStreamSynchronizationCUDA)</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>28</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/116423' target='_blank' rel='noopener noreferrer'>issue #116423: PyTorch Distributed Elastic Launch Segmentation Fault with Python 3.12</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>31</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/106164' target='_blank' rel='noopener noreferrer'>issue #106164: distributed.batch_isend_irecv() crash when send/recv refers to itself</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>31</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/85088' target='_blank' rel='noopener noreferrer'>issue #85088: reentrant torch.utils.checkpoint does not work with NamedTuple outputs</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>31</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/71303' target='_blank' rel='noopener noreferrer'>issue #71303: [RFC] Cross-Process Performance Analysis: Straggler Detection</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>32</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/75725' target='_blank' rel='noopener noreferrer'>issue #75725: Warning originating in C10 backend does not get translated to Python warning if run from subprocess</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>32</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/76449' target='_blank' rel='noopener noreferrer'>issue #76449: Enhance _verify_param_shape_across_processes</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>32</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/81684' target='_blank' rel='noopener noreferrer'>issue #81684: Message exchange failure when perform alltoallv (cpus)</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>33</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/119845' target='_blank' rel='noopener noreferrer'>issue #119845: Segmentation fault in dataloader worker sub-process</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>33</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/183459' target='_blank' rel='noopener noreferrer'>issue #183459: Regression in torch.distributed._functional_collectives.all_to_all_single in 2.9.1 -&gt; 2.11</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>33</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/180088' target='_blank' rel='noopener noreferrer'>issue #180088: [DTensor] RNG tracker does not advance state for CPU tensors on CUDA mesh, causing trunc_normal_ infinite loop</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>41</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/173921' target='_blank' rel='noopener noreferrer'>issue #173921: [Inductor] Significant numerical divergence (4.7% relative error) in Conv2d on CPU with torch.compile</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>48</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/179502' target='_blank' rel='noopener noreferrer'>issue #179502: [DTensor] sharded view incorrectly passes when redistribution is needed</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>93</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/145610' target='_blank' rel='noopener noreferrer'>issue #145610: mmap fails on 64k page aarch64 systems for AOTI model loading</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>141</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/66504' target='_blank' rel='noopener noreferrer'>issue #66504: BatchNorm runtimeError: one of the variables needed for gradient computation has been modified by an inplace operation</a></td><td>115</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>144</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188632' target='_blank' rel='noopener noreferrer'>pr #188632: Fix Dynamo opaque object staticmethod guards</a></td><td>112</td><td>ready for maintainer decision</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>0</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/186429' target='_blank' rel='noopener noreferrer'>pr #186429: Fix ONNX export of pad_sequence with symbolic split lengths</a></td><td>112</td><td>ready for maintainer decision</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>0</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/186461' target='_blank' rel='noopener noreferrer'>pr #186461: Fix SymBool equality handling in symbolic shapes</a></td><td>112</td><td>ready for maintainer decision</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>0</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/184716' target='_blank' rel='noopener noreferrer'>pr #184716: Fix conv_transpose2d meta output padding validation</a></td><td>112</td><td>ready for maintainer decision</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>0</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/185691' target='_blank' rel='noopener noreferrer'>pr #185691: Avoid UB in float to signed integer casts</a></td><td>112</td><td>ready for maintainer decision</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>1</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/184939' target='_blank' rel='noopener noreferrer'>pr #184939: Support SymInt steps for linspace/logspace export</a></td><td>112</td><td>ready for maintainer decision</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>20</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189307' target='_blank' rel='noopener noreferrer'>pr #189307: [CI/CD] Copy newer CUPTI headers into manywheel binary-build images</a></td><td>108</td><td>waiting on CI/check fix</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>0</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/186927' target='_blank' rel='noopener noreferrer'>pr #186927: [MPS] Gemv kernels</a></td><td>108</td><td>waiting on CI/check fix</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>0</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189312' target='_blank' rel='noopener noreferrer'>pr #189312: [CUDA][cuBLAS] Change cuBLAS default workspace size for SM 11.0 to 32 MiB</a></td><td>108</td><td>waiting on CI/check fix</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>0</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188301' target='_blank' rel='noopener noreferrer'>pr #188301: [inductor] Fix inductor dropping ordering dep between effectful ops with different kernel types</a></td><td>108</td><td>waiting on CI/check fix</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>0</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/184648' target='_blank' rel='noopener noreferrer'>pr #184648: [Test] Add has_sufficient_memory() hook to DeviceTypeTestBase</a></td><td>108</td><td>waiting on CI/check fix</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>1</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188237' target='_blank' rel='noopener noreferrer'>pr #188237: Optimize CSR SpMM CPU grain size and inner accumulation</a></td><td>108</td><td>waiting on CI/check fix</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>1</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/187501' target='_blank' rel='noopener noreferrer'>pr #187501: [Stable C Shim] Use error message retrieval shim if available at runtime</a></td><td>108</td><td>waiting on CI/check fix</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>13</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/185882' target='_blank' rel='noopener noreferrer'>pr #185882: [PrivateUse1] Return `False` instead of `None` when PU1 backend is not available</a></td><td>108</td><td>waiting on CI/check fix</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>15</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/153410' target='_blank' rel='noopener noreferrer'>issue #153410: [Feature request] `torch.export` .save/.load could support `safetensors` and/or `weights_only=True`</a></td><td>107</td><td>needs design decision</td><td><div class='wide'>review linked PR state before touching issue</div></td><td>171</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188892' target='_blank' rel='noopener noreferrer'>issue #188892: nvidia-cudnn-cu13==9.20.0.48 bundled by torch is incomplete → CUDNN_STATUS_SUBLIBRARY_VERSION_MISMATCH</a></td><td>106</td><td>has linked PR</td><td><div class='wide'>confirm whether existing evidence is enough; otherwise wait for a reproducer</div></td><td>1</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188970' target='_blank' rel='noopener noreferrer'>issue #188970: [MPS] SIGSEGV in MPSStream::copy with triton.cudagraphs (reduce-overhead/max-autotune) on deep fp16 graphs</a></td><td>106</td><td>has linked PR</td><td><div class='wide'>confirm whether existing evidence is enough; otherwise wait for a reproducer</div></td><td>1</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/135631' target='_blank' rel='noopener noreferrer'>pr #135631: [scan] Autograd</a></td><td>100</td><td>waiting on CI/check fix</td><td><div class='wide'>final maintainer merge/release decision</div></td><td>521</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188115' target='_blank' rel='noopener noreferrer'>pr #188115: [Inductor][TEST] Align `test_main_loop_scaling` with H100 support surface</a></td><td>98</td><td>waiting on CI/check fix</td><td><div class='wide'>identify whether block is CI, merge conflict, or review gate</div></td><td>6</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/149447' target='_blank' rel='noopener noreferrer'>issue #149447: DTensor slicing on sharded dimension leads to replication</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>200</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/164929' target='_blank' rel='noopener noreferrer'>issue #164929: get_optimizer_state_dict modifies optimizer state</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>211</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/156563' target='_blank' rel='noopener noreferrer'>issue #156563: Segmentation fault (core dumped) in `torch.profiler.profile`</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>231</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/122019' target='_blank' rel='noopener noreferrer'>issue #122019: `F.max_pool3d`: Segfault due to incorrect `dilation` check in kernel</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>285</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/159238' target='_blank' rel='noopener noreferrer'>issue #159238: Performance regression: torch.jit.trace() significantly slower on RTX 5090 than RTX 4060 (cu128 nightly)</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>333</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/121219' target='_blank' rel='noopener noreferrer'>issue #121219: Torch profiler corrupted names with Python 3.11</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>335</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/91165' target='_blank' rel='noopener noreferrer'>issue #91165: [FSDP] FSDP with CPU offload consumes `1.65X` more GPU memory when training models with most of the params frozen</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>339</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/47260' target='_blank' rel='noopener noreferrer'>issue #47260: DDP doesn&#x27;t work with retain_graph = True</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>394</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/100253' target='_blank' rel='noopener noreferrer'>issue #100253: profiler.export_stacks doesn&#x27;t return stack trace unless experimental_config is provided</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>395</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/155147' target='_blank' rel='noopener noreferrer'>issue #155147: torch.profiler raises Aborted (core dumped) failurer related with GIL (gilstate_tss_set)</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>397</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/155146' target='_blank' rel='noopener noreferrer'>issue #155146: Segmentation fault when using torch.profiler</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>398</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/153941' target='_blank' rel='noopener noreferrer'>issue #153941: TorchScript Fails Type Inference When Using .item()-Derived Scalar in Tensor Construction</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>414</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/153260' target='_blank' rel='noopener noreferrer'>issue #153260: Export shouldn&#x27;t run TS under the hood.</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>421</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/153145' target='_blank' rel='noopener noreferrer'>issue #153145: Segmentation faults with CPU FlexAttention</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>425</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/152108' target='_blank' rel='noopener noreferrer'>issue #152108: [inductor][cpu] AMP static shape default wrapper AOTInductor performance regression in 2025_04_20 nightly release</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>440</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/145088' target='_blank' rel='noopener noreferrer'>issue #145088: [torchbench] torch._dynamo.exc.Unsupported: Graph break due to unsupported builtin None.morphologyEx</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>490</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/146788' target='_blank' rel='noopener noreferrer'>issue #146788: Segmentation Fault in `torch.choose_qparams_optimized` with Invalid Parameters</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>514</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/124950' target='_blank' rel='noopener noreferrer'>issue #124950: [distributed] First NCCL barrier does not respect timeout</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>525</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/71855' target='_blank' rel='noopener noreferrer'>issue #71855: Regression in multi-node training speed with Transformers + PyTorch</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>588</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/136712' target='_blank' rel='noopener noreferrer'>issue #136712: Segmentation fault (core dumped) in `torch.ao.nn.quantized.dynamic.LSTMCell/GRUCell`</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>650</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/136728' target='_blank' rel='noopener noreferrer'>issue #136728: Aborted (core dumped) in `torch.package.package_exporter.PackageExporter`/`torch.package.PackageExporter`</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>650</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/135835' target='_blank' rel='noopener noreferrer'>issue #135835: A segmentation fault will be raised when using the `torch._C._jit_to_backend`</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>664</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/81085' target='_blank' rel='noopener noreferrer'>issue #81085: RuntimeError: required keyword attribute &#x27;value&#x27; is undefined</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>667</td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/128693' target='_blank' rel='noopener noreferrer'>issue #128693: Segmentation fault (core dumped) in `torch.fused_moving_avg_obs_fake_quant`</a></td><td>97</td><td>needs design decision</td><td><div class='wide'>review early; this can change release or regression risk</div></td><td>716</td></tr></tbody></table></div></section></div>

data/backlog_report.md

@@ -0,0 +1,149 @@
+# pytorch/pytorch Backlog Intelligence Packet
+
+## Coverage
+- Generated: `2026-07-09T03:47:02+00:00`
+- Open issues represented: `15437` of `15437` listed
+- Open PRs represented: `2770` of `2770` listed
+- Full collection: `True`
+- Work thread active coverage: `True`
+- Issue collapse coverage: `15437` open issues in `2704` groups
+- Issue collapse full coverage: `True`
+- Issue review-unit reduction: `82.5`%
+- Maintainer action bundles: `1579`
+- Patch-plan bundles: `413`
+- Existing PR decision bundles: `863`
+- Full issue/PR bundle coverage: `True`
+- No-write audit rows: `15803`
+- No-write audit boundary: `GitHub writes: none`
+- Relationship edges: `4000`
+- Public surface: `read-only static report`
+
+## Local Board
+- Covered open issues: `15437`
+- Covered open PRs: `2770`
+- Local board boundary: `no GitHub writes`
+- Resolution hash: `141b35358567a5186edb6d910eab9ccd2abd54eb84d9d80d0b16c50fa4844f6d`
+- This local board shows what the tool would review, close, merge, park, or turn into a local patch plan, without doing any of it on GitHub.
+
+## No-Write Audit Packet
+- All open issue proof rows: `15437`
+- Stale open PR proof rows: `366`
+- Proof hash: `ec5b874ff029ecfa5d4ecd204d13f40398adb089965dbb6308b8caec6b268a61`
+- These rows are attachable evidence of what the tool can triage, but they do not comment, label, close, merge, or upload anything.
+
+## Maintainer Review Threads
+- Multi-issue groups: `1957`
+- PR-backed groups: `570`
+- Singleton issue groups retained for audit: `747`
+
+- `1 issue` - Review thread: 1 issue + 1 PR - pr #188801: [MPS] Leak MetalShaderLibrary bundled singleton to avoid exit-time destructor crashes - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#188812; active PRs: PR #188801)
+- `2 issues` - Review thread: 2 issues + 1 PR - pr #189122: Fix sparse-dense mul dropping data when broadcasting a size-1 sparse dim - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#158861, #188900; active PRs: PR #189122)
+- `2 issues` - Review thread: 2 issues + 1 PR - pr #185730: Fix dynamic shapes for variadic kwargs - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#150022, #150371; active PRs: PR #185730)
+- `1 issue` - Review thread: 1 issue + 1 PR - pr #188948: Fix triangular_solve for sparse CPU tensors on non-MKL platforms - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#153410; active PRs: PR #188948)
+- `3 issues` - Review thread: 3 issues + 2 PRs - pr #181720: [MPS] Make pin_memory return CPU-aliased storage backed by a unified MTLBuffer - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#180397, #181374, #188970; active PRs: PR #181720, PR #189256)
+- `1 issue` - Review thread: 1 issue + 1 PR - pr #189043: Preload full bundled cuDNN set with RTLD_GLOBAL to prevent sublibrary version mismatch - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#188892; active PRs: PR #189043)
+- `21 issues` - Review thread: 21 issues + 14 PRs - pr #181726: [xpu][1/4]Implement scaled_mm_v2 for MXFP8/MXFP4/NVFP4 on XPU - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#178040, #183988, #186348, #187988, #188477, #188675, #188704, #188706, #188707, #188708, #188709, #188710; active PRs: PR #181726, PR #181727, PR #181728, PR #183511, PR #184276, PR #187315, PR #187318, PR #187989)
+- `3 issues` - Review thread: 3 issues + 3 PRs - pr #184481: Guard synthetic-base input alias layouts - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#93617, #178680, #188133; active PRs: PR #184481, PR #184694, PR #185891)
+- `3 issues` - Review thread: 3 issues + 2 PRs - pr #185946: [CUDA] Warn instead of assert in `reportProcessMemoryInfo` to prevent crash on Tegra OOM - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#16706, #185240, #186374; active PRs: PR #185946, PR #186375)
+- `2 issues` - Review thread: 2 issues + 3 PRs - pr #185846: [decomp] Match eager's device-dependent threshold comparison dtype - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#185470, #185484; active PRs: PR #185846, PR #186358, PR #187609)
+- `3 issues` - Review thread: 3 issues + 1 PR - pr #189045: Fix NaN gradients in nn.MultiheadAttention for fully-masked rows (#41… - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#40932, #41508, #55056; active PRs: PR #189045)
+- `3 issues` - Review thread: 3 issues + 1 PR - pr #188998: [inductor] Fix bf16 matmul accumulation precision in Triton templates (#188492) - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#188492, #188841, #189287; active PRs: PR #188998)
+- `3 issues` - Review thread: 3 issues + 1 PR - pr #184621: Fix Dynamo torch.Size from tensor shape inputs - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#182649, #182651, #182652; active PRs: PR #184621)
+- `2 issues` - Review thread: 2 issues + 2 PRs - pr #189294: [CI] Realign scipy after nvmath install to fix H100 smoke NumPy ABI crash - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#173059, #189034; active PRs: PR #189294, PR #189295)
+- `2 issues` - Review thread: 2 issues + 2 PRs - pr #187354: Match Inductor i0/i1 infinities to eager CUDA - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#187332, #188545; active PRs: PR #187354, PR #188556)
+- `2 issues` - Review thread: 2 issues + 2 PRs - pr #185756: [clamp] Fix float16 scalar overflow check inconsistency between CPU and GPU - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#171356, #187429; active PRs: PR #185756, PR #187908)
+- `2 issues` - Review thread: 2 issues + 1 PR - pr #185496: Fix export autograd.grad saved tensor tracing - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#146719, #155044; active PRs: PR #185496)
+- `2 issues` - Review thread: 2 issues + 1 PR - pr #184632: Fix CUDA fake strides for mixed dtype pointwise ops - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#182200, #184101; active PRs: PR #184632)
+- `2 issues` - Review thread: 2 issues + 1 PR - pr #184628: Fix SingletonInt static guard evaluation - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#182217, #183369; active PRs: PR #184628)
+- `2 issues` - Review thread: 2 issues + 1 PR - pr #184392: Fix ModularIndexing printer semantics - linked work thread - Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately. (#119883, #187027; active PRs: PR #184392)
+
+## Maintainer Work Threads
+- `151` pr #188801: [MPS] Leak MetalShaderLibrary bundled singleton to avoid exit-time destructor crashes - waiting on contributor - Wait for contributor update on PR #188801; keep related issue/PR context attached. (issue #188812, PR #188801)
+- `139` pr #189122: Fix sparse-dense mul dropping data when broadcasting a size-1 sparse dim - ready for maintainer decision - Review active PRs and linked issues as one maintainer work thread. (issue #158861, issue #188900, PR #189122)
+- `139` issue #186535: Windows, gloo: Access violation (0xC0000005) in ProcessGroupGloo::enqueue when calling allreduce on CUDA tensors — GlooAllreduceRegistry has no kCUDA creator - needs triage - needs triage (issue #186535)
+- `139` issue #188323: [Inductor][CPU] dynamic=True convolution lowering crashes with ValueError: Exponent must be non-negative - needs triage - needs triage (issue #188323)
+- `127` issue #157668: NCCL error caused due to use of NVLS in torch 2.7.1-cu128 on aarch64 gb200 cluster - needs triage - needs triage (issue #157668)
+- `127` issue #187912: [CPU] Concurrent `cpublas::brgemm` calls can crash in the AMX path when the underlying oneDNN ukernel is shared - needs triage - needs triage (issue #187912)
+- `121` pr #185730: Fix dynamic shapes for variadic kwargs - PR blocked - Review active PRs and linked issues as one maintainer work thread. (issue #150022, issue #150371, PR #185730)
+- `121` issue #116254: C++ API `at::quantized_max_pool2d`: Heap-buffer-overflow - needs triage - needs triage (issue #116254, issue #162476)
+- `121` issue #162422: Runtime failure when running torch.compile() & using GCC 11.5.0 on Neoverse V1 - needs triage - needs triage (issue #162422)
+- `115` issue #154297: Hangs and timeouts on dist.reduce_scatter on B200 GPU - needs maintainer decision - needs maintainer decision (issue #154297, issue #162178, issue #162745, issue #162748, issue #162820, issue #162871, issue #162897, issue #162917, issue #162940, issue #163429, issue #165170, issue #165685, issue #187158, issue #189065)
+- `115` issue #55655: Performance debugging / warning mode - stale/low urgency - stale/low urgency (issue #55655, issue #57118, issue #68768, issue #72948, issue #75725)
+- `115` issue #66504: BatchNorm runtimeError: one of the variables needed for gradient computation has been modified by an inplace operation - needs triage - needs triage (issue #66504, issue #68407, issue #73332)
+- `115` issue #144965: RuntimeError "global alloc not supported yet" when using TorchScript optimization. - needs triage - needs triage (issue #69078, issue #144965)
+- `115` issue #106164: distributed.batch_isend_irecv() crash when send/recv refers to itself - needs triage - needs triage (issue #106164)
+- `115` issue #116423: PyTorch Distributed Elastic Launch Segmentation Fault with Python 3.12 - needs triage - needs triage (issue #116423)
+- `115` issue #119845: Segmentation fault in dataloader worker sub-process - needs triage - needs triage (issue #119845)
+- `115` issue #145610: mmap fails on 64k page aarch64 systems for AOTI model loading - needs triage - needs triage (issue #145610)
+- `115` issue #153517: [CI][CUDA][Distributed] test_non_blocking_with_eager_init timeout - needs triage - needs triage (issue #153517)
+- `115` issue #162731: DTensor cached op propagation results can be mutated when propagating other ops - needs triage - needs triage (issue #162731)
+- `115` issue #167693: DISABLED test_side_stream_backward_overlap_cuda (__main__.TestAutogradStreamSynchronizationCUDA) - needs triage - needs triage (issue #167693)
+
+## Maintainer Attention Queue
+- `151` issue #188812: [[MPS] MetalShaderLibrary::getBundledLibrary singleton can crash at process exit](https://github.com/pytorch/pytorch/issues/188812) - needs triage - confirm whether existing evidence is enough; otherwise wait for a reproducer
+- `139` issue #188900: [`sparse.mul`: broadcasting a size-1 dimension of the sparse operand silently drops data](https://github.com/pytorch/pytorch/issues/188900) - needs triage - review early; this can change release or regression risk
+- `139` issue #188323: [[Inductor][CPU] dynamic=True convolution lowering crashes with ValueError: Exponent must be non-negative](https://github.com/pytorch/pytorch/issues/188323) - needs triage - review early; this can change release or regression risk
+- `139` issue #186535: [Windows, gloo: Access violation (0xC0000005) in ProcessGroupGloo::enqueue when calling allreduce on CUDA tensors — GlooAllreduceRegistry has no kCUDA creator](https://github.com/pytorch/pytorch/issues/186535) - needs triage - review early; this can change release or regression risk
+- `127` issue #187912: [[CPU] Concurrent `cpublas::brgemm` calls can crash in the AMX path when the underlying oneDNN ukernel is shared](https://github.com/pytorch/pytorch/issues/187912) - needs triage - confirm whether existing evidence is enough; otherwise wait for a reproducer
+- `127` issue #157668: [NCCL error caused due to use of NVLS in torch 2.7.1-cu128 on aarch64 gb200 cluster](https://github.com/pytorch/pytorch/issues/157668) - needs triage - confirm whether existing evidence is enough; otherwise wait for a reproducer
+- `121` issue #150022: [Dynamic Shapes with **kwargs](https://github.com/pytorch/pytorch/issues/150022) - needs triage - review early; this can change release or regression risk
+- `121` issue #162422: [Runtime failure when running torch.compile() & using GCC 11.5.0 on Neoverse V1](https://github.com/pytorch/pytorch/issues/162422) - needs triage - review early; this can change release or regression risk
+- `121` issue #162476: [heap-buffer-overflow in torch.quantized_max_pool2d via Python API](https://github.com/pytorch/pytorch/issues/162476) - needs triage - review early; this can change release or regression risk
+- `115` issue #189194: [[ROCm] torch 2.13 wheel: "Can't detect vectorized ISA for CPU" in torch.compile smoke test on non-ROCm image (regression vs 2.12.1)](https://github.com/pytorch/pytorch/issues/189194) - needs triage - review early; this can change release or regression risk
+- `115` issue #189150: [SIGSEGV / cudaErrorIllegalAddress when replaying a CUDA graph that captured multiple training iterations of two trainers over deeply unrolled recurrent modules](https://github.com/pytorch/pytorch/issues/189150) - needs triage - review early; this can change release or regression risk
+- `115` issue #189282: [[CPU] operator_benchmark: backward-pass (add, + batchnorm on aarch64) ~200-500x slower since ~Nov 2025 (x86 + aarch64; possible measurement artifact)](https://github.com/pytorch/pytorch/issues/189282) - needs triage - review early; this can change release or regression risk
+- `115` issue #189281: [[CPU] operator_benchmark: embedding / embeddingbag ~70-130x slower since ~Nov 2025 (x86 + aarch64; possible measurement issue)](https://github.com/pytorch/pytorch/issues/189281) - needs triage - review early; this can change release or regression risk
+- `115` issue #189239: [[Inductor]Unable to apply layout optimization on convolution backward causes training performance regression.](https://github.com/pytorch/pytorch/issues/189239) - needs triage - review early; this can change release or regression risk
+- `115` issue #189144: [[XPU][B580] `flex-attn-causal` performance drop with `last_level_cache_size` cache clear between runs](https://github.com/pytorch/pytorch/issues/189144) - needs triage - review early; this can change release or regression risk
+- `115` issue #144965: [RuntimeError "global alloc not supported yet" when using TorchScript optimization.](https://github.com/pytorch/pytorch/issues/144965) - needs triage - review early; this can change release or regression risk
+- `115` issue #189126: [Report a bug, as requested by the error message](https://github.com/pytorch/pytorch/issues/189126) - needs triage - review early; this can change release or regression risk
+- `115` issue #189065: [CI for distributed tests on B200 fully red since ~2026-07-01: rank 0 "CUDA-capable device(s) busy or unavailable" -> NCCL id timeout -> 1170-min job timeout](https://github.com/pytorch/pytorch/issues/189065) - needs triage - review early; this can change release or regression risk
+- `115` issue #170003: [batch_isend_irecv with nccl causes illegal memory access depending on P2P ordering](https://github.com/pytorch/pytorch/issues/170003) - needs triage - review early; this can change release or regression risk
+- `115` issue #174288: [[distributed] Batched isend/irecv with NCCL backend hangs on high load](https://github.com/pytorch/pytorch/issues/174288) - needs triage - review early; this can change release or regression risk
+
+## Issue Buckets
+- stale/low urgency: `13044`
+- needs triage: `1822`
+- needs maintainer decision: `1332`
+- has linked PR: `978`
+- high-priority/blocker: `768`
+- needs reproduction: `567`
+- needs design decision: `353`
+- good volunteer slice: `48`
+
+## PR Buckets
+- PR blocked: `1467`
+- has linked issue: `1132`
+- ready for maintainer decision: `558`
+- draft/noise: `495`
+- stale/low urgency: `366`
+- waiting on contributor: `311`
+- high-priority/blocker: `2`
+
+## Top Clusters
+- PR #188980 with linked issues: pr #188980, issue #188790. Review the issue and PR together instead of triaging both separately.
+- PR #167224 with linked issues: pr #167224, issue #134385. Review the issue and PR together instead of triaging both separately.
+- PR #189317 with linked issues: pr #189317, issue #181474. Review the issue and PR together instead of triaging both separately.
+- PR #188632 with linked issues: pr #188632, issue #188544. Review the issue and PR together instead of triaging both separately.
+- PR #186252 with linked issues: pr #186252, issue #184408. Review the issue and PR together instead of triaging both separately.
+- PR #189313 with linked issues: pr #189313, issue #102948. Review the issue and PR together instead of triaging both separately.
+- PR #189286 with linked issues: pr #189286, issue #189133. Review the issue and PR together instead of triaging both separately.
+- PR #188006 with linked issues: pr #188006, issue #168868. Review the issue and PR together instead of triaging both separately.
+- PR #186358 with linked issues: pr #186358, issue #185484. Review the issue and PR together instead of triaging both separately.
+- PR #189129 with linked issues: pr #189129, issue #188890. Review the issue and PR together instead of triaging both separately.
+- PR #186082 with linked issues: pr #186082, issue #140960. Review the issue and PR together instead of triaging both separately.
+- PR #189294 with linked issues: pr #189294, issue #189034. Review the issue and PR together instead of triaging both separately.
+- PR #189288 with linked issues: pr #189288, issue #189271. Review the issue and PR together instead of triaging both separately.
+- PR #188931 with linked issues: pr #188931, issue #188891. Review the issue and PR together instead of triaging both separately.
+- PR #188913 with linked issues: pr #188913, issue #188150. Review the issue and PR together instead of triaging both separately.
+- PR #188998 with linked issues: pr #188998, issue #188492. Review the issue and PR together instead of triaging both separately.
+- PR #189119 with linked issues: pr #189119, issue #189118. Review the issue and PR together instead of triaging both separately.
+- PR #188996 with linked issues: pr #188996, issue #188711. Review the issue and PR together instead of triaging both separately.
+- PR #189010 with linked issues: pr #189010, issue #188866. Review the issue and PR together instead of triaging both separately.
+- PR #189158 with linked issues: pr #189158, issue #189157. Review the issue and PR together instead of triaging both separately.
+
+## Limits
+- This is an offline, read-only maintainer attention report.
+- It does not comment, label, close, merge, or submit anything on GitHub.
+- Public bundle exports are reduced rows and short excerpts; full raw API responses remain local unless explicitly approved.
+- Duplicate clusters are suggestions for maintainer review, not automatic close decisions.

data/cluster_cards.html

@@ -0,0 +1,75 @@
+
+<style>
+:root { color-scheme: light; }
+.bi-wrap { font-family: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; color: #1f2328 !important; }
+.bi-wrap h1, .bi-wrap h2, .bi-wrap h3, .bi-wrap p, .bi-wrap strong, .bi-wrap td, .bi-wrap th, .bi-wrap div { color: #1f2328 !important; }
+.bi-band { border: 1px solid #d8dee4; border-radius: 8px; background: #ffffff !important; padding: 16px; margin: 0 0 12px; }
+.bi-title { display: flex; gap: 12px; align-items: flex-start; justify-content: space-between; margin-bottom: 12px; }
+.bi-title h1, .bi-title h2 { margin: 0; line-height: 1.15; letter-spacing: 0; }
+.bi-title h1 { font-size: 1.55rem; }
+.bi-title h2 { font-size: 1.12rem; }
+.bi-title p { margin: 4px 0 0; color: #57606a; max-width: 820px; }
+.status-pill { border-radius: 999px; padding: 5px 9px; font-size: .78rem; font-weight: 700; background: #dafbe1; color: #116329 !important; white-space: nowrap; }
+.metric-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(138px, 1fr)); gap: 8px; }
+.metric { border: 1px solid #d0d7de; border-radius: 8px; padding: 10px; background: #f6f8fa !important; min-height: 86px; }
+.metric-value { font-size: 1.45rem; font-weight: 800; line-height: 1.1; color: #0969da !important; }
+.metric-label { margin-top: 4px; font-size: .84rem; font-weight: 700; }
+.metric-note { margin-top: 2px; font-size: .76rem; color: #57606a !important; }
+.callout-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 8px; }
+.callout { border-left: 4px solid #0969da; background: #f6f8fa !important; border-radius: 8px; padding: 10px 12px; }
+.callout:nth-child(2) { border-left-color: #1a7f37; }
+.callout:nth-child(3) { border-left-color: #bf8700; }
+.callout:nth-child(4) { border-left-color: #cf222e; }
+.callout strong { display: block; margin-bottom: 4px; }
+.group-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(310px, 1fr)); gap: 10px; }
+.group-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.group-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.meta-row { display: flex; flex-wrap: wrap; gap: 6px; margin: 6px 0; }
+.chip-row { display: inline-flex; flex-wrap: wrap; gap: 5px; align-items: center; vertical-align: middle; }
+.chip { display: inline-flex; align-items: center; border: 1px solid #d0d7de; border-radius: 999px; padding: 2px 7px; font-size: .78rem; background: #f6f8fa !important; color: #24292f !important; }
+.chip.bucket { border-color: #9a6700; background: #fff8c5 !important; }
+.chip.kind { border-color: #0969da; background: #ddf4ff !important; }
+.workflow-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(260px, 1fr)); gap: 10px; }
+.workflow-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.workflow-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.workflow-card .move-count { font-weight: 800; color: #0969da !important; }
+.move-list { display: grid; gap: 10px; }
+.move-row { border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+.move-row summary { list-style: none; cursor: pointer; padding: 12px; }
+.move-row summary::-webkit-details-marker { display: none; }
+.move-summary { display: grid; grid-template-columns: minmax(190px, 1fr) minmax(260px, 1.45fr) minmax(260px, 1.45fr) minmax(150px, .8fr) minmax(250px, 1.35fr) auto; gap: 16px; align-items: start; }
+.move-title, .move-field { min-width: 0; }
+.move-title h3 { margin: 0 0 4px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.move-field span { display: block; color: #57606a !important; font-size: .72rem; font-weight: 800; text-transform: uppercase; letter-spacing: 0; }
+.move-field p { margin: 3px 0 0; font-size: .84rem; line-height: 1.35; overflow-wrap: anywhere; }
+.move-field .chip, .move-title { overflow-wrap: anywhere; }
+.view-bundles { align-self: center; justify-self: end; border: 1px solid #000000; border-radius: 999px; color: #000000 !important; background: #ffffff !important; padding: 4px 9px; font-size: .78rem; font-weight: 800; white-space: nowrap; }
+.move-row[open] .view-bundles { background: #f6f8fa !important; }
+.move-expanded { border-top: 1px solid #d8dee4; padding: 0 12px 12px; }
+.bundle-row { border-top: 1px solid #eef1f4; padding: 12px 0; }
+.bundle-row:first-child { border-top: none; }
+.bundle-row h4 { margin: 0 0 6px; font-size: .94rem; line-height: 1.25; letter-spacing: 0; }
+.bundle-scope { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 7px; margin-top: 8px; }
+.evidence-list { margin: 8px 0 0 18px; padding: 0; color: #57606a !important; font-size: .82rem; line-height: 1.35; }
+.evidence-list li { margin: 3px 0; }
+.muted { color: #57606a !important; }
+.evidence { font-size: .84rem; color: #57606a !important; }
+.scroll-table { overflow-x: auto; border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+table.bi-table { width: 100%; border-collapse: collapse; font-size: .86rem; }
+.bi-table th { text-align: left; background: #f6f8fa !important; color: #57606a !important; font-size: .78rem; border-bottom: 1px solid #d8dee4; padding: 8px; white-space: nowrap; }
+.bi-table td { border-bottom: 1px solid #eef1f4; padding: 8px; vertical-align: top; min-width: 86px; }
+.bi-table tr:last-child td { border-bottom: none; }
+.wide { min-width: 280px; }
+.tight { max-width: 180px; }
+a { color: #0969da !important; text-decoration: none; }
+a:hover { text-decoration: underline; }
+.bi-wrap, .bi-wrap * { color: #000000 !important; }
+@media (max-width: 720px) {
+  .bi-title { display: block; }
+  .status-pill { display: inline-block; margin-top: 8px; }
+  .group-grid { grid-template-columns: 1fr; }
+  .move-summary { grid-template-columns: 1fr; }
+  .view-bundles { justify-self: start; }
+}
+</style>
+<div class='bi-wrap'><section class='bi-band'><div class='bi-title'><div><h2>Cluster Cards</h2><p>Older linked/duplicate cards remain available under the work-thread and issue-collapse layers.</p></div></div><div class='scroll-table'><table class='bi-table'><thead><tr><th>Cluster</th><th>Members</th><th>Items</th><th>Maintainer Use</th><th>Similarity</th></tr></thead><tbody><tr><td><strong>PR #188980 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #188980, issue #188790</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #167224 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #167224, issue #134385</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189317 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189317, issue #181474</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #188632 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #188632, issue #188544</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186252 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186252, issue #184408</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189313 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189313, issue #102948</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189286 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189286, issue #189133</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #188006 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #188006, issue #168868</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186358 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186358, issue #185484</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189129 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189129, issue #188890</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186082 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186082, issue #140960</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189294 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189294, issue #189034</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189288 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189288, issue #189271</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #188931 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #188931, issue #188891</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #188913 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #188913, issue #188150</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #188998 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #188998, issue #188492</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189119 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189119, issue #189118</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #188996 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #188996, issue #188711</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189010 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189010, issue #188866</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189158 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189158, issue #189157</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #181895 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #181895, issue #181807</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189141 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189141, issue #188680</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #187989 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #187989, issue #187988</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189125 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189125, issue #169111</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #188012 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #188012, issue #187389</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #188784 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #188784, issue #188128</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189127 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189127, issue #187093</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189149 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189149, issue #186348</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189124 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189124, issue #189121</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189145 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>3</td><td>pr #189145, issue #188325, issue #188405</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189174 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189174, issue #165097</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189234 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189234, issue #187806</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186349 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186349, issue #132635</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186355 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186355, issue #131794</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #184121 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #184121, issue #149982</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #183884 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #183884, issue #186136</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #188795 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #188795, issue #150199</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186364 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186364, issue #129936</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186372 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186372, issue #129648</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186406 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186406, issue #128830</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186415 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186415, issue #128556</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186418 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186418, issue #128554</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186403 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186403, issue #129131</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189273 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189273, issue #189267</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186849 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186849, issue #176178</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #185150 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #185150, issue #161807</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186425 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186425, issue #128160</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #184319 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #184319, issue #128063</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186429 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186429, issue #127153</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186431 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186431, issue #127112</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186443 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186443, issue #126834</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #184124 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #184124, issue #148751</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #184343 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #184343, issue #125277</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186969 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186969, issue #185955</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #187171 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #187171, issue #187170</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #184347 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #184347, issue #125077</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186454 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186454, issue #124863</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186455 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186455, issue #124747</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186457 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186457, issue #124344</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186459 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186459, issue #124181</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186460 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186460, issue #124163</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186461 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186461, issue #124110</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #185433 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #185433, issue #158088</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #185776 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #185776, issue #148475</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186463 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186463, issue #123663</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186464 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186464, issue #123651</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186466 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186466, issue #123470</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186467 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186467, issue #123411</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186557 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186557, issue #90923</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186470 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186470, issue #122860</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #186471 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #186471, issue #122773</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #185926 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #185926, issue #145785</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #185929 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #185929, issue #145773</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189233 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189233, issue #175190</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189045 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #189045, issue #41508</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #184858 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #184858, issue #174541</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #189215 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>3</td><td>pr #189215, issue #186605, issue #186867</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #185947 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #185947, issue #145596</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #185952 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #185952, issue #145574</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr><tr><td><strong>PR #185971 with linked issues</strong><div class='muted'>linked_issue_pr</div></td><td>2</td><td>pr #185971, issue #145445</td><td><div class='wide'>Review the issue and PR together instead of triaging both separately.</div></td><td></td></tr></tbody></table></div></section></div>

data/dashboard.html

@@ -0,0 +1,75 @@
+
+<style>
+:root { color-scheme: light; }
+.bi-wrap { font-family: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; color: #1f2328 !important; }
+.bi-wrap h1, .bi-wrap h2, .bi-wrap h3, .bi-wrap p, .bi-wrap strong, .bi-wrap td, .bi-wrap th, .bi-wrap div { color: #1f2328 !important; }
+.bi-band { border: 1px solid #d8dee4; border-radius: 8px; background: #ffffff !important; padding: 16px; margin: 0 0 12px; }
+.bi-title { display: flex; gap: 12px; align-items: flex-start; justify-content: space-between; margin-bottom: 12px; }
+.bi-title h1, .bi-title h2 { margin: 0; line-height: 1.15; letter-spacing: 0; }
+.bi-title h1 { font-size: 1.55rem; }
+.bi-title h2 { font-size: 1.12rem; }
+.bi-title p { margin: 4px 0 0; color: #57606a; max-width: 820px; }
+.status-pill { border-radius: 999px; padding: 5px 9px; font-size: .78rem; font-weight: 700; background: #dafbe1; color: #116329 !important; white-space: nowrap; }
+.metric-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(138px, 1fr)); gap: 8px; }
+.metric { border: 1px solid #d0d7de; border-radius: 8px; padding: 10px; background: #f6f8fa !important; min-height: 86px; }
+.metric-value { font-size: 1.45rem; font-weight: 800; line-height: 1.1; color: #0969da !important; }
+.metric-label { margin-top: 4px; font-size: .84rem; font-weight: 700; }
+.metric-note { margin-top: 2px; font-size: .76rem; color: #57606a !important; }
+.callout-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 8px; }
+.callout { border-left: 4px solid #0969da; background: #f6f8fa !important; border-radius: 8px; padding: 10px 12px; }
+.callout:nth-child(2) { border-left-color: #1a7f37; }
+.callout:nth-child(3) { border-left-color: #bf8700; }
+.callout:nth-child(4) { border-left-color: #cf222e; }
+.callout strong { display: block; margin-bottom: 4px; }
+.group-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(310px, 1fr)); gap: 10px; }
+.group-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.group-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.meta-row { display: flex; flex-wrap: wrap; gap: 6px; margin: 6px 0; }
+.chip-row { display: inline-flex; flex-wrap: wrap; gap: 5px; align-items: center; vertical-align: middle; }
+.chip { display: inline-flex; align-items: center; border: 1px solid #d0d7de; border-radius: 999px; padding: 2px 7px; font-size: .78rem; background: #f6f8fa !important; color: #24292f !important; }
+.chip.bucket { border-color: #9a6700; background: #fff8c5 !important; }
+.chip.kind { border-color: #0969da; background: #ddf4ff !important; }
+.workflow-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(260px, 1fr)); gap: 10px; }
+.workflow-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.workflow-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.workflow-card .move-count { font-weight: 800; color: #0969da !important; }
+.move-list { display: grid; gap: 10px; }
+.move-row { border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+.move-row summary { list-style: none; cursor: pointer; padding: 12px; }
+.move-row summary::-webkit-details-marker { display: none; }
+.move-summary { display: grid; grid-template-columns: minmax(190px, 1fr) minmax(260px, 1.45fr) minmax(260px, 1.45fr) minmax(150px, .8fr) minmax(250px, 1.35fr) auto; gap: 16px; align-items: start; }
+.move-title, .move-field { min-width: 0; }
+.move-title h3 { margin: 0 0 4px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.move-field span { display: block; color: #57606a !important; font-size: .72rem; font-weight: 800; text-transform: uppercase; letter-spacing: 0; }
+.move-field p { margin: 3px 0 0; font-size: .84rem; line-height: 1.35; overflow-wrap: anywhere; }
+.move-field .chip, .move-title { overflow-wrap: anywhere; }
+.view-bundles { align-self: center; justify-self: end; border: 1px solid #000000; border-radius: 999px; color: #000000 !important; background: #ffffff !important; padding: 4px 9px; font-size: .78rem; font-weight: 800; white-space: nowrap; }
+.move-row[open] .view-bundles { background: #f6f8fa !important; }
+.move-expanded { border-top: 1px solid #d8dee4; padding: 0 12px 12px; }
+.bundle-row { border-top: 1px solid #eef1f4; padding: 12px 0; }
+.bundle-row:first-child { border-top: none; }
+.bundle-row h4 { margin: 0 0 6px; font-size: .94rem; line-height: 1.25; letter-spacing: 0; }
+.bundle-scope { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 7px; margin-top: 8px; }
+.evidence-list { margin: 8px 0 0 18px; padding: 0; color: #57606a !important; font-size: .82rem; line-height: 1.35; }
+.evidence-list li { margin: 3px 0; }
+.muted { color: #57606a !important; }
+.evidence { font-size: .84rem; color: #57606a !important; }
+.scroll-table { overflow-x: auto; border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+table.bi-table { width: 100%; border-collapse: collapse; font-size: .86rem; }
+.bi-table th { text-align: left; background: #f6f8fa !important; color: #57606a !important; font-size: .78rem; border-bottom: 1px solid #d8dee4; padding: 8px; white-space: nowrap; }
+.bi-table td { border-bottom: 1px solid #eef1f4; padding: 8px; vertical-align: top; min-width: 86px; }
+.bi-table tr:last-child td { border-bottom: none; }
+.wide { min-width: 280px; }
+.tight { max-width: 180px; }
+a { color: #0969da !important; text-decoration: none; }
+a:hover { text-decoration: underline; }
+.bi-wrap, .bi-wrap * { color: #000000 !important; }
+@media (max-width: 720px) {
+  .bi-title { display: block; }
+  .status-pill { display: inline-block; margin-top: 8px; }
+  .group-grid { grid-template-columns: 1fr; }
+  .move-summary { grid-template-columns: 1fr; }
+  .view-bundles { justify-self: start; }
+}
+</style>
+<div class='bi-wrap'><section class='bi-band'><div class='bi-title'><div><h1>PyTorch Backlog Intelligence</h1><p>A public static, read-only backlog compression view. It turns individual notifications into an auditable maintainer work plan with linked evidence underneath. No GitHub writes are performed.</p></div><span class='status-pill'>all open issues represented</span></div><div class='metric-grid'><div class='metric'><div class='metric-value'>15437</div><div class='metric-label'>Open issues</div><div class='metric-note'>all represented</div></div><div class='metric'><div class='metric-value'>2770</div><div class='metric-label'>Open PRs</div><div class='metric-note'>all represented</div></div><div class='metric'><div class='metric-value'>312</div><div class='metric-label'>Workflow lanes</div><div class='metric-note'>oncall/module queues</div></div><div class='metric'><div class='metric-value'>2704</div><div class='metric-label'>Review groups</div><div class='metric-note'>82.5% fewer review units</div></div><div class='metric'><div class='metric-value'>9</div><div class='metric-label'>Maintainer moves</div><div class='metric-note'>99% fewer than review groups</div></div><div class='metric'><div class='metric-value'>15739</div><div class='metric-label'>Work threads</div><div class='metric-note'>expandable evidence graph</div></div><div class='metric'><div class='metric-value'>1579</div><div class='metric-label'>Evidence bundles</div><div class='metric-note'>local-only</div></div><div class='metric'><div class='metric-value'>413</div><div class='metric-label'>Patch-plan bundles</div><div class='metric-note'>no GitHub writes</div></div></div></section><section class='bi-band'><div class='bi-title'><div><h2>Recommended Workflow</h2><p>This static run compresses 15437 open issues and 2770 open PRs into 2704 review groups, then into 9 recommended maintainer moves across 312 workflow lanes. Start with maintainer-owned unblock work, then repair-ready PRs, oncall/module triage, issue sweeps, and volunteer-ready bundles.</p></div><span class='status-pill'>local-only; no GitHub writes</span></div><div class='workflow-grid'><article class='workflow-card'><h3>Do First: Maintainer-Owned Unblocks</h3><div class='move-count'>1 action queue &middot; 259 issues affected &middot; 988 PRs</div><p class='evidence'>Stalled work where the next action is maintainer-owned, such as repo-owner test approval or final review.</p><p class='evidence'><strong>Example:</strong> PRs waiting on repo-owner test approval are separated from contributor-blocked PRs, so maintainers can unblock them without rereading the full thread.</p><ul><li><strong>make final PR decision</strong><div class='muted'>329 bundles &middot; 259 issues &middot; 988 PRs</div></li></ul></article><article class='workflow-card'><h3>Do Second: Repair-Ready PRs</h3><div class='move-count'>2 action queues &middot; 711 issues affected &middot; 1728 PRs</div><p class='evidence'>Open PRs with named failing checks, merge state, or concrete repair direction.</p><ul><li><strong>fix failing CI checks</strong><div class='muted'>484 bundles &middot; 711 issues &middot; 1718 PRs</div></li><li><strong>fix failing static/type checks</strong><div class='muted'>1 bundle &middot; 0 issues &middot; 10 PRs</div></li></ul></article><article class='workflow-card'><h3>Do Third: PR Lineage Cleanup</h3><div class='move-count'>1 action queue &middot; 329 issues affected &middot; 2179 PRs</div><p class='evidence'>Older, superseded, or overlapping PR attempts that should be reviewed together.</p><ul><li><strong>resolve overlapping PR lineage</strong><div class='muted'>272 bundles &middot; 329 issues &middot; 2179 PRs</div></li></ul></article><article class='workflow-card'><h3>Do Fourth: Batch Issue Sweeps</h3><div class='move-count'>3 action queues &middot; 10003 issues affected &middot; 46 PRs</div><p class='evidence'>Grouped issue families where one policy, duplicate, close, or clarification pass can cover several rows.</p><ul><li><strong>park or close contributor-waiting PRs</strong><div class='muted'>46 bundles &middot; 48 issues &middot; 46 PRs</div></li><li><strong>consolidate duplicate candidates</strong><div class='muted'>22 bundles &middot; 184 issues &middot; 0 PRs</div></li><li><strong>apply stale sweep policy</strong><div class='muted'>281 bundles &middot; 9771 issues &middot; 0 PRs</div></li></ul></article><article class='workflow-card'><h3>Do Fifth: Volunteer Bundles</h3><div class='move-count'>2 action queues &middot; 4443 issues affected &middot; 0 PRs</div><p class='evidence'>Contributor-ready issue slices with shared scope, labels, and test expectations.</p><ul><li><strong>prepare patch-plan bundle</strong><div class='muted'>398 bundles &middot; 4421 issues &middot; 0 PRs</div></li><li><strong>prepare volunteer-ready bundle</strong><div class='muted'>15 bundles &middot; 22 issues &middot; 0 PRs</div></li></ul></article></div></section><section class='bi-band'><div class='bi-title'><div><h2>How This Reduces Maintainer Time</h2><p>The output does not act on GitHub. It turns individual notifications into evidence-bound review units.</p></div></div><div class='callout-grid'><div class='callout'><strong>Maintainer moves first</strong>Shows what a maintainer should do first, second, third, and why.</div><div class='callout'><strong>Threads stay expandable</strong>Review threads, relationship edges, and raw rows remain underneath for audit.</div><div class='callout'><strong>Specific blockers</strong>PR states are split into owner gates, contributor waits, check fixes, review decisions, and branch updates.</div><div class='callout'><strong>Batchable issue work</strong>Duplicate, stale, waiting-info, and volunteer-ready rows are handled as grouped sessions.</div></div></section></div>

data/downloads.html

@@ -0,0 +1,75 @@
+
+<style>
+:root { color-scheme: light; }
+.bi-wrap { font-family: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; color: #1f2328 !important; }
+.bi-wrap h1, .bi-wrap h2, .bi-wrap h3, .bi-wrap p, .bi-wrap strong, .bi-wrap td, .bi-wrap th, .bi-wrap div { color: #1f2328 !important; }
+.bi-band { border: 1px solid #d8dee4; border-radius: 8px; background: #ffffff !important; padding: 16px; margin: 0 0 12px; }
+.bi-title { display: flex; gap: 12px; align-items: flex-start; justify-content: space-between; margin-bottom: 12px; }
+.bi-title h1, .bi-title h2 { margin: 0; line-height: 1.15; letter-spacing: 0; }
+.bi-title h1 { font-size: 1.55rem; }
+.bi-title h2 { font-size: 1.12rem; }
+.bi-title p { margin: 4px 0 0; color: #57606a; max-width: 820px; }
+.status-pill { border-radius: 999px; padding: 5px 9px; font-size: .78rem; font-weight: 700; background: #dafbe1; color: #116329 !important; white-space: nowrap; }
+.metric-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(138px, 1fr)); gap: 8px; }
+.metric { border: 1px solid #d0d7de; border-radius: 8px; padding: 10px; background: #f6f8fa !important; min-height: 86px; }
+.metric-value { font-size: 1.45rem; font-weight: 800; line-height: 1.1; color: #0969da !important; }
+.metric-label { margin-top: 4px; font-size: .84rem; font-weight: 700; }
+.metric-note { margin-top: 2px; font-size: .76rem; color: #57606a !important; }
+.callout-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 8px; }
+.callout { border-left: 4px solid #0969da; background: #f6f8fa !important; border-radius: 8px; padding: 10px 12px; }
+.callout:nth-child(2) { border-left-color: #1a7f37; }
+.callout:nth-child(3) { border-left-color: #bf8700; }
+.callout:nth-child(4) { border-left-color: #cf222e; }
+.callout strong { display: block; margin-bottom: 4px; }
+.group-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(310px, 1fr)); gap: 10px; }
+.group-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.group-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.meta-row { display: flex; flex-wrap: wrap; gap: 6px; margin: 6px 0; }
+.chip-row { display: inline-flex; flex-wrap: wrap; gap: 5px; align-items: center; vertical-align: middle; }
+.chip { display: inline-flex; align-items: center; border: 1px solid #d0d7de; border-radius: 999px; padding: 2px 7px; font-size: .78rem; background: #f6f8fa !important; color: #24292f !important; }
+.chip.bucket { border-color: #9a6700; background: #fff8c5 !important; }
+.chip.kind { border-color: #0969da; background: #ddf4ff !important; }
+.workflow-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(260px, 1fr)); gap: 10px; }
+.workflow-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.workflow-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.workflow-card .move-count { font-weight: 800; color: #0969da !important; }
+.move-list { display: grid; gap: 10px; }
+.move-row { border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+.move-row summary { list-style: none; cursor: pointer; padding: 12px; }
+.move-row summary::-webkit-details-marker { display: none; }
+.move-summary { display: grid; grid-template-columns: minmax(190px, 1fr) minmax(260px, 1.45fr) minmax(260px, 1.45fr) minmax(150px, .8fr) minmax(250px, 1.35fr) auto; gap: 16px; align-items: start; }
+.move-title, .move-field { min-width: 0; }
+.move-title h3 { margin: 0 0 4px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.move-field span { display: block; color: #57606a !important; font-size: .72rem; font-weight: 800; text-transform: uppercase; letter-spacing: 0; }
+.move-field p { margin: 3px 0 0; font-size: .84rem; line-height: 1.35; overflow-wrap: anywhere; }
+.move-field .chip, .move-title { overflow-wrap: anywhere; }
+.view-bundles { align-self: center; justify-self: end; border: 1px solid #000000; border-radius: 999px; color: #000000 !important; background: #ffffff !important; padding: 4px 9px; font-size: .78rem; font-weight: 800; white-space: nowrap; }
+.move-row[open] .view-bundles { background: #f6f8fa !important; }
+.move-expanded { border-top: 1px solid #d8dee4; padding: 0 12px 12px; }
+.bundle-row { border-top: 1px solid #eef1f4; padding: 12px 0; }
+.bundle-row:first-child { border-top: none; }
+.bundle-row h4 { margin: 0 0 6px; font-size: .94rem; line-height: 1.25; letter-spacing: 0; }
+.bundle-scope { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 7px; margin-top: 8px; }
+.evidence-list { margin: 8px 0 0 18px; padding: 0; color: #57606a !important; font-size: .82rem; line-height: 1.35; }
+.evidence-list li { margin: 3px 0; }
+.muted { color: #57606a !important; }
+.evidence { font-size: .84rem; color: #57606a !important; }
+.scroll-table { overflow-x: auto; border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+table.bi-table { width: 100%; border-collapse: collapse; font-size: .86rem; }
+.bi-table th { text-align: left; background: #f6f8fa !important; color: #57606a !important; font-size: .78rem; border-bottom: 1px solid #d8dee4; padding: 8px; white-space: nowrap; }
+.bi-table td { border-bottom: 1px solid #eef1f4; padding: 8px; vertical-align: top; min-width: 86px; }
+.bi-table tr:last-child td { border-bottom: none; }
+.wide { min-width: 280px; }
+.tight { max-width: 180px; }
+a { color: #0969da !important; text-decoration: none; }
+a:hover { text-decoration: underline; }
+.bi-wrap, .bi-wrap * { color: #000000 !important; }
+@media (max-width: 720px) {
+  .bi-title { display: block; }
+  .status-pill { display: inline-block; margin-top: 8px; }
+  .group-grid { grid-template-columns: 1fr; }
+  .move-summary { grid-template-columns: 1fr; }
+  .view-bundles { justify-self: start; }
+}
+</style>
+<div class='bi-wrap'><section class='bi-band'><div class='bi-title'><div><h2>Downloads</h2><p>The bundle includes these reduced public-safe artifacts. Full raw API payloads, diffs, and patches are not included.</p></div></div><div class='scroll-table'><table class='bi-table'><thead><tr><th>Artifact</th><th>Role</th></tr></thead><tbody><tr><td><code>data/summary_metrics.json</code></td><td>run metrics and safety flags</td></tr><tr><td><code>data/workflow_lanes.csv</code></td><td>oncall/module/CI workflow lanes</td></tr><tr><td><code>data/evidence_bundles.csv</code></td><td>local-only evidence and patch-plan bundles</td></tr><tr><td><code>data/maintainer_moves.csv</code></td><td>workflow-first maintainer moves grouped by same next action</td></tr><tr><td><code>data/dry_run_proof_actions.csv</code></td><td>no-write audit rows for all issues and stale PRs</td></tr><tr><td><code>data/issue_collapse_groups.csv</code></td><td>public-safe maintainer review threads</td></tr><tr><td><code>data/work_threads.csv</code></td><td>public-safe graph work threads</td></tr><tr><td><code>data/relationship_edges.csv</code></td><td>evidence-backed item links</td></tr><tr><td><code>data/issues.csv</code></td><td>all represented open issues</td></tr><tr><td><code>data/pull_requests.csv</code></td><td>all represented open pull requests</td></tr><tr><td><code>data/attention_queue.csv</code></td><td>ranked issue and PR attention queue</td></tr><tr><td><code>data/cluster_cards.csv</code></td><td>legacy linked and duplicate candidate cards</td></tr><tr><td><code>data/backlog_report.md</code></td><td>readable report</td></tr></tbody></table></div></section></div>

data/dry_run_proof.html

@@ -0,0 +1,75 @@
+
+<style>
+:root { color-scheme: light; }
+.bi-wrap { font-family: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; color: #1f2328 !important; }
+.bi-wrap h1, .bi-wrap h2, .bi-wrap h3, .bi-wrap p, .bi-wrap strong, .bi-wrap td, .bi-wrap th, .bi-wrap div { color: #1f2328 !important; }
+.bi-band { border: 1px solid #d8dee4; border-radius: 8px; background: #ffffff !important; padding: 16px; margin: 0 0 12px; }
+.bi-title { display: flex; gap: 12px; align-items: flex-start; justify-content: space-between; margin-bottom: 12px; }
+.bi-title h1, .bi-title h2 { margin: 0; line-height: 1.15; letter-spacing: 0; }
+.bi-title h1 { font-size: 1.55rem; }
+.bi-title h2 { font-size: 1.12rem; }
+.bi-title p { margin: 4px 0 0; color: #57606a; max-width: 820px; }
+.status-pill { border-radius: 999px; padding: 5px 9px; font-size: .78rem; font-weight: 700; background: #dafbe1; color: #116329 !important; white-space: nowrap; }
+.metric-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(138px, 1fr)); gap: 8px; }
+.metric { border: 1px solid #d0d7de; border-radius: 8px; padding: 10px; background: #f6f8fa !important; min-height: 86px; }
+.metric-value { font-size: 1.45rem; font-weight: 800; line-height: 1.1; color: #0969da !important; }
+.metric-label { margin-top: 4px; font-size: .84rem; font-weight: 700; }
+.metric-note { margin-top: 2px; font-size: .76rem; color: #57606a !important; }
+.callout-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 8px; }
+.callout { border-left: 4px solid #0969da; background: #f6f8fa !important; border-radius: 8px; padding: 10px 12px; }
+.callout:nth-child(2) { border-left-color: #1a7f37; }
+.callout:nth-child(3) { border-left-color: #bf8700; }
+.callout:nth-child(4) { border-left-color: #cf222e; }
+.callout strong { display: block; margin-bottom: 4px; }
+.group-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(310px, 1fr)); gap: 10px; }
+.group-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.group-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.meta-row { display: flex; flex-wrap: wrap; gap: 6px; margin: 6px 0; }
+.chip-row { display: inline-flex; flex-wrap: wrap; gap: 5px; align-items: center; vertical-align: middle; }
+.chip { display: inline-flex; align-items: center; border: 1px solid #d0d7de; border-radius: 999px; padding: 2px 7px; font-size: .78rem; background: #f6f8fa !important; color: #24292f !important; }
+.chip.bucket { border-color: #9a6700; background: #fff8c5 !important; }
+.chip.kind { border-color: #0969da; background: #ddf4ff !important; }
+.workflow-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(260px, 1fr)); gap: 10px; }
+.workflow-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.workflow-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.workflow-card .move-count { font-weight: 800; color: #0969da !important; }
+.move-list { display: grid; gap: 10px; }
+.move-row { border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+.move-row summary { list-style: none; cursor: pointer; padding: 12px; }
+.move-row summary::-webkit-details-marker { display: none; }
+.move-summary { display: grid; grid-template-columns: minmax(190px, 1fr) minmax(260px, 1.45fr) minmax(260px, 1.45fr) minmax(150px, .8fr) minmax(250px, 1.35fr) auto; gap: 16px; align-items: start; }
+.move-title, .move-field { min-width: 0; }
+.move-title h3 { margin: 0 0 4px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.move-field span { display: block; color: #57606a !important; font-size: .72rem; font-weight: 800; text-transform: uppercase; letter-spacing: 0; }
+.move-field p { margin: 3px 0 0; font-size: .84rem; line-height: 1.35; overflow-wrap: anywhere; }
+.move-field .chip, .move-title { overflow-wrap: anywhere; }
+.view-bundles { align-self: center; justify-self: end; border: 1px solid #000000; border-radius: 999px; color: #000000 !important; background: #ffffff !important; padding: 4px 9px; font-size: .78rem; font-weight: 800; white-space: nowrap; }
+.move-row[open] .view-bundles { background: #f6f8fa !important; }
+.move-expanded { border-top: 1px solid #d8dee4; padding: 0 12px 12px; }
+.bundle-row { border-top: 1px solid #eef1f4; padding: 12px 0; }
+.bundle-row:first-child { border-top: none; }
+.bundle-row h4 { margin: 0 0 6px; font-size: .94rem; line-height: 1.25; letter-spacing: 0; }
+.bundle-scope { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 7px; margin-top: 8px; }
+.evidence-list { margin: 8px 0 0 18px; padding: 0; color: #57606a !important; font-size: .82rem; line-height: 1.35; }
+.evidence-list li { margin: 3px 0; }
+.muted { color: #57606a !important; }
+.evidence { font-size: .84rem; color: #57606a !important; }
+.scroll-table { overflow-x: auto; border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+table.bi-table { width: 100%; border-collapse: collapse; font-size: .86rem; }
+.bi-table th { text-align: left; background: #f6f8fa !important; color: #57606a !important; font-size: .78rem; border-bottom: 1px solid #d8dee4; padding: 8px; white-space: nowrap; }
+.bi-table td { border-bottom: 1px solid #eef1f4; padding: 8px; vertical-align: top; min-width: 86px; }
+.bi-table tr:last-child td { border-bottom: none; }
+.wide { min-width: 280px; }
+.tight { max-width: 180px; }
+a { color: #0969da !important; text-decoration: none; }
+a:hover { text-decoration: underline; }
+.bi-wrap, .bi-wrap * { color: #000000 !important; }
+@media (max-width: 720px) {
+  .bi-title { display: block; }
+  .status-pill { display: inline-block; margin-top: 8px; }
+  .group-grid { grid-template-columns: 1fr; }
+  .move-summary { grid-template-columns: 1fr; }
+  .view-bundles { justify-self: start; }
+}
+</style>
+<div class='bi-wrap'><section class='bi-band'><div class='bi-title'><div><h2>No-Write Audit Packet</h2><p>This is the attachable no-write proof: all open issues plus stale open PRs get a proposed maintainer-facing action, with evidence, but no GitHub write is allowed.</p></div><span class='status-pill'>no GitHub actions</span></div><div class='metric-grid'><div class='metric'><div class='metric-value'>15437</div><div class='metric-label'>All issue audit rows</div><div class='metric-note'>every open issue</div></div><div class='metric'><div class='metric-value'>366</div><div class='metric-label'>Stale PR audit rows</div><div class='metric-note'>open PRs with stale bucket</div></div><div class='metric'><div class='metric-value'>0</div><div class='metric-label'>GitHub writes</div><div class='metric-note'>dry-run only</div></div><div class='metric'><div class='metric-value'>ec5b874ff029</div><div class='metric-label'>Proof hash</div><div class='metric-note'>full hash in metrics</div></div></div><div class='scroll-table'><table class='bi-table'><thead><tr><th>Scope</th><th>Item</th><th>Dry-Run Action</th><th>Would Do</th><th>Why It Helps</th><th>GitHub Action Allowed</th></tr></thead><tbody><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/88' target='_blank' rel='noopener noreferrer'>issue #88: expose backend selection and cudnn settings to the end user</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/285' target='_blank' rel='noopener noreferrer'>issue #285: Keyword arguments passed to module&#x27;s __call__ aren&#x27;t forwarded to the hooks</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/499' target='_blank' rel='noopener noreferrer'>issue #499: Feature Request: Locally Connected Layer</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/630' target='_blank' rel='noopener noreferrer'>issue #630: Add Peephole connections for LSTMs?</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/634' target='_blank' rel='noopener noreferrer'>issue #634: Feature Request: NegativeSampling and HierarchicalSoftmax loss functions</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/711' target='_blank' rel='noopener noreferrer'>issue #711: Feature Request: Easier to extend base RNN implementation</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/842' target='_blank' rel='noopener noreferrer'>issue #842: BCELoss doesn&#x27;t accept LongTensor targets</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/934' target='_blank' rel='noopener noreferrer'>issue #934: dataloader parallels over elements vs over batches</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1178' target='_blank' rel='noopener noreferrer'>issue #1178: Batched sparse QR factorizations and solves with cusolver</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1249' target='_blank' rel='noopener noreferrer'>issue #1249: Dice Loss PR</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1328' target='_blank' rel='noopener noreferrer'>issue #1328: Unhelpful CrossEntropyLoss dimension error message</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1329' target='_blank' rel='noopener noreferrer'>issue #1329: [proposed feature] Eve: Improving Stochastic Gradient Descent with Feedback</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1359' target='_blank' rel='noopener noreferrer'>issue #1359: Conjugate Gradient mentioned in docs, but not implemented</a></td><td>review_pr_backed_thread</td><td><div class='wide'>Review active PRs and linked issues as one maintainer work thread.</div></td><td><div class='wide'>Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1362' target='_blank' rel='noopener noreferrer'>issue #1362: Feature Request: noise contrastive estimation/negative sampling</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1369' target='_blank' rel='noopener noreferrer'>issue #1369: &quot;Sparsified&quot; mathematical operations</a></td><td>review_pr_backed_thread</td><td><div class='wide'>Review active PRs and linked issues as one maintainer work thread.</div></td><td><div class='wide'>Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1410' target='_blank' rel='noopener noreferrer'>issue #1410: BatchNorm should use Bessel&#x27;s correction consistently</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1442' target='_blank' rel='noopener noreferrer'>issue #1442: Optimizers can&#x27;t be moved to a different GPU</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1462' target='_blank' rel='noopener noreferrer'>issue #1462: Change sparse_mask to take indexing mask, rather than entire sparse tensor</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1468' target='_blank' rel='noopener noreferrer'>issue #1468: avg_pool functions hold input for backward</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1487' target='_blank' rel='noopener noreferrer'>issue #1487: [feature request/proposal] Relax scale_factor for nearest neighbor upsampling</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1489' target='_blank' rel='noopener noreferrer'>issue #1489: Optimizer should track parameter names and not id</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1505' target='_blank' rel='noopener noreferrer'>issue #1505: Cannot find Intel MKL</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1512' target='_blank' rel='noopener noreferrer'>issue #1512: [feature request] Support tensors of different sizes as batch elements in DataLoader</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1529' target='_blank' rel='noopener noreferrer'>issue #1529: [feature request] Caching allocator diagnostics and memory allocation tracing/visualization</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1591' target='_blank' rel='noopener noreferrer'>issue #1591: Pad PackedSequences to original batch length</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1641' target='_blank' rel='noopener noreferrer'>issue #1641: [Feature request] In-place &#x27;max&#x27; method for Tensor</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1642' target='_blank' rel='noopener noreferrer'>issue #1642: feature request: On-the-fly Operation Batching in Dynamic Computation Graphs</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1686' target='_blank' rel='noopener noreferrer'>issue #1686: Feature Request: load_state_dict should take filenames</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1736' target='_blank' rel='noopener noreferrer'>issue #1736: Expose optimizer options as attributes when there&#x27;s a single param group</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1747' target='_blank' rel='noopener noreferrer'>issue #1747: In-place bernoulli_ has more functionality than torch.bernoulli with output parameter</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1750' target='_blank' rel='noopener noreferrer'>issue #1750: Add support for colors (and maybe other attributes) to NVTX API</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1794' target='_blank' rel='noopener noreferrer'>issue #1794: Feature request: reverse_padded_sequence</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1861' target='_blank' rel='noopener noreferrer'>issue #1861: Factorized Output Layer</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1927' target='_blank' rel='noopener noreferrer'>issue #1927: [feature request] time-distributed layers for application of normal layers to sequence data</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/1932' target='_blank' rel='noopener noreferrer'>issue #1932: Feature Request: ReLU on LSTMs and GRUs</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2001' target='_blank' rel='noopener noreferrer'>issue #2001: Implement similar PyTorch function as model.summary() in keras?</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2129' target='_blank' rel='noopener noreferrer'>issue #2129: [Feature request] truncated normal initializer(sampler)</a></td><td>review_pr_backed_thread</td><td><div class='wide'>Review active PRs and linked issues as one maintainer work thread.</div></td><td><div class='wide'>Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2180' target='_blank' rel='noopener noreferrer'>issue #2180: Discrepancy in BCEWithLogitsLoss and ClassNLLLoss</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2312' target='_blank' rel='noopener noreferrer'>issue #2312: DataParallel is not compatible with pack_padded_sequence</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2400' target='_blank' rel='noopener noreferrer'>issue #2400: Add CRF Layer</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2407' target='_blank' rel='noopener noreferrer'>issue #2407: Hard-negative mining using __getitem__ directive in Dataset class</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2466' target='_blank' rel='noopener noreferrer'>issue #2466: Autograd test failure on ppc64le</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2478' target='_blank' rel='noopener noreferrer'>issue #2478: ReduceLROnPlateau with a naive Backtracking</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2482' target='_blank' rel='noopener noreferrer'>issue #2482: &quot;Shared memory manager connection has timed out&quot;</a></td><td>human_triage_decision</td><td><div class='wide'>Decide issue category, owner, and response pattern; no automated GitHub action.</div></td><td><div class='wide'>The issue is represented and auditable, but the evidence does not justify an automated action.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2512' target='_blank' rel='noopener noreferrer'>issue #2512: detach_() variant that affects all past uses too</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2539' target='_blank' rel='noopener noreferrer'>issue #2539: DataLoader converts cuda FloatTensor into cpu DoubleTensor when shape is (n,)</a></td><td>human_triage_decision</td><td><div class='wide'>Decide issue category, owner, and response pattern; no automated GitHub action.</div></td><td><div class='wide'>The issue is represented and auditable, but the evidence does not justify an automated action.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2545' target='_blank' rel='noopener noreferrer'>issue #2545: Counter-intuitive Patience &amp; Cooldown of ReduceLROnPlateau</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2569' target='_blank' rel='noopener noreferrer'>issue #2569: Add a [strong] hash function for tensor data</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2575' target='_blank' rel='noopener noreferrer'>issue #2575: ImportError: dlopen: cannot load any more object with static TLS</a></td><td>priority_human_review</td><td><div class='wide'>Review early as release/regression risk; no automated GitHub action.</div></td><td><div class='wide'>High priority or blocker signal should stay in a human-owned queue.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2576' target='_blank' rel='noopener noreferrer'>issue #2576: CUDA multinomial is limited to 2^24 categories</a></td><td>priority_human_review</td><td><div class='wide'>Review early as release/regression risk; no automated GitHub action.</div></td><td><div class='wide'>High priority or blocker signal should stay in a human-owned queue.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2628' target='_blank' rel='noopener noreferrer'>issue #2628: BatchNorm{1-2-3}d are redundant</a></td><td>review_pr_backed_thread</td><td><div class='wide'>Review active PRs and linked issues as one maintainer work thread.</div></td><td><div class='wide'>Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2629' target='_blank' rel='noopener noreferrer'>issue #2629: [feature request] add pairwise ranking loss</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2691' target='_blank' rel='noopener noreferrer'>issue #2691: Feature Request: Support grad of grad in fused RNNs</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2703' target='_blank' rel='noopener noreferrer'>issue #2703: Have ppc64le docker images?</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2724' target='_blank' rel='noopener noreferrer'>issue #2724: DataLoader gets stuck after model initialization</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2732' target='_blank' rel='noopener noreferrer'>issue #2732: support grid_sample with batch=1 but supprting batch affine parameters</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2739' target='_blank' rel='noopener noreferrer'>issue #2739: Proposal: simplify overloaded Tensor function signatures</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2849' target='_blank' rel='noopener noreferrer'>issue #2849: Learning rate scheduler have different APIs</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/2854' target='_blank' rel='noopener noreferrer'>issue #2854: Add the new lr_scheduler which called poly</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3076' target='_blank' rel='noopener noreferrer'>issue #3076: BN slows down double-backprop enormously</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3152' target='_blank' rel='noopener noreferrer'>issue #3152: Sparse tensor .new(size) can be confusing</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3176' target='_blank' rel='noopener noreferrer'>issue #3176: Add safety checks in `index_add`/`scatter_add`</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3281' target='_blank' rel='noopener noreferrer'>issue #3281: DataLoader &quot;casting&quot; non statndard objects to lists</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3356' target='_blank' rel='noopener noreferrer'>issue #3356: Data sampling seems to be more complicated than necessary</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3364' target='_blank' rel='noopener noreferrer'>issue #3364: type of torch.bernoulli and torch.multinomial inconsistent</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3390' target='_blank' rel='noopener noreferrer'>issue #3390: High CPU use by clock_gettime syscall</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3396' target='_blank' rel='noopener noreferrer'>issue #3396: CUDA topk is slow for some input sizes</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3468' target='_blank' rel='noopener noreferrer'>issue #3468: improve performance of common CPU clone / contiguous calls with HPTT</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3473' target='_blank' rel='noopener noreferrer'>issue #3473: Feature Request: Distributed send arbitrary objects</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3600' target='_blank' rel='noopener noreferrer'>issue #3600: making .cuda() falls back to an identity function when gpu is not available</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3619' target='_blank' rel='noopener noreferrer'>issue #3619: Multiprocessing with torch.solve hangs</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3625' target='_blank' rel='noopener noreferrer'>issue #3625: Proposal: combine requires_grad and retain_grad()</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3667' target='_blank' rel='noopener noreferrer'>issue #3667: Exposing CuDNN benchmark strategy selection</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3697' target='_blank' rel='noopener noreferrer'>issue #3697: Deprecate inplace argument in torch.nn.functional</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3790' target='_blank' rel='noopener noreferrer'>issue #3790: Add SGDR, SGDW, AdamW and AdamWR</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3791' target='_blank' rel='noopener noreferrer'>issue #3791: Make pytest stop printing docstrings in its default diagnostic output</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3818' target='_blank' rel='noopener noreferrer'>issue #3818: Memory leak when doing backward with grad as yourself</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3823' target='_blank' rel='noopener noreferrer'>issue #3823: Fuse bias to CuDNN convolution</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3863' target='_blank' rel='noopener noreferrer'>issue #3863: Considerable slowdown in Adam.step after a number of epochs with multiple losses</a></td><td>human_triage_decision</td><td><div class='wide'>Decide issue category, owner, and response pattern; no automated GitHub action.</div></td><td><div class='wide'>The issue is represented and auditable, but the evidence does not justify an automated action.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3867' target='_blank' rel='noopener noreferrer'>issue #3867: [Feature Request] Implement &quot;same&quot; padding for convolution operations?</a></td><td>human_triage_decision</td><td><div class='wide'>Decide issue category, owner, and response pattern; no automated GitHub action.</div></td><td><div class='wide'>The issue is represented and auditable, but the evidence does not justify an automated action.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3877' target='_blank' rel='noopener noreferrer'>issue #3877: Wrap Cephes library for mathematical special functions</a></td><td>review_pr_backed_thread</td><td><div class='wide'>Review active PRs and linked issues as one maintainer work thread.</div></td><td><div class='wide'>Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3898' target='_blank' rel='noopener noreferrer'>issue #3898: Sparse matrices in dataloader error</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3904' target='_blank' rel='noopener noreferrer'>issue #3904: Implement DE in pytorch.optim</a></td><td>human_triage_decision</td><td><div class='wide'>Decide issue category, owner, and response pattern; no automated GitHub action.</div></td><td><div class='wide'>The issue is represented and auditable, but the evidence does not justify an automated action.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3920' target='_blank' rel='noopener noreferrer'>issue #3920: [docs] Tensor.new is not documented</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3931' target='_blank' rel='noopener noreferrer'>issue #3931: Suppress hidden state output of RNNs?</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/3990' target='_blank' rel='noopener noreferrer'>issue #3990: Raise an error when using magma built against wrong version of cuda</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4073' target='_blank' rel='noopener noreferrer'>issue #4073: Feature request: Correlation module</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4085' target='_blank' rel='noopener noreferrer'>issue #4085: GridSampler behaviours</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4102' target='_blank' rel='noopener noreferrer'>issue #4102: Make the generator tools data model more explicit</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4123' target='_blank' rel='noopener noreferrer'>issue #4123: Use the int64 version of MKL calls</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4132' target='_blank' rel='noopener noreferrer'>issue #4132: x.grad should be 0 but get NaN after x/0</a></td><td>human_triage_decision</td><td><div class='wide'>Decide issue category, owner, and response pattern; no automated GitHub action.</div></td><td><div class='wide'>The issue is represented and auditable, but the evidence does not justify an automated action.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4145' target='_blank' rel='noopener noreferrer'>issue #4145: [Proposal] Consistent `batch_first` effect for RNN modules</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4148' target='_blank' rel='noopener noreferrer'>issue #4148: ATen explicitly differentiated native function resolution hazard (call is ambiguous)</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4159' target='_blank' rel='noopener noreferrer'>issue #4159: Variable outputs of stochastic functions should never require grad</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4181' target='_blank' rel='noopener noreferrer'>issue #4181: Fused RNN refactor plan</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4186' target='_blank' rel='noopener noreferrer'>issue #4186: Feature Request: CPU performance optimization with MKL-DNN</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4241' target='_blank' rel='noopener noreferrer'>issue #4241: single-gpu works but multi-gpu hangs</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4247' target='_blank' rel='noopener noreferrer'>issue #4247: Feature request: sparse matrix max(axis)</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4334' target='_blank' rel='noopener noreferrer'>issue #4334: torch.cuda.device_count() returns 1 using 4 TitanX setup.</a></td><td>human_triage_decision</td><td><div class='wide'>Decide issue category, owner, and response pattern; no automated GitHub action.</div></td><td><div class='wide'>The issue is represented and auditable, but the evidence does not justify an automated action.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4358' target='_blank' rel='noopener noreferrer'>issue #4358: Invoking MKL in multiprocessing with importing torch causes blocking</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4392' target='_blank' rel='noopener noreferrer'>issue #4392: Cache CuDNN benchmark selection, turn it on by default, use it across PyTorch runs</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4400' target='_blank' rel='noopener noreferrer'>issue #4400: Installation Optimise For Chinese Users Who Behind the Wall</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4406' target='_blank' rel='noopener noreferrer'>issue #4406: DistributedDataParallel doesn&#x27;t converge well when using MPI</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4474' target='_blank' rel='noopener noreferrer'>issue #4474: [docs] Return value type hint to be specified in docs for torch.is_grad_enabled (i.e. -&gt; bool)</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4501' target='_blank' rel='noopener noreferrer'>issue #4501: Bind in Python _backward ATen functions</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4530' target='_blank' rel='noopener noreferrer'>issue #4530: descriptor &#x27;add&#x27; of &#x27;torch._C._VariableBase&#x27; object needs an argument</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4542' target='_blank' rel='noopener noreferrer'>issue #4542: Carefully audit contiguity requirements of code</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4551' target='_blank' rel='noopener noreferrer'>issue #4551: Met &#x27;cudnnDestroyDropoutDescriptor&#x27; while run multiply gpu-based models in multiply processes</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4564' target='_blank' rel='noopener noreferrer'>issue #4564: Mixed Tensor/TensorList arguments in ATen functions with explicit derivatives</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4574' target='_blank' rel='noopener noreferrer'>issue #4574: `from` keyword in `random_` gives error</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4622' target='_blank' rel='noopener noreferrer'>issue #4622: Protect user from No module named _C import error</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4632' target='_blank' rel='noopener noreferrer'>issue #4632: CUDNN_STATUS_INTERNAL_ERROR when training with conv3d</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4636' target='_blank' rel='noopener noreferrer'>issue #4636: Better header hygiene in ATen</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4660' target='_blank' rel='noopener noreferrer'>issue #4660: [Feature proposal] Add MC-derived optimizers</a></td><td>human_triage_decision</td><td><div class='wide'>Decide issue category, owner, and response pattern; no automated GitHub action.</div></td><td><div class='wide'>The issue is represented and auditable, but the evidence does not justify an automated action.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4703' target='_blank' rel='noopener noreferrer'>issue #4703: Very slow on CPU</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4716' target='_blank' rel='noopener noreferrer'>issue #4716: Compilation issue: problem with GPU capability check</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4731' target='_blank' rel='noopener noreferrer'>issue #4731: Consider disallowing Variables that require grad in NCCL/comm functions</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4758' target='_blank' rel='noopener noreferrer'>issue #4758: Clang color diagnostics don&#x27;t work with ninja</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4825' target='_blank' rel='noopener noreferrer'>issue #4825: [feature request]Support AVX512F intrinstics to vectorize operations</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4829' target='_blank' rel='noopener noreferrer'>issue #4829: [Feature Request] clip_grad_norm for sparse gradients</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4906' target='_blank' rel='noopener noreferrer'>issue #4906: Rebuild from no-CUDA to CUDA leads to: error: #error &quot;Expected GLOO_USE_CUDA to be defined&quot;</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4927' target='_blank' rel='noopener noreferrer'>issue #4927: [Feature Request]would PackedSequence support unsorted sequences?</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4930' target='_blank' rel='noopener noreferrer'>issue #4930: [feature request] Type-1 Multi-layer bidirectional RNN</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4952' target='_blank' rel='noopener noreferrer'>issue #4952: [Feature request] Optimize autograd/ATen when a gradient is clearly zero</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4954' target='_blank' rel='noopener noreferrer'>issue #4954: [docs] Docs website search finds duplicates and produces bad snippets</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4958' target='_blank' rel='noopener noreferrer'>issue #4958: [Feature Request] Extract glimpses from a batch of images (as in tf.image.extract_glimpse)</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4959' target='_blank' rel='noopener noreferrer'>issue #4959: Speed up data loading for `TensorDataset` if the underlying dataset supports index by a list of indices</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/4987' target='_blank' rel='noopener noreferrer'>issue #4987: MultiGPU hangs Titan Xp in multiprocessing/queue.py</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5070' target='_blank' rel='noopener noreferrer'>issue #5070: NVIDIA_DRIVER_CAPABILITIES env variable is missing in pytorch docker images</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5096' target='_blank' rel='noopener noreferrer'>issue #5096: [feature request]Add an env variable to cover different pathes when testing code with openmp</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5106' target='_blank' rel='noopener noreferrer'>issue #5106: Saving model with runtime code changes</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5157' target='_blank' rel='noopener noreferrer'>issue #5157: BCELoss - weight parameter shape incorrect</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5159' target='_blank' rel='noopener noreferrer'>issue #5159: TakeBackward taking a significant portion of backward time</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5161' target='_blank' rel='noopener noreferrer'>issue #5161: ASAN detected leaks on python -c &#x27;import torch&#x27;</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5212' target='_blank' rel='noopener noreferrer'>issue #5212: Weird error message in torch.split_size_or_sections</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5231' target='_blank' rel='noopener noreferrer'>issue #5231: [feature request] Stratified splits in random_split function</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5266' target='_blank' rel='noopener noreferrer'>issue #5266: [feature request] warnings for functions with unspecified dim arguments</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5272' target='_blank' rel='noopener noreferrer'>issue #5272: Unsafe out= keyword argument with tensors sharing storage</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5280' target='_blank' rel='noopener noreferrer'>issue #5280: BatchNorm1d raises RuntimeError (CUDNN_STATUS_BAD_PARAM) on 3D input.</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5326' target='_blank' rel='noopener noreferrer'>issue #5326: TestMultiprocessing.test_fd_sharing hangs with ASAN</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5353' target='_blank' rel='noopener noreferrer'>issue #5353: Bugs: Score Function approach in REINFORCE for PONG</a></td><td>priority_human_review</td><td><div class='wide'>Review early as release/regression risk; no automated GitHub action.</div></td><td><div class='wide'>High priority or blocker signal should stay in a human-owned queue.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5385' target='_blank' rel='noopener noreferrer'>issue #5385: Handle python_arg_parser dtype constants better</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5388' target='_blank' rel='noopener noreferrer'>issue #5388: Perf regression: indexing 1-d tensor</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5405' target='_blank' rel='noopener noreferrer'>issue #5405: scatter_add_ should support scalar source (including Python scalar)</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5434' target='_blank' rel='noopener noreferrer'>issue #5434: RuntimeError: $ Torch: not enough memory: you tried to allocate 72GB. Buy new RAM!</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5435' target='_blank' rel='noopener noreferrer'>issue #5435: Gaussian Sampling</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5489' target='_blank' rel='noopener noreferrer'>issue #5489: torch.jit.trace(network, data) fails if data is an OrderedDict</a></td><td>human_triage_decision</td><td><div class='wide'>Decide issue category, owner, and response pattern; no automated GitHub action.</div></td><td><div class='wide'>The issue is represented and auditable, but the evidence does not justify an automated action.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5524' target='_blank' rel='noopener noreferrer'>issue #5524: Redo torch.nn.functional docstring strategy</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5528' target='_blank' rel='noopener noreferrer'>issue #5528: TestNN.test_data_parallel takes 10G of memory</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5565' target='_blank' rel='noopener noreferrer'>issue #5565: [feature request] Different interpolation algos for &#x27;grid_sample&#x27; function</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5580' target='_blank' rel='noopener noreferrer'>issue #5580: [feature request] F.interpolate to support integral data types: bool, int8, int32, int16, int64 ||| support uint8 on CUDA</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5737' target='_blank' rel='noopener noreferrer'>issue #5737: [feature request] Callback on learning rate drop in torch.optim.lr_scheduler.ReduceLROnPlateau</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5740' target='_blank' rel='noopener noreferrer'>issue #5740: [feature request] scalar input to scatter_add_ like scatter_</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5790' target='_blank' rel='noopener noreferrer'>issue #5790: Add hookable weights</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5857' target='_blank' rel='noopener noreferrer'>issue #5857: Change THCudaCheck to suggest that device-side asserts likely mean that you have out of bound indices</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5900' target='_blank' rel='noopener noreferrer'>issue #5900: Build Fails on Gentoo with CUDA 9.1, GCC 6.4, Python 3.5</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5912' target='_blank' rel='noopener noreferrer'>issue #5912: Conv-RNN combination slow in backward pass</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5924' target='_blank' rel='noopener noreferrer'>issue #5924: Install doesn&#x27;t work with spaces in directory</a></td><td>batch_stale_issue_sweep</td><td><div class='wide'>Review in its collapse group for close/reproduce/docs-update decision; do not post automatically.</div></td><td><div class='wide'>Old or low-urgency issue can be handled in a component sweep instead of rediscovering context alone.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/5953' target='_blank' rel='noopener noreferrer'>issue #5953: LBFGS always give nan results, why</a></td><td>human_triage_decision</td><td><div class='wide'>Decide issue category, owner, and response pattern; no automated GitHub action.</div></td><td><div class='wide'>The issue is represented and auditable, but the evidence does not justify an automated action.</div></td><td>False</td></tr><tr><td>all_open_issues</td><td><a href='https://github.com/pytorch/pytorch/issues/6007' target='_blank' rel='noopener noreferrer'>issue #6007: Clean up the extra copies that occur in the execution engine and other places</a></td><td>human_triage_decision</td><td><div class='wide'>Decide issue category, owner, and response pattern; no automated GitHub action.</div></td><td><div class='wide'>The issue is represented and auditable, but the evidence does not justify an automated action.</div></td><td>False</td></tr></tbody></table></div></section></div>

data/dry_run_proof_actions.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11bb3014ceea96ac5cabc75a3d77cacf8f44f8ada1ebcd57265f25c4f4e8d02c
+size 12258994

data/issues.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8bb02ea7cd2de5290825dddd91b839bd43ae21c8d70c3462da77c53e5f1f2054
+size 12173666

data/issues.html

@@ -0,0 +1,75 @@
+
+<style>
+:root { color-scheme: light; }
+.bi-wrap { font-family: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; color: #1f2328 !important; }
+.bi-wrap h1, .bi-wrap h2, .bi-wrap h3, .bi-wrap p, .bi-wrap strong, .bi-wrap td, .bi-wrap th, .bi-wrap div { color: #1f2328 !important; }
+.bi-band { border: 1px solid #d8dee4; border-radius: 8px; background: #ffffff !important; padding: 16px; margin: 0 0 12px; }
+.bi-title { display: flex; gap: 12px; align-items: flex-start; justify-content: space-between; margin-bottom: 12px; }
+.bi-title h1, .bi-title h2 { margin: 0; line-height: 1.15; letter-spacing: 0; }
+.bi-title h1 { font-size: 1.55rem; }
+.bi-title h2 { font-size: 1.12rem; }
+.bi-title p { margin: 4px 0 0; color: #57606a; max-width: 820px; }
+.status-pill { border-radius: 999px; padding: 5px 9px; font-size: .78rem; font-weight: 700; background: #dafbe1; color: #116329 !important; white-space: nowrap; }
+.metric-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(138px, 1fr)); gap: 8px; }
+.metric { border: 1px solid #d0d7de; border-radius: 8px; padding: 10px; background: #f6f8fa !important; min-height: 86px; }
+.metric-value { font-size: 1.45rem; font-weight: 800; line-height: 1.1; color: #0969da !important; }
+.metric-label { margin-top: 4px; font-size: .84rem; font-weight: 700; }
+.metric-note { margin-top: 2px; font-size: .76rem; color: #57606a !important; }
+.callout-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 8px; }
+.callout { border-left: 4px solid #0969da; background: #f6f8fa !important; border-radius: 8px; padding: 10px 12px; }
+.callout:nth-child(2) { border-left-color: #1a7f37; }
+.callout:nth-child(3) { border-left-color: #bf8700; }
+.callout:nth-child(4) { border-left-color: #cf222e; }
+.callout strong { display: block; margin-bottom: 4px; }
+.group-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(310px, 1fr)); gap: 10px; }
+.group-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.group-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.meta-row { display: flex; flex-wrap: wrap; gap: 6px; margin: 6px 0; }
+.chip-row { display: inline-flex; flex-wrap: wrap; gap: 5px; align-items: center; vertical-align: middle; }
+.chip { display: inline-flex; align-items: center; border: 1px solid #d0d7de; border-radius: 999px; padding: 2px 7px; font-size: .78rem; background: #f6f8fa !important; color: #24292f !important; }
+.chip.bucket { border-color: #9a6700; background: #fff8c5 !important; }
+.chip.kind { border-color: #0969da; background: #ddf4ff !important; }
+.workflow-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(260px, 1fr)); gap: 10px; }
+.workflow-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.workflow-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.workflow-card .move-count { font-weight: 800; color: #0969da !important; }
+.move-list { display: grid; gap: 10px; }
+.move-row { border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+.move-row summary { list-style: none; cursor: pointer; padding: 12px; }
+.move-row summary::-webkit-details-marker { display: none; }
+.move-summary { display: grid; grid-template-columns: minmax(190px, 1fr) minmax(260px, 1.45fr) minmax(260px, 1.45fr) minmax(150px, .8fr) minmax(250px, 1.35fr) auto; gap: 16px; align-items: start; }
+.move-title, .move-field { min-width: 0; }
+.move-title h3 { margin: 0 0 4px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.move-field span { display: block; color: #57606a !important; font-size: .72rem; font-weight: 800; text-transform: uppercase; letter-spacing: 0; }
+.move-field p { margin: 3px 0 0; font-size: .84rem; line-height: 1.35; overflow-wrap: anywhere; }
+.move-field .chip, .move-title { overflow-wrap: anywhere; }
+.view-bundles { align-self: center; justify-self: end; border: 1px solid #000000; border-radius: 999px; color: #000000 !important; background: #ffffff !important; padding: 4px 9px; font-size: .78rem; font-weight: 800; white-space: nowrap; }
+.move-row[open] .view-bundles { background: #f6f8fa !important; }
+.move-expanded { border-top: 1px solid #d8dee4; padding: 0 12px 12px; }
+.bundle-row { border-top: 1px solid #eef1f4; padding: 12px 0; }
+.bundle-row:first-child { border-top: none; }
+.bundle-row h4 { margin: 0 0 6px; font-size: .94rem; line-height: 1.25; letter-spacing: 0; }
+.bundle-scope { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 7px; margin-top: 8px; }
+.evidence-list { margin: 8px 0 0 18px; padding: 0; color: #57606a !important; font-size: .82rem; line-height: 1.35; }
+.evidence-list li { margin: 3px 0; }
+.muted { color: #57606a !important; }
+.evidence { font-size: .84rem; color: #57606a !important; }
+.scroll-table { overflow-x: auto; border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+table.bi-table { width: 100%; border-collapse: collapse; font-size: .86rem; }
+.bi-table th { text-align: left; background: #f6f8fa !important; color: #57606a !important; font-size: .78rem; border-bottom: 1px solid #d8dee4; padding: 8px; white-space: nowrap; }
+.bi-table td { border-bottom: 1px solid #eef1f4; padding: 8px; vertical-align: top; min-width: 86px; }
+.bi-table tr:last-child td { border-bottom: none; }
+.wide { min-width: 280px; }
+.tight { max-width: 180px; }
+a { color: #0969da !important; text-decoration: none; }
+a:hover { text-decoration: underline; }
+.bi-wrap, .bi-wrap * { color: #000000 !important; }
+@media (max-width: 720px) {
+  .bi-title { display: block; }
+  .status-pill { display: inline-block; margin-top: 8px; }
+  .group-grid { grid-template-columns: 1fr; }
+  .move-summary { grid-template-columns: 1fr; }
+  .view-bundles { justify-self: start; }
+}
+</style>
+<div class='bi-wrap'><section class='bi-band'><div class='bi-title'><div><h2>Open Issues</h2><p>Public-safe normalized rows for auditability.</p></div></div><div class='scroll-table'><table class='bi-table'><thead><tr><th>Issue</th><th>Bucket</th><th>Score</th><th>Linked</th><th>Labels</th><th>Excerpt</th></tr></thead><tbody><tr><td><a href='https://github.com/pytorch/pytorch/issues/180948' target='_blank' rel='noopener noreferrer'>Issue #180948: DISABLED test_emulate_precision_casts_preserves_explicit_precision_cast_float16 (__main__.CudaReproTests)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: tests, triaged, skipped, oncall: pt2, module: inductor, module: xpu, bot-triaged</td><td><div class='wide'>Platforms: xpu This test was disabled because it is failing on main branch (recent examples). cc @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @kadeng @muchulee8 @amjames @aakhundov @coconutruben @jataylo @gujinghui @fe...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189322' target='_blank' rel='noopener noreferrer'>Issue #189322: DISABLED test_aoti_debug_printer_codegen_cuda (__main__.AOTInductorTestDualWrapper)</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: flaky-tests, skipped, oncall: pt2, module: inductor</td><td><div class='wide'>Platforms: inductor This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 3 workflow(s) with 3 failures and 3 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSUME...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189323' target='_blank' rel='noopener noreferrer'>Issue #189323: DISABLED test_fake_export_contiguous_cpu_float32 (__main__.TestExportOpInfoCPU)</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: flaky-tests, skipped, oncall: pt2, oncall: export</td><td><div class='wide'>Platforms: linux This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 7 workflow(s) with 7 failures and 7 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSUME THI...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/119537' target='_blank' rel='noopener noreferrer'>Issue #119537: DISABLED test_source_multithreaded_open_in_scope_work_in_main_thread_False (__main__.TestProfiler)</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: flaky-tests, skipped, oncall: profiler</td><td><div class='wide'>Platforms: dynamo, linux, mac, macos, rocm, slow, asan This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 3 hours, it has been determined flaky in 4 workflow(s) with 12 failures and 4 successes. Debugging instructions (after clicking on the...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/119576' target='_blank' rel='noopener noreferrer'>Issue #119576: DISABLED test_source_multithreaded_multiple_preexisting_work_in_main_thread_True (__main__.TestProfiler)</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: flaky-tests, skipped, oncall: profiler</td><td><div class='wide'>Platforms: linux, mac, macos, rocm, asan, slow This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 3 hours, it has been determined flaky in 7 workflow(s) with 21 failures and 7 successes. Debugging instructions (after clicking on the recent s...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/180194' target='_blank' rel='noopener noreferrer'>Issue #180194: DISABLED test_vjp_grad_cpu (__main__.TestComposabilityCPU)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, skipped, oncall: pt2, module: functorch</td><td><div class='wide'>Platforms: dynamo This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 3 workflow(s) with 3 failures and 3 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSUME TH...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/174507' target='_blank' rel='noopener noreferrer'>Issue #174507: PyTorch Stable ABI support</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>feature, module: abi, triaged, bot-triaged</td><td><div class='wide'>🐛 Describe the bug Hello @janeyx99 I am one of owners of https://github.com/amd/quark and I am super excited in converting our runtime-compiled Torch C++ extensions into pre-compiled libraries shipped with Quark using the PyTorch&#x27;s stable ABI When I checked our code that uses torch C++, I have identified APIs that i...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189164' target='_blank' rel='noopener noreferrer'>Issue #189164: [CI][B200] B200 Runner Maintenance Notification</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: ci, triaged, module: infra, bot-triaged</td><td><div class='wide'>We are performing B200 runner maintenance one by one for the six runners. Jobs may encounter failures during this time and may need to be restarted. cc @malfet @pytorch/pytorch-dev-infra @huydhn @atalman @ptrblck @eqy @tinglvv</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189194' target='_blank' rel='noopener noreferrer'>Issue #189194: [ROCm] torch 2.13 wheel: &quot;Can&#x27;t detect vectorized ISA for CPU&quot; in torch.compile smoke test on non-ROCm image (regression vs 2.12.1)</a></td><td>needs design decision</td><td>115</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>high priority, triage review, module: binaries, module: rocm, triaged, module: vectorization, module: regression, oncall: pt2, module: inductor, bot-triaged</td><td><div class='wide'>🐛 Describe the bug The torch==2.13.0+rocm7.2 wheel fails the torch.compile smoke test with RuntimeError: Can&#x27;t detect vectorized ISA for CPU when it is imported/run on a non-ROCm image (e.g. pytorch/almalinux-builder:cpu-main), where torch.cuda.is_available() is False and the compile check exercises the CPU / Induct...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189308' target='_blank' rel='noopener noreferrer'>Issue #189308: `torch.distributed.checkpoint`: `.metadata` loaded with raw `pickle.load` on the public `load()` path — same class you already warn about on `torch.load`</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>oncall: distributed, oncall: distributed checkpointing, bot-triaged</td><td><div class='wide'>🐛 Describe the bug I first reported this privately as a security advisory (GHSA-ppvg-47fv-5f4w). It was closed under the SECURITY.md policy. In that thread I asked whether it might still be worth tracking as a documentation/API-hardening issue and whether you would prefer that I open a regular issue; I did not recei...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189309' target='_blank' rel='noopener noreferrer'>Issue #189309: [Stable ABI] Add a has_side_effects op tag. DCE silently deletes side-effect-only op calls</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>oncall: jit, bot-triaged</td><td><div class='wide'>🚀 The feature, motivation and pitch While porting torchvision to stable ABI, I noticed that ops that return nothing and exist only for their side effects cannot be registered safely. StableLibrary::def only takes a schema string so the op gets FROM_SCHEMA alias analysis, and TorchScript DCE deletes calls to an op wi...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/150199' target='_blank' rel='noopener noreferrer'>Issue #150199: `asarray`: device does not propagate from input to output after `set_default_device`</a></td><td>has linked PR</td><td>44</td><td><span class='chip-row'><span class='chip'>PR #188795</span></span></td><td>triaged, module: python array api, module: python frontend</td><td><div class='wide'>🐛 Describe the bug The documentation of asarray states: device (torch.device, optional) – the device of the returned tensor. Default: None, which causes the device of obj to be used. Or, if obj is a Python sequence, the current default device will be used. The described behaviour is coherent with the specification o...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/141287' target='_blank' rel='noopener noreferrer'>Issue #141287: MPS operator coverage tracking issue (2.6+ version)</a></td><td>has linked PR</td><td>44</td><td><span class='chip-row'><span class='chip'>PR #178031</span></span></td><td>feature, triaged, tracker, module: mps</td><td><div class='wide'>🐛 Describe the bug This issue is to have a centralized place to list and track work on adding support to new ops for the MPS backend. Most requested ops (extracted from comments to this issue and #77764) PyTorch MPS Ops Project : Project to track all the ops for MPS backend. There are a very large number of operator...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189298' target='_blank' rel='noopener noreferrer'>Issue #189298: torch.jit.trace fails with MaxUnpool in 2.13</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>oncall: jit, bot-triaged</td><td><div class='wide'>🐛 Describe the bug Reproducible Example import torch def main(): pool = torch.nn.MaxPool2d(2, return_indices=True) out, indices = pool(torch.randn(1, 10, 10, 10)) unpool = torch.nn.MaxUnpool2d(2) _ = torch.jit.trace(unpool, (out, indices)) if __name__ == &quot;__main__&quot;: main() error message: TypeError: cond must be a bo...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189293' target='_blank' rel='noopener noreferrer'>Issue #189293: AttributeError: Can&#x27;t get local object &#x27;WeakValueDictionary.__init__.&lt;locals&gt;.remove&#x27; during AOTAutograd graph caching on Python 3.13</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triage review, module: pickle, tensor subclass, oncall: pt2, module: aotdispatch, module: pt2-dispatcher, module: python version, bot-triaged</td><td><div class='wide'>🐛 Describe the bug When utilizing torch.compile with graph caching enabled (TORCHINDUCTOR_FX_GRAPH_CACHE=1) on Python 3.13 and PyTorch 2.14 Nightly, the compilation disk cache subsystem crashes during the serialization phase of tensor subclass metadata. Specifically, when a custom tensor subclass (such as an advance...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189150' target='_blank' rel='noopener noreferrer'>Issue #189150: SIGSEGV / cudaErrorIllegalAddress when replaying a CUDA graph that captured multiple training iterations of two trainers over deeply unrolled recurrent modules</a></td><td>needs design decision</td><td>115</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triage review, module: crash, module: cuda, module: cuda graphs, bot-triaged</td><td><div class='wide'>🐛 Describe the bug Capturing N training iterations (forward + backward + AdamW(capturable=True).step()) of TWO independent trainers into one torch.cuda.CUDAGraph produces a graph whose replay crashes. In the script below each trainer owns M=4 plain nn.Modules called in a Python loop - no functorch anywhere - and the...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188898' target='_blank' rel='noopener noreferrer'>Issue #188898: DISABLED test_vmap_over_vmap_two_inputs (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, skipped, oncall: pt2, module: inductor</td><td><div class='wide'>Platforms: linux, rocm, slow This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 3 workflow(s) with 6 failures and 3 successes. Debugging instructions (after clicking on the recent samples link): DO NO...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/119526' target='_blank' rel='noopener noreferrer'>Issue #119526: DISABLED test_source_multithreaded_multiple_preexisting_work_in_main_thread_False (__main__.TestProfiler)</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: flaky-tests, skipped, oncall: profiler</td><td><div class='wide'>Platforms: dynamo, linux, mac, macos, rocm, slow, asan This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 3 hours, it has been determined flaky in 8 workflow(s) with 24 failures and 8 successes. Debugging instructions (after clicking on the...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189287' target='_blank' rel='noopener noreferrer'>Issue #189287: [vllm] vllm-x-pytorch-test multimodal pooling tests fail with EngineCore &quot;CUDA driver initialization failed&quot; since Triton 3.8 bump (#188251)</a></td><td>high-priority/blocker</td><td>70</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: cuda, module: ci, module: tests, triaged, module: regression, oncall: pt2, module: vllm, bot-triaged</td><td><div class='wide'>Job vllm-test / vllm-x-pytorch-test / test (vllm_multi_model_test_28_failure_test, 1, 1, linux.g6.4xlarge.experimental.nvidia.gpu, unstable) Symptom 13 failed / 23 passed / 29 skipped. All 13 failures are vLLM multimodal pooling model tests, and every one dies the same way: vLLM EngineCore fails to start with Runtim...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189282' target='_blank' rel='noopener noreferrer'>Issue #189282: [CPU] operator_benchmark: backward-pass (add, + batchnorm on aarch64) ~200-500x slower since ~Nov 2025 (x86 + aarch64; possible measurement artifact)</a></td><td>needs design decision</td><td>115</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triage review, module: performance, module: autograd, module: cpu, module: regression, bot-triaged</td><td><div class='wide'>Possible / low-confidence. add_M64_N64_K*_cpu_bwd*_BACKWARD, 9 cases, ~200-400x slower vs the prior baseline, stable across runs. Flagged as low-confidence: a real 400x regression in add backward would be catastrophic and caught in training, so this looks more like a backward-pass measurement artifact than a genuine...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189281' target='_blank' rel='noopener noreferrer'>Issue #189281: [CPU] operator_benchmark: embedding / embeddingbag ~70-130x slower since ~Nov 2025 (x86 + aarch64; possible measurement issue)</a></td><td>needs design decision</td><td>115</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triage review, module: performance, module: cpu, module: regression, op-bench, module: embedding, bot-triaged</td><td><div class='wide'>Possible regression. embedding and embeddingbag, ~48 cases, ~70-90x slower vs the prior baseline, stable across runs. Flagged as possible because embeddingbag time is roughly flat vs bag count (10 vs 2300 bags both ~8500us), which suggests the number may be dominated by a fixed per-call cost rather than the embeddin...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189283' target='_blank' rel='noopener noreferrer'>Issue #189283: [CPU] operator_benchmark: isolated ~2x regressions (index_add, baddbmm) since ~Nov 2025</a></td><td>high-priority/blocker</td><td>70</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: performance, module: cpu, triaged, module: regression, op-bench, bot-triaged</td><td><div class='wide'>Possible / low-confidence. Two isolated single-shape cases flagged by the op-benchmark CPU rebaseline, each stable across two runs but not part of a broader family pattern. index_add__M8_N32_K1_dim0_cpu_dtypetorch.float32 -- ~2.4x baddbmm_B128_M64_N32_K64_cpu_dtypetorch.float32 -- ~2.0x. Note: other baddbmm shapes a...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189280' target='_blank' rel='noopener noreferrer'>Issue #189280: [CPU][quant] operator_benchmark: qcat ~3-4x slower (likely) and fake_quantize ~700x slower (possible) since ~Nov 2025</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>oncall: quantization, bot-triaged</td><td><div class='wide'>Two quantization families flagged by the rebaseline, both stable across runs. qcat (quantized cat), 18 cases, ~3-4x slower -- likely regression. fake_quantize per-tensor/channel (original_kernel/learnable_kernel), ~700x slower -- possible regression or measurement artifact; the magnitude is extreme and needs an owne...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188833' target='_blank' rel='noopener noreferrer'>Issue #188833: [pipelining] Interleaved1F1B warmup head-of-line bubble from front-loaded RECV</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>oncall: distributed</td><td><div class='wide'>🐛 Describe the bug TL;DR On the _PipelineScheduleRuntime path (which ScheduleInterleaved1F1B now uses), the communication lowering places each RECV_F at the position where its sender is scheduled by a dependency simulation, not at the timestep where the data is actually produced/needed. https://github.com/pytorch/py...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189279' target='_blank' rel='noopener noreferrer'>Issue #189279: [CPU] operator_benchmark: eager transcendental unary (cosh/sinh/atan/log2/log10) ~3-5x slower since ~Nov 2025</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td></td><td><div class='wide'>Likely regression. 8 cases of eager transcendental unary math are ~3-5x slower vs the prior baseline, stable across two runs. Suspected: the vectorized/Sleef CPU math path. Please confirm. sinh_M512_N512_cpu old 81.5 -&gt; 432.0 / 388.9 (~5x) log10__M512_N512_cpu old 11.8 -&gt; 46.2 / 45.1 (~4x) log2__M512_N512_cpu old 11...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189262' target='_blank' rel='noopener noreferrer'>Issue #189262: DISABLED rocm-nightly / linux-noble-rocm-nightly-py3.12-gfx942 / test</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: rocm, module: ci, triaged</td><td><div class='wide'>Reason MI300 cluster is having outages, hence capacity is reduced, leading to queueing on 4-GPU MI300 runners. cc @jeffdaily @sunway513 @pruthvistony @ROCmSupport @jataylo @hongxiayang @naromero77amd @pragupta @jerrymannil @xinyazhang @malfet @pytorch/pytorch-dev-infra</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/167822' target='_blank' rel='noopener noreferrer'>Issue #167822: [RFD] Composable Kernel (CK) moving to C++20 — request for feedback/concerns</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: build, triaged</td><td><div class='wide'>New Feature for Release Composable Kernel (CK) plans to upgrade its language standard from C++17 → C++20 in the next few months. This will change our default CMake from C++17 to C++20. We’d like to confirm there are no blockers for PyTorch builds and downstreams. Point(s) of contact ThomasNing Release Mode (pytorch/...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188756' target='_blank' rel='noopener noreferrer'>Issue #188756: [MPS] index_select / F.embedding silently return all-zero rows when the source tensor exceeds 2**31 elements (fixed on main, still broken in release/2.13)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: 64-bit, module: correctness (silent), module: scatter &amp; gather ops, module: embedding, module: mps, bot-triaged</td><td><div class='wide'>🐛 Describe the bug On MPS, F.embedding / nn.Embedding / torch.index_select silently return all-zero rows (no error, no NaN) for every index i &gt;= ceil(2**31 / dim) when the source weight tensor has more than 2**31 elements. Indices below that boundary return bit-exact rows, so the corruption is partial and easy to mi...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188836' target='_blank' rel='noopener noreferrer'>Issue #188836: DISABLED test_dropout_parity_dynamic_shapes (__main__.TestDropoutAlignRandomEager)</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>skipped, oncall: pt2</td><td><div class='wide'>This test was disabled because it is failing on main branch (recent examples). cc @chauhang @penguinwu</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188896' target='_blank' rel='noopener noreferrer'>Issue #188896: DISABLED test_launch_elastic_worker_raise_exception (__main__.ElasticLaunchTest)</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>oncall: distributed, module: flaky-tests, skipped, oncall: r2p</td><td><div class='wide'>Platforms: linux This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 3 workflow(s) with 0 failures and 3 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSUME THI...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189267' target='_blank' rel='noopener noreferrer'>Issue #189267: nn.functional.threshold OpInfo reference raises OverflowError on unsigned dtypes with NumPy &gt;= 2.5</a></td><td>has linked PR</td><td>94</td><td><span class='chip-row'><span class='chip'>PR #189273</span></span></td><td>module: nn, module: tests, triaged, module: numpy, module: regression, module: unsigned int, bot-triaged</td><td><div class='wide'>🐛 Describe the bug The nn.functional.threshold OpInfo reference in torch/testing/_internal/common_methods_invocations.py fails for unsigned dtypes under NumPy &gt;= 2.5. test_reference_numerics_{small,normal}_nn_functional_threshold_{cpu,cuda}_uint8 (and the _refs variants) fail with: OverflowError: Python integer -9 o...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189245' target='_blank' rel='noopener noreferrer'>Issue #189245: [cpp_extension] Stale .lock file from FileBaton causes permanent deadlock when compilation process is killed</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triage review, module: cpp-extensions, module: deadlock, bot-triaged</td><td><div class='wide'>🐛 Describe the bug If the process holding the lock is forcefully terminated (e.g., by OOM Killer, SIGKILL, Ctrl+C, or Slurm job cancellation), the .lock file is never cleaned up. On subsequent runs, all processes detect the stale .lock file and enter an infinite wait loop, causing a permanent deadlock. To Reproduce...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189271' target='_blank' rel='noopener noreferrer'>Issue #189271: [inductor] KeyError in Scheduler.compute_ancestors: self-edge from realizing a reshaped in-place-mutated buffer</a></td><td>has linked PR</td><td>44</td><td><span class='chip-row'><span class='chip'>PR #189288</span></span></td><td>triaged, has workaround, module: custom-operators, module: viewing and reshaping, oncall: pt2, module: dynamic shapes, module: inductor, module: pt2-dispatcher, bot-triaged</td><td><div class='wide'>🐛 Describe the bug torch.compile fails with InductorError: KeyError in Scheduler.compute_ancestors when a custom op mutates freshly-allocated (torch.empty) output buffers in place and those buffers are then reshaped in a way Inductor cannot prove is a free view (under dynamic shapes). This is the shape of a flash_at...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189270' target='_blank' rel='noopener noreferrer'>Issue #189270: [inductor] Emit per-kernel physical op structure (with rematerialization) in provenance</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, enhancement, oncall: pt2, module: inductor, module: trace, bot-triaged</td><td><div class='wide'>Summary kernel_information.json (#160540, extended in #183952) and the profiler-timeline provenance (#186230, #187328) map each generated kernel to a set of pre/post-grad node names, stack traces, and now extern I/O shapes. What is still missing is the physical, per-kernel op structure: the ordered list of ops that...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/177952' target='_blank' rel='noopener noreferrer'>Issue #177952: Fix onnx tests running on GPU/CUDA</a></td><td>needs reproduction</td><td>32</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>needs reproduction, module: onnx, triaged, OSS contribution wanted, bot-triaged</td><td><div class='wide'>🐛 Describe the bug cd test/onnx python test_pytorch_onnx_onnxruntime_cuda.py -k test_arithmetic_bfp16 onnxruntime.capi.onnxruntime_pybind11_state.NotImplemented: [ONNXRuntimeError] : 9 : NOT_IMPLEMENTED : Could not find an implementation for Add(14) node with name &#x27;/Add&#x27; Versions torch 2.7.1 onnx 1.17.0 onnxruntime...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188496' target='_blank' rel='noopener noreferrer'>Issue #188496: torch.fft.irfft2 ONNX export produces graph that onnxruntime cannot execute.</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: onnx, triaged, has workaround, module: fft, bot-triaged</td><td><div class='wide'>🐛 Describe the bug torch.fft.irfft2 ONNX export now succeeds on nightly (fixed in #164698), but the exported graph cannot be run by onnxruntime in practice. PyTorch&#x27;s dynamo exporter currently produces ONNX IR version 10 graphs. The newest onnxruntime release that supports CUDA 11.8 is onnxruntime-gpu==1.16.3, which...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189261' target='_blank' rel='noopener noreferrer'>Issue #189261: DISABLED periodic-rocm-mi300 / linux-noble-rocm-py3.12-mi300 / test</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: rocm, module: ci, triaged, bot-triaged</td><td><div class='wide'>Reason MI300 cluster is having outages, hence capacity is reduced, leading to queueing on 4-GPU MI300 runners. cc @jeffdaily @sunway513 @pruthvistony @ROCmSupport @jataylo @hongxiayang @naromero77amd @pragupta @jerrymannil @xinyazhang @malfet @pytorch/pytorch-dev-infra</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/129131' target='_blank' rel='noopener noreferrer'>Issue #129131: Torch compile initialises CUDA context, even compiling CPU functions</a></td><td>has linked PR</td><td>44</td><td><span class='chip-row'><span class='chip'>PR #186403</span></span></td><td>triaged, oncall: pt2, module: dynamo, module: pt2-dispatcher</td><td><div class='wide'>🐛 Describe the bug When trying to compile a simple function that uses cpu tensors, torch inductor initialises a context on cuda:0. If used in a multiprocessing context (E.G.) a torch dataloader, this quickly results in OOM (simple repro example allocates 500MB of VRAM per process). CUDA_VISIBLE_DEVICES=&quot;&quot; works as a...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189255' target='_blank' rel='noopener noreferrer'>Issue #189255: DISABLED test_delayed_optim_step_offload_true_no_shard_cuda (__main__.TestParityWithDDPCUDA)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>oncall: distributed, module: rocm, triaged, module: flaky-tests, skipped</td><td><div class='wide'>Platforms: rocm This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 3 workflow(s) with 0 failures and 3 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSUME THIN...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/119536' target='_blank' rel='noopener noreferrer'>Issue #119536: DISABLED test_source_multithreaded_complex_work_in_main_thread_True (__main__.TestProfiler)</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: flaky-tests, skipped, oncall: profiler</td><td><div class='wide'>Platforms: asan, linux, mac, macos, rocm, slow This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 3 hours, it has been determined flaky in 14 workflow(s) with 42 failures and 14 successes. Debugging instructions (after clicking on the recent...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/179799' target='_blank' rel='noopener noreferrer'>Issue #179799: DISABLED test_grad_has_aux (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, skipped, oncall: pt2, module: inductor</td><td><div class='wide'>Platforms: linux, slow, rocm This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 6 workflow(s) with 12 failures and 6 successes. Debugging instructions (after clicking on the recent samples link): DO N...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/160740' target='_blank' rel='noopener noreferrer'>Issue #160740: MPS torch.argmax/torch.argmin fail for non-contiguous input</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: correctness (silent), module: mps</td><td><div class='wide'>🕵️‍♂️ Detected with FACTO 🐛 Describe the bug As the title says: import torch def test_argmax(x): device = x.device try: output = torch.argmax(x) print(f&quot; argmax test succeeds for device: {device}. output: {output}&quot;) except Exception as e: print(f&quot; argmax test fails for device: {device}: {e}&quot;) x_cpu_contiguous = torc...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/143954' target='_blank' rel='noopener noreferrer'>Issue #143954: No ONNX function found for &lt;OpOverload(op=&#x27;quantized_decomposed.dequantize_per_channel&#x27;, overload=&#x27;default&#x27;)&gt;</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: onnx, triaged, OSS contribution wanted</td><td><div class='wide'>🐛 Describe the bug We tried to leverage per_channel quantization in QAT and exported the trained model in onnx format. model = dummy pytorch model export_model = torch.export.export_for_training( model, example_inputs).module() quantizer = XNNPACKQuantizer().set_global(get_symmetric_quantization_config( is_per_chann...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/187455' target='_blank' rel='noopener noreferrer'>Issue #187455: [MPS] Migrate remaining MPSGraph-backed ops to native Metal to bound graph-cache memory</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: performance, module: memory usage, triaged, enhancement, module: mps, bot-triaged</td><td><div class='wide'>Several MPS ops still dispatch through MPSGraph (LookUpOrCreateCachedGraph), which compiles and caches a graph per input shape with no eviction. On shape-varying workloads -- GNNs, variable-length attention, hyperparameter sweeps -- the cache grows without bound and the process eventually OOMs on Apple Silicon. (tor...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/168995' target='_blank' rel='noopener noreferrer'>Issue #168995: MXFP8 + torch.compile error: NotImplementedError: make_reindexer NYI on DtypeView</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, oncall: pt2, module: inductor</td><td><div class='wide'>🐛 Describe the bug Using the latest pytorch nightly, torch.compile throws an error with torchao mxfp8 MoE training code. Repro Use machine with access to local B200 GPUs Clone https://github.com/pytorch/ao Install from source: USE_CUDA=1 pip install -e . --no-build-isolation Run benchmark script: rm -rf /tmp/torchin...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189244' target='_blank' rel='noopener noreferrer'>Issue #189244: DISABLED test_cat_pointwise (__main__.NumBytesMetricTests)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: rocm, triaged, module: flaky-tests, skipped, oncall: pt2, module: inductor</td><td><div class='wide'>Platforms: rocm This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 4 workflow(s) with 4 failures and 4 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSUME THIN...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189223' target='_blank' rel='noopener noreferrer'>Issue #189223: DISABLED test_vmap_free_tensor (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, skipped, oncall: pt2, module: inductor</td><td><div class='wide'>Platforms: linux, slow This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 4 workflow(s) with 8 failures and 4 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSU...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/119490' target='_blank' rel='noopener noreferrer'>Issue #119490: DISABLED test_source_multithreaded_complex_work_in_main_thread_False (__main__.TestProfiler)</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: flaky-tests, skipped, oncall: profiler</td><td><div class='wide'>Platforms: dynamo, asan, linux, mac, macos, rocm, slow This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 3 hours, it has been determined flaky in 4 workflow(s) with 12 failures and 4 successes. Debugging instructions (after clicking on the...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/187289' target='_blank' rel='noopener noreferrer'>Issue #187289: DISABLED test_col2im_uint32_overflow (__main__.TestMPS)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, module: macos, skipped, module: mps</td><td><div class='wide'>Platforms: mac, macos This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 3 workflow(s) with 4 failures and 3 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSUM...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/152632' target='_blank' rel='noopener noreferrer'>Issue #152632: DISABLED test_torchvision_models_efficientnet_v2_l (__main__.TestVisionTracing)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, skipped, module: fx</td><td><div class='wide'>Platforms: asan, linux, slow This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 3 hours, it has been determined flaky in 3 workflow(s) with 3 failures and 3 successes. Debugging instructions (after clicking on the recent samples link): DO NO...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/179879' target='_blank' rel='noopener noreferrer'>Issue #179879: DISABLED test_vmap_vmap_cpu (__main__.TestComposabilityCPU)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, skipped, oncall: pt2, module: functorch</td><td><div class='wide'>Platforms: dynamo This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 12 workflow(s) with 12 failures and 12 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSUME...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189239' target='_blank' rel='noopener noreferrer'>Issue #189239: [Inductor]Unable to apply layout optimization on convolution backward causes training performance regression.</a></td><td>needs design decision</td><td>115</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triage review, module: performance, module: convolution, module: regression, module: memory format, oncall: pt2, module: inductor, bot-triaged</td><td><div class='wide'>🐛 Describe the bug decide_layout_opt() only counts aten.convolution.default nodes when deciding whether to enable channels-last layout optimization. When AOTAutograd splits a model into separate forward and backward graphs, the backward graph may contain only convolution_backward without a corresponding forward conv...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189236' target='_blank' rel='noopener noreferrer'>Issue #189236: [Feature Request] Add &#x27;symmetric&#x27; padding_mode (repeats boundary sample)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: nn, triaged, function request, module: padding, bot-triaged</td><td><div class='wide'>🚀 The feature, motivation and pitch PyTorch&#x27;s current &#x27;reflect&#x27; padding_mode uses half-sample reflection, i.e. it does not repeat the boundary sample: input: ABCDEF reflect: DCB|ABCDEF|EDC (boundaries not repeated) I&#x27;d like to add a &#x27;symmetric&#x27; padding_mode that does repeat the boundary sample: input: ABCDEF symmetr...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/185405' target='_blank' rel='noopener noreferrer'>Issue #185405: [Inductor] AssertionError: len(index) == len(stride) in ir.py when using torch.compile(mode=&quot;max-autotune&quot;) on CPU BMM/GEMM</a></td><td>high-priority/blocker</td><td>70</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: crash, module: cpu, triaged, module: complex, oncall: pt2, module: inductor, bot-triaged</td><td><div class='wide'>🐛 Describe the bug When compiling an attention-like function containing complex numbers (torch.polar, torch.view_as_complex) and batched matrix multiplication (torch.matmul) on CPU with torch.compile(mode=&quot;max-autotune&quot;), the compiler crashes with an AssertionError. The crash occurs during the C++ backend&#x27;s BMM/GEMM...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/187175' target='_blank' rel='noopener noreferrer'>Issue #187175: [feature request] Autotune TMA and non-TMA path for pointwise and reduction kernels</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: performance, feature, module: cuda, triaged, oncall: pt2, module: inductor, bot-triaged</td><td><div class='wide'>🚀 The feature, motivation and pitch Summary When autotuning is enabled for GEMM kernels, both the TMA (Tensor Memory Accelerator) path and the non-TMA path are included as candidates and benchmarked against each other. This ensures the fastest path is always selected based on actual measured performance. However, fo...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189228' target='_blank' rel='noopener noreferrer'>Issue #189228: [symm_mem] one/two-shot all-reduce deadlocks when issued from multiple streams: ops lack channel plumbing</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>oncall: distributed, bot-triaged, oncall: distributed infra</td><td><div class='wide'>🐛 Describe the bug symm_mem::two_shot_all_reduce_ (and the one-shot / multimem variants) synchronize the group with barriers that require the device execution order of the ops to match their group issue order. CUDA does not order kernels across streams, so issuing these ops from more than one stream — combined with...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/147326' target='_blank' rel='noopener noreferrer'>Issue #147326: torch.export.export fails when one input is a class inheriting from torch.nn.Module</a></td><td>needs design decision</td><td>89</td><td><span class='chip-row'><span class='chip'>PR #185804</span></span></td><td>oncall: pt2, export-triaged, oncall: export</td><td><div class='wide'>🐛 Describe the bug transformers.cache_utils.DynamicCache inherits from torch.nn.Module. It seems to confuse torch.export.export and gives the following error: File &quot;.../site-packages/torch/export/_trace.py&quot;, line 1697, in _export_to_aten_ir_make_fx raise UserError(UserErrorType.CONSTRAINT_VIOLATION, str(e)) # noqa:...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189144' target='_blank' rel='noopener noreferrer'>Issue #189144: [XPU][B580] `flex-attn-causal` performance drop with `last_level_cache_size` cache clear between runs</a></td><td>needs design decision</td><td>115</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triage review, module: performance, module: regression, has workaround, oncall: pt2, module: inductor, module: xpu, module: flex attention, bot-triaged</td><td><div class='wide'>This PR #184567 introduced regression in a few cases for flex-attn-causal with triton, for example: H_q=16, H_kv=2, N_CTX_q=1, N_CTX_kv=1024, D_HEAD_qk=128, D_HEAD_v=128 case dropped from 0.81 -&gt; 0.41 TFLOPS More details: intel/intel-xpu-backend-for-triton#7302 TL;DR; 18MB of cache clear between runs seems like not...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/187811' target='_blank' rel='noopener noreferrer'>Issue #187811: DISABLED test_combo_kernel_dynamic_scale_rblock (__main__.ComboKernelTests)</a></td><td>has linked PR</td><td>44</td><td><span class='chip-row'><span class='chip'>PR #189224</span></span></td><td>triaged, skipped, module: xpu</td><td><div class='wide'>Platforms: xpu This test was disabled because it is failing on main branch (recent examples). cc @gujinghui @EikanWang @fengyuan14 @guangyey cc @gujinghui @EikanWang @fengyuan14 @guangyey</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/187810' target='_blank' rel='noopener noreferrer'>Issue #187810: DISABLED test_combo_kernel_dynamic_scale_rblock (__main__.ComboKernelTestsPerSubkernelBlocks)</a></td><td>has linked PR</td><td>44</td><td><span class='chip-row'><span class='chip'>PR #189224</span></span></td><td>triaged, skipped, module: xpu</td><td><div class='wide'>Platforms: xpu This test was disabled because it is failing on main branch (recent examples). cc @gujinghui @EikanWang @fengyuan14 @guangyey</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/163650' target='_blank' rel='noopener noreferrer'>Issue #163650: [DO NOT CLOSE] Autorevert actions shadow mode stream</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged</td><td><div class='wide'>This PR can be subscribed by anyone who is interested. Autorevert bot should post here, as comments, all its Autorevert decisions while running in shadow mode. This is, anytime Autorevert decided to act to revert a PR (this means faster than a human). The goal of this PR is to enable the team to quickly troubleshoot...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189221' target='_blank' rel='noopener noreferrer'>Issue #189221: DISABLED test_nested_dict_partial_load (__main__.TestCheckpointer)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: rocm, triaged, module: flaky-tests, skipped, oncall: distributed checkpointing</td><td><div class='wide'>Platforms: rocm This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 3 workflow(s) with 0 failures and 3 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSUME THIN...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/119514' target='_blank' rel='noopener noreferrer'>Issue #119514: DISABLED test_source_multithreaded_close_in_scope_work_in_main_thread_True (__main__.TestProfiler)</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: flaky-tests, skipped, oncall: profiler</td><td><div class='wide'>Platforms: asan, linux, mac, macos, rocm, slow This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 3 hours, it has been determined flaky in 26 workflow(s) with 78 failures and 26 successes. Debugging instructions (after clicking on the recent...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/179742' target='_blank' rel='noopener noreferrer'>Issue #179742: DISABLED test_grad_freevar_python_scalar (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, skipped, oncall: pt2, module: inductor</td><td><div class='wide'>Platforms: linux, slow, rocm This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 9 workflow(s) with 18 failures and 9 successes. Debugging instructions (after clicking on the recent samples link): DO N...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/180213' target='_blank' rel='noopener noreferrer'>Issue #180213: DISABLED test_vmap_vjp_cpu (__main__.TestComposabilityCPU)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, skipped, oncall: pt2, module: functorch</td><td><div class='wide'>Platforms: dynamo This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 7 workflow(s) with 7 failures and 7 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSUME TH...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/76185' target='_blank' rel='noopener noreferrer'>Issue #76185: Ensure custom Function are correct in double backward setting</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: autograd, triaged, enhancement, actionable</td><td><div class='wide'>Problem statement Today, if intermediary Tensors are saved during the forward and used during the backward, the double backward will be silently wrong because their contribution will be ignored. A small repro is: import torch from torch.autograd import Function, gradcheck, gradgradcheck class MyFn(Function): @static...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/187073' target='_blank' rel='noopener noreferrer'>Issue #187073: torch.compile recompiles when input alternates between AsyncCollectiveTensor and Tensor</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: performance, oncall: distributed, triaged, has workaround, module: fsdp, oncall: pt2, module: dynamo, module: guards, bot-triaged, oncall: distributed parallelisms</td><td><div class='wide'>torch.compile recompiles when input alternates between AsyncCollectiveTensor and Tensor Summary When a compiled region&#x27;s input alternates between AsyncCollectiveTensor (ACT) and a resolved Tensor across steps, Dynamo guards on the tensor class and recompiles each time the class changes. With FSDP + TP this happens d...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189189' target='_blank' rel='noopener noreferrer'>Issue #189189: Subtraction between two `torch.complex64` tensors fails on modules compiled with `aoti_compile_and_package`</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: complex, has workaround, oncall: pt2, oncall: export, module: aotinductor, bot-triaged</td><td><div class='wide'>🐛 Describe the bug I am trying to export an nn.Module that involves manipulation of complex-valued tensors with torch.export.export and then compiling with torch._inductor.aoti_compile_and_package. Compilation succeeds but I run into errors when loading the package and running. In the following minimal reproducing e...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/186796' target='_blank' rel='noopener noreferrer'>Issue #186796: weakrefs don&#x27;t when tensor swap is used</a></td><td>has linked PR</td><td>44</td><td><span class='chip-row'><span class='chip'>PR #186801</span></span></td><td>triaged, oncall: pt2, module: dynamo, module: guards, bot-triaged</td><td><div class='wide'>🐛 Describe the bug I thought #180566 would fix it but I was wrong. A live torch.compile&#x27;d module keeps weakrefs on its parameters (Dynamo guards). swap_tensors() refuses to operate on tensors that have weakrefs. The follow should reproduce the issue: &quot;&quot;&quot;Minimal reproducer for why group offloading cannot use torch.ut...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/50282' target='_blank' rel='noopener noreferrer'>Issue #50282: Add MultiScheduler</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: optimizer, triaged, enhancement</td><td><div class='wide'>🚀 Feature I would like to have a MultiScheduler that allows me to use multiple schedulers after each other, with the same optimizer. Motivation It is not uncommon to have a learning rate schedule that looks like the following example: First have a burn in learning rate for 1000-2000 batches (eg. quadratic or exponen...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/167725' target='_blank' rel='noopener noreferrer'>Issue #167725: `torch._inductor.utils.fresh_cache` fails due to persistant process pool</a></td><td>has linked PR</td><td>44</td><td><span class='chip-row'><span class='chip'>PR #184007</span></span></td><td>triaged, oncall: pt2, module: inductor, repro:agent:h100:repro_fail:not_reproduced, repro:agent:h100:incorrect</td><td><div class='wide'>🐛 Describe the bug Consider this code: import torch from torch._inductor.utils import fresh_cache def my_addmm(x, a, b, alpha, beta): return torch.addmm(x, a, b, alpha=beta, beta=alpha) x = torch.randn((128, 128)).cuda().half() a = torch.randn(128, 128).cuda().half() b = torch.randn(128, 128).cuda().half().t() with...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/187781' target='_blank' rel='noopener noreferrer'>Issue #187781: [RFC] Topology-aware nn.Module primitives for memory optimization in torch.compile</a></td><td>has linked PR</td><td>62</td><td><span class='chip-row'><span class='chip'>PR #187790</span></span></td><td>feature, module: nn, module: memory usage, triaged, needs research, oncall: pt2, module: compile ux, bot-triaged</td><td><div class='wide'>Hi PyTorch team, I am working on a research prototype called TopologyAwareJointOptimizer. I would like feedback on whether topology-aware memory optimization is a direction that should be explored around torch.compile, torch.nn, torchao, or an out-of-tree PyTorch backend. The problem I am trying to address is traini...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189161' target='_blank' rel='noopener noreferrer'>Issue #189161: [Inductor] Remove Redundant NaN Propagation in `max` op in Softmax</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: performance, triaged, enhancement, oncall: pt2, module: inductor, bot-triaged</td><td><div class='wide'>🚀 The feature, motivation and pitch Redundant NaN Propagation in max op in non-onlined Softmax Issue Current Inductor softmax can do redundant NaN propagation in the softmax-internal max reduction. For the fallback/persistent softmax form, the computation is effectively: xmax = max(x, dim=-1, keepdim=True) xexp = ex...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189157' target='_blank' rel='noopener noreferrer'>Issue #189157: GHA stats artifact downloader does not skip mismatched workflow run attempts</a></td><td>has linked PR</td><td>44</td><td><span class='chip-row'><span class='chip'>PR #189158</span></span></td><td>module: ci, triaged, bot-triaged</td><td><div class='wide'>🐛 Describe the bug tools.stats.upload_stats_lib._download_artifact() is intended to skip GitHub-hosted fallback artifacts whose embedded runattemptN token does not match the requested workflow_run_attempt. Today the mismatch branch only prints a Skipping ... message. After that, the function continues to download th...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189151' target='_blank' rel='noopener noreferrer'>Issue #189151: DISABLED test_pending_fusion_pro_and_epi_use_async_compile_False (__main__.TestPrologueFusion)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: rocm, triaged, module: flaky-tests, skipped, oncall: pt2, module: inductor</td><td><div class='wide'>Platforms: rocm This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 4 workflow(s) with 4 failures and 4 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSUME THIN...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189133' target='_blank' rel='noopener noreferrer'>Issue #189133: DISABLED test_vmap (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd)</a></td><td>has linked PR</td><td>44</td><td><span class='chip-row'><span class='chip'>PR #189286</span></span></td><td>triaged, module: flaky-tests, skipped, oncall: pt2, module: inductor</td><td><div class='wide'>Platforms: linux, rocm, slow This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 12 workflow(s) with 24 failures and 12 successes. Debugging instructions (after clicking on the recent samples link): DO...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/152944' target='_blank' rel='noopener noreferrer'>Issue #152944: DISABLED test_compiler_collectives_automatic_dynamic_tensor (__main__.TestMultiProc)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, skipped, oncall: pt2, module: dynamic shapes, module: dynamo, PT2-Bug-Bash</td><td><div class='wide'>Platforms: inductor, rocm This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 3 hours, it has been determined flaky in 3 workflow(s) with 3 failures and 3 successes. Debugging instructions (after clicking on the recent samples link): DO NOT A...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/119364' target='_blank' rel='noopener noreferrer'>Issue #119364: DISABLED test_source_multithreaded_close_in_scope_work_in_main_thread_False (__main__.TestProfiler)</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: flaky-tests, skipped, oncall: profiler</td><td><div class='wide'>Platforms: asan, dynamo, linux, mac, macos, rocm, slow This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 3 hours, it has been determined flaky in 10 workflow(s) with 30 failures and 10 successes. Debugging instructions (after clicking on th...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/186865' target='_blank' rel='noopener noreferrer'>Issue #186865: DISABLED test_dynamic_shape_backend_eager_device_cpu (__main__.TestPackage)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, skipped, oncall: pt2, module: dynamo</td><td><div class='wide'>Platforms: linux This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 3 workflow(s) with 3 failures and 3 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSUME THI...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/180195' target='_blank' rel='noopener noreferrer'>Issue #180195: DISABLED test_vjp_vmap_cpu (__main__.TestComposabilityCPU)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, skipped, oncall: pt2, module: functorch</td><td><div class='wide'>Platforms: dynamo This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 3 workflow(s) with 3 failures and 3 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSUME TH...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189137' target='_blank' rel='noopener noreferrer'>Issue #189137: [Sub-RFC] PrivateUse1 Inductor codegen registration interfaces (RFC 3/3)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>feature, triaged, module: backend, oncall: pt2, module: inductor, oncall: export, module: aotinductor, module: PrivateUse1, bot-triaged</td><td><div class='wide'>TL;DR This sub-RFC addresses the third slice of the umbrella RFC: providing registration interfaces for Inductor codegen so that a PrivateUse1 backend can be auto-discovered, configured, compiled, and run without monkeypatching upstream Inductor code. Four things are addressed: (1) backend discovery via entry-point...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189136' target='_blank' rel='noopener noreferrer'>Issue #189136: [Sub-RFC] PrivateUse1 Dynamo tracing via registry splicing (RFC 2/3)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, enhancement, oncall: pt2, module: dynamo, module: PrivateUse1, bot-triaged</td><td><div class='wide'>TL;DR This sub-RFC addresses the second slice of the umbrella RFC: making Dynamo&#x27;s six hardcoded lookup tables recognize PrivateUse1 device types and runtime functions by splicing them from the device interface registry. We introduce get_interface_for_device__PrivateUse1() — an iterable that returns only PrivateUse1...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189135' target='_blank' rel='noopener noreferrer'>Issue #189135: [Sub-RFC] PrivateUse1 device identity/capability in torch.compile (RFC 1/3)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>feature, triaged, module: backend, oncall: pt2, module: inductor, module: dynamo, module: PrivateUse1, bot-triaged</td><td><div class='wide'>TL;DR This sub-RFC addresses the first slice of the umbrella RFC: making torch.compile query the DeviceInterface contract for device identity (is_gpu) and capabilities (is_triton_capable, get_multi_processor_count, exposes_streams) instead of hardcoded cuda/xpu/mps lists. We introduce register_interface_for_device_P...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189138' target='_blank' rel='noopener noreferrer'>Issue #189138: [RFC] DeviceInterface: Registry Support for PrivateUse1 Backends in torch.compile</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>feature, triaged, large, module: backend, oncall: pt2, module: inductor, module: dynamo, module: PrivateUse1, bot-triaged</td><td><div class='wide'>TL;DR This RFC document aims to propose and discuss making PrivateUse1 (out-of-tree) backends first-class citizens in torch.compile. Our focus is on eliminating the hardcoded cuda/xpu/mps device lists and if-branches that permeate the Dynamo and Inductor codebase, replacing them with a unified registry-based contrac...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/75171' target='_blank' rel='noopener noreferrer'>Issue #75171: torch.jit.load fails when path contains non-ascii characters</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>oncall: jit, module: windows</td><td><div class='wide'>🐛 Describe the bug openai/CLIP#227 torch.jit.load fails when path contains non-ascii characters import torch torch.jit.load(&#x27;C:\\Users\\流星暴雨/.cache/clip\\ViT-B-16.pt&#x27;, &quot;cpu&quot;) Traceback (most recent call last): File &quot;&lt;string&gt;&quot;, line 1, in &lt;module&gt; File &quot;[C:\Users\]()流星暴雨\AppData\Local\Programs\Python\Python39\lib\sit...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188892' target='_blank' rel='noopener noreferrer'>Issue #188892: nvidia-cudnn-cu13==9.20.0.48 bundled by torch is incomplete → CUDNN_STATUS_SUBLIBRARY_VERSION_MISMATCH</a></td><td>has linked PR</td><td>106</td><td><span class='chip-row'><span class='chip'>PR #189043</span></span></td><td>needs reproduction, module: binaries, module: crash, module: cudnn, module: convolution, triaged, has workaround, bot-triaged</td><td><div class='wide'>🐛 Describe the bug torch==2.12.1+cu130 pins nvidia-cudnn-cu13==9.20.0.48, whose libcudnn_graph.so.9 dlopens libcudnn_engines_tensor_ir.so by bare soname - but that wheel doesn&#x27;t ship it (it&#x27;s only in ≥9.23). With no other cuDNN around this fails silently and falls back. But if a different system cuDNN 9.x is on the...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/186867' target='_blank' rel='noopener noreferrer'>Issue #186867: DISABLED test_group_norm_backward_large_input_float16_trigger_32bit_overflow_True_mps_float16 (__main__.TestConsistencyMPS)</a></td><td>has linked PR</td><td>44</td><td><span class='chip-row'><span class='chip'>PR #189215</span></span></td><td>triaged, module: flaky-tests, module: macos, skipped, module: mps</td><td><div class='wide'>Platforms: mac, macos This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 19 workflow(s) with 38 failures and 19 successes. Debugging instructions (after clicking on the recent samples link): DO NOT AS...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188798' target='_blank' rel='noopener noreferrer'>Issue #188798: DISABLED test_linearize_jvp_fn (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, skipped, oncall: pt2, module: inductor</td><td><div class='wide'>Platforms: linux, rocm, slow This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 3 workflow(s) with 6 failures and 3 successes. Debugging instructions (after clicking on the recent samples link): DO NO...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/119357' target='_blank' rel='noopener noreferrer'>Issue #119357: DISABLED test_source_multithreaded_basic_work_in_main_thread_True (__main__.TestProfiler)</a></td><td>needs design decision</td><td>65</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: flaky-tests, skipped, oncall: profiler</td><td><div class='wide'>Platforms: linux, rocm, asan, mac, macos, slow This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 3 hours, it has been determined flaky in 11 workflow(s) with 33 failures and 11 successes. Debugging instructions (after clicking on the recent...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/176966' target='_blank' rel='noopener noreferrer'>Issue #176966: DISABLED test_trace_return_namedtuple (__main__.TestFX)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, skipped, module: fx, oncall: pt2</td><td><div class='wide'>Platforms: dynamo This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 6 hours, it has been determined flaky in 9 workflow(s) with 9 failures and 9 successes. Debugging instructions (after clicking on the recent samples link): DO NOT ASSUME TH...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/124152' target='_blank' rel='noopener noreferrer'>Issue #124152: DISABLED test_torchvision_models_efficientnet_v2_m (__main__.TestVisionTracing)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, module: flaky-tests, skipped, module: fx</td><td><div class='wide'>Platforms: asan, linux, rocm, slow This test was disabled because it is failing in CI. See recent examples and the most recent trunk workflow logs. Over the past 3 hours, it has been determined flaky in 2 workflow(s) with 2 failures and 2 successes. Debugging instructions (after clicking on the recent samples link):...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188332' target='_blank' rel='noopener noreferrer'>Issue #188332: Inductor: allgather fusion creates rank-specific cache that crashes when loaded by other ranks</a></td><td>high-priority/blocker</td><td>70</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>high priority, triaged, oncall: pt2</td><td><div class='wide'>🐛 Describe the bug Minimal repro: Inductor FX graph cache reuses rank-specific code. Run twice with two GPUs. The first run cold-compiles and should pass; the second run reloads the shared warm FX graph cache and should fail because one rank can load the other rank&#x27;s device/rank-specific generated wrapper. rm -rf /t...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/144965' target='_blank' rel='noopener noreferrer'>Issue #144965: RuntimeError &quot;global alloc not supported yet&quot; when using TorchScript optimization.</a></td><td>needs design decision</td><td>115</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>high priority, triage review, oncall: jit</td><td><div class='wide'>🐛 Describe the bug Calling the forward method on TorchScript models can, under some specific conditions, raise a RuntimeError with message: &quot;Global alloc not supported yet.&quot;. I think this is linked to an old issue: #69078, however, I managed to consistently reproduce this error. The code that reproduces the bug is q...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188987' target='_blank' rel='noopener noreferrer'>Issue #188987: DISABLED test_optimizer_non_static_param (__main__.LoggingTests)</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: logging, module: tests, triaged, skipped, oncall: pt2, module: xpu, bot-triaged</td><td><div class='wide'>Platforms: xpu This test was disabled because it is failing on main branch (recent examples). cc @gujinghui @EikanWang @fengyuan14 @guangyey cc @mruberry @chauhang @penguinwu @gujinghui @EikanWang @fengyuan14 @guangyey</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/188970' target='_blank' rel='noopener noreferrer'>Issue #188970: [MPS] SIGSEGV in MPSStream::copy with triton.cudagraphs (reduce-overhead/max-autotune) on deep fp16 graphs</a></td><td>has linked PR</td><td>106</td><td><span class='chip-row'><span class='chip'>PR #181720</span></span></td><td>needs reproduction, module: crash, triaged, module: macos, module: mps, module: inductor, bot-triaged</td><td><div class='wide'>🐛 Describe the bug For a family of deep fp16 inference models, torch.compile on MPS deterministically crashes the process (SIGSEGV, exit 139) whenever the compile mode enables triton.cudagraphs, and works (bit-exact correct outputs) whenever it does not: configuration triton.cudagraphs result eager — ✅ OK torch.comp...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189126' target='_blank' rel='noopener noreferrer'>Issue #189126: Report a bug, as requested by the error message</a></td><td>needs design decision</td><td>115</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triage review, module: crash, module: cuda, module: memory usage, bot-triaged</td><td><div class='wide'>🐛 Describe the bug terminate called after throwing an instance of &#x27;c10::Error&#x27; what(): inserted INTERNAL ASSERT FAILED at &quot;/pytorch/c10/cuda/CUDACachingAllocator.cpp&quot;:3252, ... Versions curl -sL https://raw.githubusercontent.com/pytorch/pytorch/main/torch/utils/collect_env.py | python command returns: Collecting env...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/187093' target='_blank' rel='noopener noreferrer'>Issue #187093: [inductor] baddbmm with broadcast bias: autotune never considers the bmm + pointwise-bias decomposition (1.3-1.45x left on the table at MoE expert-GEMM shapes)</a></td><td>has linked PR</td><td>44</td><td><span class='chip-row'><span class='chip'>PR #187106</span><span class='chip'>PR #189127</span></span></td><td>module: performance, triaged, enhancement, oncall: pt2, module: inductor, bot-triaged</td><td><div class='wide'>🐛 Describe the bug For baddbmm with a broadcast bias (bias[B, 1, N], stride [N, 0, 1] after broadcasting), Inductor&#x27;s max-autotune chooses between (a) Triton templates and (b) the extern ATen kernel — but never considers the decomposition bmm(x, w) + bias, which is faster than both of its existing choices whenever t...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189121' target='_blank' rel='noopener noreferrer'>Issue #189121: [inductor] Single-config reductions extended by _dynamic_scale_rblock are permanently uncacheable — re-benchmark {R0_BLOCK, R0_BLOCK/2} every process; near-tied winners flip perf and numerics</a></td><td>needs design decision</td><td>89</td><td><span class='chip-row'><span class='chip'>PR #189124</span></span></td><td>triage review, module: performance, module: determinism, module: reductions, oncall: pt2, module: inductor, module: compile-time, bot-triaged</td><td><div class='wide'>🐛 Describe the bug Follow-up to #181789 (related: #172819 / #173391): one more case where _dynamic_scale_rblock mutates the autotune candidate set after the cache contract is frozen, defeating autotune_local_cache=True. Verified on current nightly. check_autotune_cache only creates an AutotuneCache when there is mor...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/185839' target='_blank' rel='noopener noreferrer'>Issue #185839: torch.linalg.qr (float64) repeated in a loop crashes the process (SIGSEGV, exit 139) on ROCm 7.x</a></td><td>high-priority/blocker</td><td>70</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: crash, module: rocm, triaged, module: regression, has workaround, module: linear algebra, module: CUDACachingAllocator, bot-triaged</td><td><div class='wide'>🐛 Describe the bug Calling torch.linalg.qr on a float64 matrix repeatedly in a loop causes the process to be killed by SIGSEGV (exit code 139) after ~31 iterations on ROCm 7.x (AMD Instinct MI300A). No Python exception is raised — the process dies silently with no traceback. The crash is specific to torch.linalg.qr...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/issues/189046' target='_blank' rel='noopener noreferrer'>Issue #189046: Treat PrivateUse1 as CUDA-like under env override to reuse CUDA optimization paths in shared `at::native` operators</a></td><td>needs maintainer decision</td><td>28</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>feature, module: autograd, triaged, module: advanced indexing, module: dispatch, module: backend, module: reductions, module: PrivateUse1, bot-triaged</td><td><div class='wide'>🚀 The feature, motivation and pitch Allow CUDA-like PrivateUse1 backends to opt into existing CUDA optimization paths in shared at::native operators via an environment variable, avoiding code duplication and performance loss. 🚀 Feature Both this issue and #176715 share the same goal: third-party privateuse1 backends...</div></td></tr></tbody></table></div></section></div>

data/manifest.html

@@ -0,0 +1,75 @@
+
+<style>
+:root { color-scheme: light; }
+.bi-wrap { font-family: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; color: #1f2328 !important; }
+.bi-wrap h1, .bi-wrap h2, .bi-wrap h3, .bi-wrap p, .bi-wrap strong, .bi-wrap td, .bi-wrap th, .bi-wrap div { color: #1f2328 !important; }
+.bi-band { border: 1px solid #d8dee4; border-radius: 8px; background: #ffffff !important; padding: 16px; margin: 0 0 12px; }
+.bi-title { display: flex; gap: 12px; align-items: flex-start; justify-content: space-between; margin-bottom: 12px; }
+.bi-title h1, .bi-title h2 { margin: 0; line-height: 1.15; letter-spacing: 0; }
+.bi-title h1 { font-size: 1.55rem; }
+.bi-title h2 { font-size: 1.12rem; }
+.bi-title p { margin: 4px 0 0; color: #57606a; max-width: 820px; }
+.status-pill { border-radius: 999px; padding: 5px 9px; font-size: .78rem; font-weight: 700; background: #dafbe1; color: #116329 !important; white-space: nowrap; }
+.metric-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(138px, 1fr)); gap: 8px; }
+.metric { border: 1px solid #d0d7de; border-radius: 8px; padding: 10px; background: #f6f8fa !important; min-height: 86px; }
+.metric-value { font-size: 1.45rem; font-weight: 800; line-height: 1.1; color: #0969da !important; }
+.metric-label { margin-top: 4px; font-size: .84rem; font-weight: 700; }
+.metric-note { margin-top: 2px; font-size: .76rem; color: #57606a !important; }
+.callout-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 8px; }
+.callout { border-left: 4px solid #0969da; background: #f6f8fa !important; border-radius: 8px; padding: 10px 12px; }
+.callout:nth-child(2) { border-left-color: #1a7f37; }
+.callout:nth-child(3) { border-left-color: #bf8700; }
+.callout:nth-child(4) { border-left-color: #cf222e; }
+.callout strong { display: block; margin-bottom: 4px; }
+.group-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(310px, 1fr)); gap: 10px; }
+.group-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.group-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.meta-row { display: flex; flex-wrap: wrap; gap: 6px; margin: 6px 0; }
+.chip-row { display: inline-flex; flex-wrap: wrap; gap: 5px; align-items: center; vertical-align: middle; }
+.chip { display: inline-flex; align-items: center; border: 1px solid #d0d7de; border-radius: 999px; padding: 2px 7px; font-size: .78rem; background: #f6f8fa !important; color: #24292f !important; }
+.chip.bucket { border-color: #9a6700; background: #fff8c5 !important; }
+.chip.kind { border-color: #0969da; background: #ddf4ff !important; }
+.workflow-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(260px, 1fr)); gap: 10px; }
+.workflow-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.workflow-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.workflow-card .move-count { font-weight: 800; color: #0969da !important; }
+.move-list { display: grid; gap: 10px; }
+.move-row { border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+.move-row summary { list-style: none; cursor: pointer; padding: 12px; }
+.move-row summary::-webkit-details-marker { display: none; }
+.move-summary { display: grid; grid-template-columns: minmax(190px, 1fr) minmax(260px, 1.45fr) minmax(260px, 1.45fr) minmax(150px, .8fr) minmax(250px, 1.35fr) auto; gap: 16px; align-items: start; }
+.move-title, .move-field { min-width: 0; }
+.move-title h3 { margin: 0 0 4px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.move-field span { display: block; color: #57606a !important; font-size: .72rem; font-weight: 800; text-transform: uppercase; letter-spacing: 0; }
+.move-field p { margin: 3px 0 0; font-size: .84rem; line-height: 1.35; overflow-wrap: anywhere; }
+.move-field .chip, .move-title { overflow-wrap: anywhere; }
+.view-bundles { align-self: center; justify-self: end; border: 1px solid #000000; border-radius: 999px; color: #000000 !important; background: #ffffff !important; padding: 4px 9px; font-size: .78rem; font-weight: 800; white-space: nowrap; }
+.move-row[open] .view-bundles { background: #f6f8fa !important; }
+.move-expanded { border-top: 1px solid #d8dee4; padding: 0 12px 12px; }
+.bundle-row { border-top: 1px solid #eef1f4; padding: 12px 0; }
+.bundle-row:first-child { border-top: none; }
+.bundle-row h4 { margin: 0 0 6px; font-size: .94rem; line-height: 1.25; letter-spacing: 0; }
+.bundle-scope { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 7px; margin-top: 8px; }
+.evidence-list { margin: 8px 0 0 18px; padding: 0; color: #57606a !important; font-size: .82rem; line-height: 1.35; }
+.evidence-list li { margin: 3px 0; }
+.muted { color: #57606a !important; }
+.evidence { font-size: .84rem; color: #57606a !important; }
+.scroll-table { overflow-x: auto; border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+table.bi-table { width: 100%; border-collapse: collapse; font-size: .86rem; }
+.bi-table th { text-align: left; background: #f6f8fa !important; color: #57606a !important; font-size: .78rem; border-bottom: 1px solid #d8dee4; padding: 8px; white-space: nowrap; }
+.bi-table td { border-bottom: 1px solid #eef1f4; padding: 8px; vertical-align: top; min-width: 86px; }
+.bi-table tr:last-child td { border-bottom: none; }
+.wide { min-width: 280px; }
+.tight { max-width: 180px; }
+a { color: #0969da !important; text-decoration: none; }
+a:hover { text-decoration: underline; }
+.bi-wrap, .bi-wrap * { color: #000000 !important; }
+@media (max-width: 720px) {
+  .bi-title { display: block; }
+  .status-pill { display: inline-block; margin-top: 8px; }
+  .group-grid { grid-template-columns: 1fr; }
+  .move-summary { grid-template-columns: 1fr; }
+  .view-bundles { justify-self: start; }
+}
+</style>
+<div class='bi-wrap'><section class='bi-band'><div class='bi-title'><div><h2>Manifest</h2><p>Release boundary for this public static bundle.</p></div></div><div class='scroll-table'><table class='bi-table'><thead><tr><th>Field</th><th>Value</th></tr></thead><tbody><tr><td>Repo</td><td>pytorch/pytorch</td></tr><tr><td>Space visibility</td><td>Public</td></tr><tr><td>Generated</td><td>2026-07-09T03:47:02+00:00</td></tr><tr><td>Packet hash</td><td>bb2ad53aa6630f3381c2a30f2e4225d98ab5e28172ceb254137002bd27dd863c</td></tr><tr><td>Metrics hash</td><td>401994156ad917ca0d3219a4265c760111d5a39007e87a8e368a224bcac23395</td></tr><tr><td>Full collection</td><td>True</td></tr><tr><td>Raw API payloads included</td><td>False</td></tr><tr><td>GitHub writes</td><td>None</td></tr></tbody></table></div></section></div>

data/pull_requests.html

@@ -0,0 +1,75 @@
+
+<style>
+:root { color-scheme: light; }
+.bi-wrap { font-family: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; color: #1f2328 !important; }
+.bi-wrap h1, .bi-wrap h2, .bi-wrap h3, .bi-wrap p, .bi-wrap strong, .bi-wrap td, .bi-wrap th, .bi-wrap div { color: #1f2328 !important; }
+.bi-band { border: 1px solid #d8dee4; border-radius: 8px; background: #ffffff !important; padding: 16px; margin: 0 0 12px; }
+.bi-title { display: flex; gap: 12px; align-items: flex-start; justify-content: space-between; margin-bottom: 12px; }
+.bi-title h1, .bi-title h2 { margin: 0; line-height: 1.15; letter-spacing: 0; }
+.bi-title h1 { font-size: 1.55rem; }
+.bi-title h2 { font-size: 1.12rem; }
+.bi-title p { margin: 4px 0 0; color: #57606a; max-width: 820px; }
+.status-pill { border-radius: 999px; padding: 5px 9px; font-size: .78rem; font-weight: 700; background: #dafbe1; color: #116329 !important; white-space: nowrap; }
+.metric-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(138px, 1fr)); gap: 8px; }
+.metric { border: 1px solid #d0d7de; border-radius: 8px; padding: 10px; background: #f6f8fa !important; min-height: 86px; }
+.metric-value { font-size: 1.45rem; font-weight: 800; line-height: 1.1; color: #0969da !important; }
+.metric-label { margin-top: 4px; font-size: .84rem; font-weight: 700; }
+.metric-note { margin-top: 2px; font-size: .76rem; color: #57606a !important; }
+.callout-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 8px; }
+.callout { border-left: 4px solid #0969da; background: #f6f8fa !important; border-radius: 8px; padding: 10px 12px; }
+.callout:nth-child(2) { border-left-color: #1a7f37; }
+.callout:nth-child(3) { border-left-color: #bf8700; }
+.callout:nth-child(4) { border-left-color: #cf222e; }
+.callout strong { display: block; margin-bottom: 4px; }
+.group-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(310px, 1fr)); gap: 10px; }
+.group-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.group-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.meta-row { display: flex; flex-wrap: wrap; gap: 6px; margin: 6px 0; }
+.chip-row { display: inline-flex; flex-wrap: wrap; gap: 5px; align-items: center; vertical-align: middle; }
+.chip { display: inline-flex; align-items: center; border: 1px solid #d0d7de; border-radius: 999px; padding: 2px 7px; font-size: .78rem; background: #f6f8fa !important; color: #24292f !important; }
+.chip.bucket { border-color: #9a6700; background: #fff8c5 !important; }
+.chip.kind { border-color: #0969da; background: #ddf4ff !important; }
+.workflow-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(260px, 1fr)); gap: 10px; }
+.workflow-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.workflow-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.workflow-card .move-count { font-weight: 800; color: #0969da !important; }
+.move-list { display: grid; gap: 10px; }
+.move-row { border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+.move-row summary { list-style: none; cursor: pointer; padding: 12px; }
+.move-row summary::-webkit-details-marker { display: none; }
+.move-summary { display: grid; grid-template-columns: minmax(190px, 1fr) minmax(260px, 1.45fr) minmax(260px, 1.45fr) minmax(150px, .8fr) minmax(250px, 1.35fr) auto; gap: 16px; align-items: start; }
+.move-title, .move-field { min-width: 0; }
+.move-title h3 { margin: 0 0 4px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.move-field span { display: block; color: #57606a !important; font-size: .72rem; font-weight: 800; text-transform: uppercase; letter-spacing: 0; }
+.move-field p { margin: 3px 0 0; font-size: .84rem; line-height: 1.35; overflow-wrap: anywhere; }
+.move-field .chip, .move-title { overflow-wrap: anywhere; }
+.view-bundles { align-self: center; justify-self: end; border: 1px solid #000000; border-radius: 999px; color: #000000 !important; background: #ffffff !important; padding: 4px 9px; font-size: .78rem; font-weight: 800; white-space: nowrap; }
+.move-row[open] .view-bundles { background: #f6f8fa !important; }
+.move-expanded { border-top: 1px solid #d8dee4; padding: 0 12px 12px; }
+.bundle-row { border-top: 1px solid #eef1f4; padding: 12px 0; }
+.bundle-row:first-child { border-top: none; }
+.bundle-row h4 { margin: 0 0 6px; font-size: .94rem; line-height: 1.25; letter-spacing: 0; }
+.bundle-scope { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 7px; margin-top: 8px; }
+.evidence-list { margin: 8px 0 0 18px; padding: 0; color: #57606a !important; font-size: .82rem; line-height: 1.35; }
+.evidence-list li { margin: 3px 0; }
+.muted { color: #57606a !important; }
+.evidence { font-size: .84rem; color: #57606a !important; }
+.scroll-table { overflow-x: auto; border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+table.bi-table { width: 100%; border-collapse: collapse; font-size: .86rem; }
+.bi-table th { text-align: left; background: #f6f8fa !important; color: #57606a !important; font-size: .78rem; border-bottom: 1px solid #d8dee4; padding: 8px; white-space: nowrap; }
+.bi-table td { border-bottom: 1px solid #eef1f4; padding: 8px; vertical-align: top; min-width: 86px; }
+.bi-table tr:last-child td { border-bottom: none; }
+.wide { min-width: 280px; }
+.tight { max-width: 180px; }
+a { color: #0969da !important; text-decoration: none; }
+a:hover { text-decoration: underline; }
+.bi-wrap, .bi-wrap * { color: #000000 !important; }
+@media (max-width: 720px) {
+  .bi-title { display: block; }
+  .status-pill { display: inline-block; margin-top: 8px; }
+  .group-grid { grid-template-columns: 1fr; }
+  .move-summary { grid-template-columns: 1fr; }
+  .view-bundles { justify-self: start; }
+}
+</style>
+<div class='bi-wrap'><section class='bi-band'><div class='bi-title'><div><h2>Open Pull Requests</h2><p>Public-safe normalized rows for auditability.</p></div></div><div class='scroll-table'><table class='bi-table'><thead><tr><th>Pull Request</th><th>Bucket</th><th>Score</th><th>Linked</th><th>Labels</th><th>Excerpt</th></tr></thead><tbody><tr><td><a href='https://github.com/pytorch/pytorch/pull/189096' target='_blank' rel='noopener noreferrer'>Pull Request #189096: [aot_autograd] Make CPU-side stream syncs real cross-stream barriers</a></td><td>waiting on contributor</td><td>43</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: inductor, ciflow/inductor, release notes: inductor</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #189096 #189095 synchronize_stream / synchronize_device / synchronize_event block the CPU until a stream (or event) drains, so every subsequent kernel launch is host-ordered after them. Unlike device-side waits (wait_event / wait_stream), nothing on a consumer&#x27;s stream waits...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189024' target='_blank' rel='noopener noreferrer'>Pull Request #189024: [Dynamo] Support exception attribute access (StopIteration/AttributeError/NameError)</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, ciflow/trunk, module: dynamo, ciflow/inductor, release notes: dynamo, merging, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #189024 #188825 #188834 #188824 #157149 #188639 #188638 #188004 #187744 #187690 Dynamo wrapped exceptions in a single ExceptionVariable that only tracked args/context/cause/traceback. Reading exception-specific attributes (StopIteration.value, AttributeError.name/.obj, NameE...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189318' target='_blank' rel='noopener noreferrer'>Pull Request #189318: Bump pip from 26.0.1 to 26.1.2 in /.ci/docker</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, topic: not user facing, python, dependencies, ciflow/docker</td><td><div class='wide'>Bumps pip from 26.0.1 to 26.1.2. Changelog Sourced from pip&#x27;s changelog. 26.1.2 (2026-05-31) Bug Fixes Reject console_scripts and gui_scripts entry points whose name would install a script outside the scripts directory. ([#14000](https://github.com/pypa/pip/issues/14000) &lt;https://github.com/pypa/pip/issues/14000&gt;_)...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189329' target='_blank' rel='noopener noreferrer'>Pull Request #189329: [DEBUG] Profile the XPU SYCL device-link bottleneck</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>ciflow/xpu</td><td><div class='wide'>Do not merge - temporary diagnostic. Why linux-noble-xpu-n-py3.10 / build takes ~35-45 min vs 5-15 min for the other builds in the same workflow. Profiling pull run 28986110660 showed the Build step is ~33 min even with sccache at 99.93% hit rate - because sccache caches only host C/C++, not SYCL device code. The si...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188597' target='_blank' rel='noopener noreferrer'>Pull Request #188597: [ROCm] Add initial support for gfx1250</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: rocm, triaged, open source, Merged, Reverted, ciflow/binaries, ciflow/trunk, release notes: releng, module: inductor, ciflow/inductor, merging, ci-no-td, ciflow/rocm-mi300, ciflow/h100, ciflow/b200, ciflow/torchtitan, ciflow/docker</td><td><div class='wide'>This PR adds initial ROCm support for gfx1250 (CDNA5). What this PR enables: Gates gfx1250-specific behavior on ROCm 7.14+ (the support floor), applied consistently across CUDABlas.cpp, ScaledBlas.cpp, CUDAHooks.cpp, and the test helpers. Adds gfx1250 to the hipBLASLt preferred/supported arch lists and the hipSparse...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188963' target='_blank' rel='noopener noreferrer'>Pull Request #188963: [XPU] Enable XPU testing for 50 test classes across 40 test files</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, release notes: quantization, topic: not user facing, module: inductor, module: dynamo, ciflow/inductor, keep-going, ciflow/xpu, ci-no-td, ciflow/rocm-mi300, ciflow/h100, ciflow/b200</td><td><div class='wide'>Summary Enable XPU (allow_xpu=True) for test classes that are confirmed working on Intel GPU (XPU). Each class was validated by running pytest -k &quot;&lt;ClassName&gt; and xpu&quot; in an XPU-enabled PyTorch environment. Changes by file File Classes Enabled test/complex_tensor/test_complex_tensor.py TestComplexTensor, TestComplex...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189173' target='_blank' rel='noopener noreferrer'>Pull Request #189173: torchfuzz: detach batch_norm running stats</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, open source, ciflow/trunk, topic: not user facing, merging</td><td><div class='wide'>The fuzzer marks float tensor args requires_grad=True for the backward pass, but torch.nn.functional.batch_norm is not differentiable w.r.t. running_mean / running_var, so when those inputs are grad-requiring args the generated program raises RuntimeError: ... &#x27;running_mean&#x27; ... cannot have requires_grad True in eag...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189111' target='_blank' rel='noopener noreferrer'>Pull Request #189111: [torchcomms hash update] update the pinned torchcomms hash</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, ciflow/trunk, topic: not user facing, ciflow/inductor, merging, ci-no-td</td><td><div class='wide'>This PR is auto-generated nightly by this action. Update the pinned torchcomms hash.</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/184218' target='_blank' rel='noopener noreferrer'>Pull Request #184218: Support decorator/curried calling convention for torch.utils.checkpoint.checkpoint</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>ciflow/trunk, release notes: distributed (fsdp), topic: new features, module: inductor, module: dynamo, ciflow/inductor, release notes: distributed (dtensor), ciflow/torchtitan, ciflow/dtensor</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #184218 checkpoint() can now be called without a function argument to produce a decorator, keeping checkpoint configuration separate from the checkpointed function&#x27;s arguments: checkpointed_fn = checkpoint(use_reentrant=False)(fn) out = checkpointed_fn(*args, **kwargs) This...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188980' target='_blank' rel='noopener noreferrer'>Pull Request #188980: Fix inconsistent accelerator guard in checkpoint staging test</a></td><td>waiting on CI/check fix</td><td>73</td><td><span class='chip-row'><span class='chip'>#188790</span></span></td><td>open source, ciflow/trunk, topic: not user facing, release notes: distributed (checkpoint), merging</td><td><div class='wide'>Fixes #188790 Summary In test/distributed/checkpoint/_experimental/test_staging.py, the block that appends the async-staging and non-blocking-copy test cases was guarded by torch.cuda.is_available(), while the config values constructed inside that same block already use the backend-generic torch.accelerator.is_avail...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/187940' target='_blank' rel='noopener noreferrer'>Pull Request #187940: [xpu][inductor] fix the fallback behavior of maxpool backward for performance</a></td><td>waiting on contributor</td><td>43</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, open source, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, ciflow/xpu, ciflow/torchtitan</td><td><div class='wide'>Fix max_pool2d_backward fallback to eager op behavior. Previously, the decomposition pass was executed before lowering, causing max_pool2d_backward to be decomposed before the fallback mechanism could be applied(described in #182619). As a result, the eager fallback path was bypassed, and the operation was fused int...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189328' target='_blank' rel='noopener noreferrer'>Pull Request #189328: Add retry with backoff to download_mnist.py</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>topic: not user facing</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #189328 Note from human: this is not a big deal but it&#x27;s also the kind of thing coding agents eat for lunch, so whatever. MNIST download in CI hits a single S3 mirror via urlretrieve with no retry, so a transient network blip fails the whole job. This adds up to 3 attempts p...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189284' target='_blank' rel='noopener noreferrer'>Pull Request #189284: Add torch.autograd.graph.node_creation_hook</a></td><td>waiting on contributor</td><td>43</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>release notes: autograd, topic: new features</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #189284 Adds a public context manager that registers a thread-local hook fired once for every autograd Node created within it, with the freshly created Node as argument. This gives users a creation-time dual of saved_tensors_hooks: the hook can inspect the node, record it, s...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/181680' target='_blank' rel='noopener noreferrer'>Pull Request #181680: [triton hash update] update the pinned triton hash</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, ciflow/trunk, topic: not user facing, ciflow/inductor, ci-no-td, ciflow/docker</td><td><div class='wide'>This PR is auto-generated nightly by this action. Update the pinned triton hash.</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189217' target='_blank' rel='noopener noreferrer'>Pull Request #189217: [vision hash update] update the pinned vision hash</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, ciflow/trunk, topic: not user facing, ciflow/inductor, ci-no-td</td><td><div class='wide'>This PR is auto-generated nightly by this action. Update the pinned vision hash.</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/187288' target='_blank' rel='noopener noreferrer'>Pull Request #187288: [vllm hash update] update the pinned vllm hash</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, ciflow/trunk, topic: not user facing, ciflow/inductor, ci-no-td, ciflow/vllm, ci-refresh-hf-cache</td><td><div class='wide'>This PR is auto-generated nightly by this action. Update the pinned vllm hash.</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188262' target='_blank' rel='noopener noreferrer'>Pull Request #188262: [CI] fix OpenBLAS build for non-AArch64</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, open source, Merged, Reverted, ciflow/trunk, topic: not user facing, ci-no-td, ciflow/docker</td><td><div class='wide'>TARGET=ARMV8 was hardcoded, breaking non-aarch64 builds. Now it&#x27;s only set on aarch64; other platforms (like riscv64) rely on DYNAMIC_ARCH instead. Also forcibly enable the float/double/complex precision variants and pass build flags to make install consistently.</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/179651' target='_blank' rel='noopener noreferrer'>Pull Request #179651: [torchtitan hash update] update the pinned torchtitan hash</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, ciflow/trunk, topic: not user facing, ciflow/inductor, ci-no-td, ciflow/torchtitan</td><td><div class='wide'>This PR is auto-generated nightly by this action. Update the pinned torchtitan hash.</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/187600' target='_blank' rel='noopener noreferrer'>Pull Request #187600: [ROCm][Inductor] Enable NHWC convs by default on CDNA with an Inductor layout-opt gate</a></td><td>waiting on contributor</td><td>43</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: rocm, triaged, open source, ciflow/trunk, module: inductor, ciflow/inductor, keep-going, release notes: inductor, ciflow/rocm-mi300, ciflow/inductor-perf-test-nightly-rocm-mi300, ciflow/inductor-perf-test-nightly-rocm-mi350, ciflow/inductor-rocm-mi300, ciflow/torchtitan, ciflow/docker</td><td><div class='wide'>Enable NHWC convs by default on CDNA (gfx9xx) and gate Inductor layout-opt so grouped convs that hit MIOpen&#x27;s slow naive kernel stay NCHW. Builds on / supersedes #149039 (closed): keeps its NHWC-on-CDNA enablement, fixes the arch-guard bug, makes ConvUtils.h ATEN_CPU-lint-clean via a runtime CUDAHooks hook, adds a c...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189083' target='_blank' rel='noopener noreferrer'>Pull Request #189083: [profiler][cupti] Drive periodic cuptiActivityFlushAll from the decode thread</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>topic: not user facing</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189185 -&gt; #189083 #188621 #189109 The CUPTI monitor&#x27;s background flush loop woke on a Python thread every background_flush_period_s just to call cuptiActivityFlushAll -- a pure-C call that needs no GIL -- so it held the GIL for no reason and contended with the training thread....</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189185' target='_blank' rel='noopener noreferrer'>Pull Request #189185: [profiler][cupti] Make CuptiMonitor a singleton configured via a global configure()</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td></td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #189185 #189083 #188621 #189109 Ports the config-only change from commit 4a24d3b207 (drop the TORCH_CUPTI_MONITOR_* env vars) and turns CuptiMonitor into a process-wide singleton, mirroring PmSampler. buffer_size, background_flush_period_s, and background_drain_period_s were...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188621' target='_blank' rel='noopener noreferrer'>Pull Request #188621: [profiler][cupti] Generate CUPTI field-id catalogs from the ABI as Field(id, ctype)</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>topic: not user facing</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189185 #189083 -&gt; #188621 #189109 The v2 CUPTI user-defined-record path selects records by field id (CUpti_ActivityFieldIds), which cupti-python does not expose. Instead of hand-maintaining those ids, generate them from cupti_activity.h at build time, and model each field as a...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189229' target='_blank' rel='noopener noreferrer'>Pull Request #189229: [DO NOT MERGE][ROCm] Use torch.version.hip to detect ROCm when /opt/rocm is absent</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: rocm, open source, release notes: releng, module: inductor, module: dynamo, ciflow/inductor, release notes: distributed (checkpoint), keep-going, ci-no-td, ciflow/rocm-nightly, ciflow/dtensor, ciflow/docker</td><td><div class='wide'>cc @jeffdaily @sunway513 @jithunnair-amd @pruthvistony @ROCmSupport @jataylo @hongxiayang @naromero77amd @pragupta @jerrymannil @xinyazhang @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @kadeng @muchulee8 @amjames @chauhang @aakhundov @coc...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188100' target='_blank' rel='noopener noreferrer'>Pull Request #188100: Support callable msg lambda in TestCase.assert* methods beyond assertEqual()</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>ciflow/trunk, topic: not user facing, mergedog, make tests great again</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189286 #188573 #188299 -&gt; #188100 Allows for lazy building of error messages only when tests fail! Saves on overhead, especially when serialized tensors are involved.</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189275' target='_blank' rel='noopener noreferrer'>Pull Request #189275: Consolidate CD wheel build pipelines under .ci/wheel/ and unify the numpy pin</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, release notes: releng, ciflow/binaries_wheel</td><td><div class='wide'>Summary The three CD wheel-build pipelines (Linux, Windows, macOS) grew as three independent Python ports and drifted apart: copy-pasted helpers, three different numpy build-pin tables, and different directory homes. This PR consolidates them under a common .ci/wheel/ root with a single shared _common.py, and reconc...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/115316' target='_blank' rel='noopener noreferrer'>Pull Request #115316: Automated submodule update: FBGEMM</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, open source, topic: not user facing, ciflow/rocm-mi300, ciflow/h100, ciflow/b200</td><td><div class='wide'>This is an automated pull request to update the first-party submodule for pytorch/FBGEMM. New submodule commit: pytorch/FBGEMM@a75b43f Test Plan: Ensure that CI jobs succeed on GitHub before landing.</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189175' target='_blank' rel='noopener noreferrer'>Pull Request #189175: torchfuzz: fix clamp_min/max emitting None bound</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, open source, ciflow/trunk, topic: not user facing</td><td><div class='wide'>ClampMinOperator/ClampMaxOperator stored the random bound as mutable instance state in fuzz_inputs_specs and read it back in codegen. The same operator instance serves multiple clamp nodes, so during code generation the first codegen resets the stored bound to None and later clamp nodes emit clamp_min(x, min=None) o...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/167224' target='_blank' rel='noopener noreferrer'>Pull Request #167224: [FlopCounter] Add skip_unsupported parameter for graceful handling of unsupported HOPs</a></td><td>waiting on contributor</td><td>53</td><td><span class='chip-row'><span class='chip'>#134385</span></span></td><td>triaged, open source, release notes: python_frontend</td><td><div class='wide'>Fixes #134385 FlopCounterMode returns NotImplemented when it encounters a Higher Order Operator it does not handle, which makes it hard to use for benchmarking real-world models that mix standard PyTorch ops with custom HOPs or kernels. What this PR does Adds a skip_unsupported flag (default False) to FlopCounterMod...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188299' target='_blank' rel='noopener noreferrer'>Pull Request #188299: Lazily build error messages for assert*() calls across tests to reduce overhead</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>release notes: quantization, topic: not user facing, ciflow/mps, module: inductor, module: dynamo, ciflow/inductor, release notes: distributed (checkpoint), ciflow/rocm-mi300, ciflow/h100, ciflow/h100-symm-mem, ciflow/b200, ciflow/dtensor, mergedog, make tests great again</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189286 #188573 -&gt; #188299 #188100 Apply new lazy error message feature to call-sites throughout test files. cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @kadeng @muchulee8 @amjames @chauhang @aak...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189192' target='_blank' rel='noopener noreferrer'>Pull Request #189192: [MPS] Add `geqrf` and refactor `linalg_qr`</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, release notes: mps, ciflow/mps, ciflow/inductor, release notes: inductor (aoti)</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #189192 This PR splits out the geqrf code from the linalg_qr kernel and refactors linalg_qr for MPS to call the device-agnostic linalg_qr_out function that CPU and CUDA use.</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188573' target='_blank' rel='noopener noreferrer'>Pull Request #188573: Cleanup dead code across tests</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>release notes: quantization, topic: not user facing, module: inductor, module: dynamo, ciflow/inductor, ciflow/rocm-mi300, ciflow/h100, ciflow/b200, ciflow/dtensor, mergedog, make tests great again</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189286 -&gt; #188573 #188299 #188100 I don&#x27;t expect huge perf improvements, but why not? cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @kadeng @muchulee8 @amjames @chauhang @aakhundov @coconutruben @...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189264' target='_blank' rel='noopener noreferrer'>Pull Request #189264: [ROCm][CI] Make rocm-nightly workflow fully mi300-specific (name, ciflow tag, filename)</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: rocm, triaged, open source, ciflow/binaries, topic: not user facing, ciflow/rocm-mi300</td><td><div class='wide'>Summary Make the rocm-nightly workflow fully arch-specific for mi300: Renamed the workflow file to .github/workflows/rocm-nightly-mi300.yml. Set the workflow display name to rocm-nightly-mi300 and its push-trigger ciflow tag to ciflow/rocm-nightly-mi300/*. Job name and build-environment are linux-noble-rocm-nightly-...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189193' target='_blank' rel='noopener noreferrer'>Pull Request #189193: [XCCL] Allow XCCL to be set as default in `_new_process_group_helper`</a></td><td>waiting on CI/check fix</td><td>73</td><td><span class='chip-row'><span class='chip'>#4098</span></span></td><td>open source, ciflow/trunk, release notes: distributed (c10d), ciflow/torchtitan</td><td><div class='wide'>Fixes intel/torch-xpu-ops#4098 For DeviceMesh which doesn&#x27;t specify the backend type, _new_process_group_helper is called with the &quot;undefined&quot; backend, which is set based on available devices. NCCL backend can be set as default, as well as CUSTOM backends. This PR also allows the XCCL backend to be set as default if...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/187465' target='_blank' rel='noopener noreferrer'>Pull Request #187465: [Symm_mem] Support expandable segements inside PyTorch Symmetric memory for NCCL backend</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, ciflow/trunk, release notes: distributed (c10d), ciflow/h100-symm-mem, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #187465 #189088 This PR allows the NCCL symmetric-memory allocator to back its allocations with the CUDA caching allocator&#x27;s expandable segments; when expandable_segments is enabled. In that mode, NCCLSymmetricMemoryAllocator now allocates/frees the symmetric buffer via CUDA...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189302' target='_blank' rel='noopener noreferrer'>Pull Request #189302: [ROCm][CI] Make ROCM_SO_FILES a single shared source of truth for wheel bundling + preload</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: rocm, triaged, open source, ciflow/binaries_wheel, ciflow/rocm-mi300</td><td><div class='wide'>Summary The ROCm shared-library basename list was duplicated in two places that had drifted apart: .ci/manywheel/repair_wheel.py::ROCM_SO_FILES (27 entries) — the libs bundled into the wheel. torch/__init__.py::_rocm_core_libs (22 entries) — the libs preloaded RTLD_GLOBAL, leaf-first, before import torch._C (added i...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189314' target='_blank' rel='noopener noreferrer'>Pull Request #189314: flex_gemm: recognize 2-D block local-reduce contracts</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: inductor, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189316 #189315 -&gt; #189314 #189190 #189188 #188739 #188112 #188470 #188469 #188468 Recognize structural 2-D block local-reduce aux outputs of the form acc.view(M // bm, bm, N // bn, bn) reduced over both grouped dims. This threads block geometry through the output plan, templat...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189315' target='_blank' rel='noopener noreferrer'>Pull Request #189315: flex_gemm: store 2-D block local reductions</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: inductor, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189316 -&gt; #189315 #189314 #189190 #189188 #188739 #188112 #188470 #188469 #188468 Enable the first QUACK-backed 2-D block local-reduce STORE path for 128x128 block-scale aux outputs. The compiler now threads block geometry into runtime dispatch with explicit block kwargs, requ...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/187699' target='_blank' rel='noopener noreferrer'>Pull Request #187699: Add a debug log when we skip sync_decision_cross_ranks (#187699)</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>ciflow/trunk, ciflow/inductor, meta-exported, ciflow/torchtitan</td><td><div class='wide'>Summary: When there is collectives in the graph but there is no sync decision cross ranks, it could potentially have issue of inconsistent decisions on different ranks on what to save. So adding this debug log to facilitate debug when the inconsistent issue happens. Test Plan: Existing tests already cover it Reviewe...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189316' target='_blank' rel='noopener noreferrer'>Pull Request #189316: flex_gemm: support mx block-scale stores</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: inductor, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #189316 #189315 #189314 #189190 #189188 #188739 #188112 #188470 #188469 #188468 Compose 2-D block local-reduce stores with physical-only finalizers such as mx_e8m0_scale. Block aux plan classification now traces a shape-preserving pointwise finalizer back to its base block r...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189320' target='_blank' rel='noopener noreferrer'>Pull Request #189320: [inductor] Enable nested Inductor configs for invoke_subgraph regions</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: inductor, module: dynamo, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189321 -&gt; #189320 The normal Inductor invoke_subgraph path compiled each nested region under the surrounding graph&#x27;s Inductor config, so per-region config patches were dropped and nested options had no effect on the region. This threads the nested-region config through the inv...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189307' target='_blank' rel='noopener noreferrer'>Pull Request #189307: [CI/CD] Copy newer CUPTI headers into manywheel binary-build images</a></td><td>waiting on CI/check fix</td><td>108</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>topic: not user facing, ciflow/docker</td><td><div class='wide'>#189214 added install_cupti_headers, which stages the newer standalone CUPTI redist headers into /usr/local/cupti-headers-&lt;major.minor&gt; for the CUDA-13 build containers. However, the manywheel binary-build images only COPY /usr/local/cuda-&lt;ver&gt; out of the cuda build stage, so the staged headers never reach the final...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189324' target='_blank' rel='noopener noreferrer'>Pull Request #189324: [T278829272][Claude Code] Gate TLX flex_attention choices on tlx_mode</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: inductor, ciflow/inductor, meta-exported, ciflow/torchtitan</td><td><div class='wide'>Summary: X-link: facebookexperimental/triton#1961 15 caffe2/test/inductor:flex_attention tests broke on Hopper/Blackwell CI, root-caused to a missing gate against tlx-mode in flex_attention code. TLXInductorChoices.append_flex_attention_choices was the only method on the class that did not gate on config.triton.tlx_...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189321' target='_blank' rel='noopener noreferrer'>Pull Request #189321: [inductor] Enable per-region cudagraphs for invoke_subgraph via auto graph partition</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: inductor, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #189321 #189320 Building on nested Inductor configs, a region can set triton.cudagraphs to opt into or out of cudagraphs independently of the enclosing graph. When a region&#x27;s preference differs from the enclosing graph and graph partition is globally off, compile_fx enables...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189266' target='_blank' rel='noopener noreferrer'>Pull Request #189266: [ROCm][CI] Fix py3.15 ROCm manywheel numpy source build (pkg-config python3)</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: rocm, triaged, open source, ciflow/binaries_wheel</td><td><div class='wide'>Summary Python 3.15 has no prebuilt numpy wheel on PyPI yet, so pip builds numpy from source for every arch. meson&#x27;s Cython sanity check resolves the python3 pkg-config dependency to get Python include/link flags. The ROCm manylinux2_28-builder image ships the system Python 3.6 pkg-config file on the default search...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189291' target='_blank' rel='noopener noreferrer'>Pull Request #189291: Add an accurate Metal erfc for MPS (eager + inductor)</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, open source, ciflow/mps, module: inductor, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Splits the MPS backend portion out of #189234 (gh-187806) per review. Problem c10::metal::erfc is 1.0 - erf(x): once erf saturates in fp32 (x ~ 3.9), erfc returns 0.0, i.e. 100% relative error over the entire upper tail. Everything consuming it on MPS inherits the flush: eager torch.erfc / torch.special.erfc igammac...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188862' target='_blank' rel='noopener noreferrer'>Pull Request #188862: [inductor] Fix bool minimum/maximum crash in index propagation (#188862) (#188862)</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: cpu, ciflow/trunk, module: inductor, ciflow/inductor, meta-exported, ciflow/torchtitan</td><td><div class='wide'>Summary: Context torch.minimum/torch.maximum on boolean tensors crash under torch.compile(backend=&quot;inductor&quot;) with ValueError: The argument &#x27;False&#x27; is not comparable., while eager mode works. Reported in #188230 (repro: threshold -&gt; torch.eq -&gt; torch.empty_like -&gt; torch.minimum). The root cause is that inductor feed...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188643' target='_blank' rel='noopener noreferrer'>Pull Request #188643: [AOTI] Add external constants API for zero-copy weight sharing (#188643)</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor (aoti), meta-exported, ciflow/torchtitan</td><td><div class='wide'>Summary: Add a new AOTI C API, AOTInductorModelContainerCreateWithExternalConstants, that creates a model container with caller-provided weight tensors instead of loading from the .so file. The caller retains ownership (user_managed=true), enabling zero-copy weight sharing via CUDA IPC. The new ExternalConstantsTag...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189109' target='_blank' rel='noopener noreferrer'>Pull Request #189109: [profiler][cupti] Resolve a version-gated CUPTI header and conditionally install libclang for the field-id codegen</a></td><td>waiting on contributor</td><td>43</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>topic: not user facing</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189185 #189083 #188621 -&gt; #189109 The CUPTI field-id codegen (tools/gen_cupti_stubs.py, stacked on top) parses the CUPTI ABI header cupti_activity.h with libclang to emit the _cupti_stubs field-id catalogs. This commit adds the header resolver + the conditional libclang build...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/181559' target='_blank' rel='noopener noreferrer'>Pull Request #181559: [XPU] Enable _flash_attention_forward/_flash_attention_backward</a></td><td>waiting on CI/check fix</td><td>73</td><td><span class='chip-row'><span class='chip'>#2442</span></span></td><td>open source, intel priority, ciflow/trunk, ciflow/inductor, ciflow/xpu, release notes: inductor (aoti)</td><td><div class='wide'>Fix intel/torch-xpu-ops#2442 intel/torch-xpu-ops#2853 at::_flash_attention_forward and at::_flash_attention_backward are two ops which is one-one mapping to Tridao&#x27;s FlashAttention2 API. at::_scaled_dot_product_flash_attention and at::_scaled_dot_product_flash_attention_backward are two ops of SDPA FlashAttention ba...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/181726' target='_blank' rel='noopener noreferrer'>Pull Request #181726: [xpu][1/4]Implement scaled_mm_v2 for MXFP8/MXFP4/NVFP4 on XPU</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: cpu, triaged, open source, intel priority</td><td><div class='wide'>Extend the XPU scaled_mm_v2 dispatch to support MXFP8 (BlockWise1x32 with e8m0fnu scales), MXFP4 (BlockWise1x32 with e8m0fnu scales on packed FP4 data), and NVFP4(BlockWise1x16 with float8_e4m3fn scales on packed FP4 data) Updated after blockwise FP8 support #173630 is merged. PR Stack: Since I don&#x27;t have ghstack pe...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189051' target='_blank' rel='noopener noreferrer'>Pull Request #189051: [dynamo] Reject keyword args to set()/frozenset() constructors</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, ciflow/trunk, module: dynamo, ciflow/inductor, release notes: dynamo, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #185424 #189053 #189052 #189022 -&gt; #189051 #189021 set(a=1) / set().init(a=1) silently returned an empty set under Dynamo instead of raising TypeError. BuiltinVariable.call_set / call_frozenset checked the keyword arguments only after a len(args) == 0 early-return, so the zero-...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189317' target='_blank' rel='noopener noreferrer'>Pull Request #189317: [doc] Add accelerator test reuse guide (#181474)</a></td><td>waiting on CI/check fix</td><td>73</td><td><span class='chip-row'><span class='chip'>#181474</span></span></td><td>open source, topic: not user facing</td><td><div class='wide'>Adds a documentation page for the Accelerator Integration series explaining how out-of-tree backend developers can reuse PyTorch&#x27;s built-in test cases. Covers: bypass_device_restrictions for @OnlyOn decorator bypass (#177248) op_skips / op_decorators for operator-level skipping (#177256) skipped_testcases for test c...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188632' target='_blank' rel='noopener noreferrer'>Pull Request #188632: Fix Dynamo opaque object staticmethod guards</a></td><td>ready for maintainer decision</td><td>112</td><td><span class='chip-row'><span class='chip'>#188544</span></span></td><td>topic: not user facing, module: dynamo, ciflow/inductor, ciflow/torchtitan, agentic, needs-review</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #188632 Value opaque objects are represented as TorchScriptObjectVariable, whose constructor accepted a source but did not pass it to UserDefinedObjectVariable. That left the base class without the instance source it needs to derive cls_source for descriptor lookup. When Dyn...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189325' target='_blank' rel='noopener noreferrer'>Pull Request #189325: [Dynamo] Fix self-referential bw_compiler on backend re-entry</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, module: dynamo, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Why When one aot_autograd backend compiles a second, distinct graph, AotAutograd.__call__ re-enters with bw_compiler already wrapped. The _is_wrapped_bw_compiler branch then ran bw_compiler.compiler_fn = bw_compiler, grafting a self-referential attribute (and needing a type: ignore) instead of leaving the wrapper un...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/187404' target='_blank' rel='noopener noreferrer'>Pull Request #187404: [inductor][cutlass] Normalize reshaped external reads in EVT epilogue fusion</a></td><td>ready for maintainer decision</td><td>45</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, topic: not user facing, module: inductor, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #187404 Motivation Running the MobileBertForMaskedLM HuggingFace benchmark on XPU with CUTLASS epilogue fusion crashed during codegen: RuntimeError: Dimension mismatch between accum(1, 16384, 128), buf19(128, 128, 128) CUTLASS EVT treats the 2D GEMM output (M, N) = (16384, 1...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/179286' target='_blank' rel='noopener noreferrer'>Pull Request #179286: Use C++20 concepts where it improves readability</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, Merged, NNC, Reverted, ciflow/trunk, release notes: jit, topic: not user facing, ciflow/inductor, ci-no-td, no-runner-experiments, ciflow/vllm</td><td><div class='wide'>C++20 introduces concepts which can simplify some template code. In this PR, I have changed some instances of enable_if with concepts where it&#x27;s not hard to verify correctness. #176662 cc @EikanWang @jgong5</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/178393' target='_blank' rel='noopener noreferrer'>Pull Request #178393: [DO NOT MERGE] Dynamo 3.15 support</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, release notes: releng, module: dynamo, ciflow/inductor, ciflow/dynamo, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #178393 #179094 #188029 #187631 #189306</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189249' target='_blank' rel='noopener noreferrer'>Pull Request #189249: [Dynamo] Remove dead code in tvm and common backends</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, open source, ciflow/trunk, topic: not user facing, module: dynamo, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Why Two branches in torch/_dynamo/backends can never execute. tvm() assigns options a default when it is None, then immediately re-checks it for None. AotAutograd.__call__ assigns use_fallback = False and then branches on it, so the fallback and its not_ok counter increment are dead. How Drop the redundant if option...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188742' target='_blank' rel='noopener noreferrer'>Pull Request #188742: [ROCm][inductor] Enable scaled grouped GEMM Triton lowering</a></td><td>waiting on contributor</td><td>43</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: rocm, triaged, open source, ciflow/trunk, release notes: rocm, release notes: linalg_frontend, module: inductor, ciflow/inductor, release notes: inductor, ciflow/rocm-mi300, ciflow/h100, ciflow/rocm-mi355, ciflow/b200, ciflow/torchtitan, ciflow/rocm, ciflow/rocm-mi350</td><td><div class='wide'>Summary Build on the ROCm grouped GEMM Triton lowering change so legacy aten._scaled_grouped_mm can use the shared scaled grouped Triton template on ROCm. Gate ROCm scaled grouped GEMM Triton selection to FP8-capable hardware and the expected FP8 encoding (fnuz on gfx94, OCP FP8 on newer FP8-capable arches). Enable...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189299' target='_blank' rel='noopener noreferrer'>Pull Request #189299: [MemoryViz] Decimate over-long stacked-area polygons so the summarize…</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td></td><td><div class='wide'>…d band renders The &quot;Active Memory Timeline&quot; / &quot;Allocated Memory (incl. Private Pools)&quot; views build one stacked-area per allocation series and advance a point onto each live series at every alloc/free action. A series that stays live across the whole trace therefore gains one point per timestep. The global summarize...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189311' target='_blank' rel='noopener noreferrer'>Pull Request #189311: [wip] Reduce hackery for Dynamo appeasement</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: dynamo, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): (to be filled) cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @kadeng @chauhang @amjames @jataylo @azahed98</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/186055' target='_blank' rel='noopener noreferrer'>Pull Request #186055: Add CUDA graph support for torch.while_loop</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, ciflow/trunk, release notes: cuda</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #186056 -&gt; #186055 Motivation is to allow us to handle some forms of dynamic shapes inside of a cuda graph. This does not necessarily improve performance, but it can make your entire workload cuda graph compatible, which is good for developer experience For example, if you want...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189088' target='_blank' rel='noopener noreferrer'>Pull Request #189088: [SymmMem] Place the signal pad at the front of symmetric memory allocations</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, release notes: distributed (c10d), release notes: distributed (symm_mem), ciflow/h100-symm-mem, ciflow/b200-symm-mem</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #187465 -&gt; #189088 We place the signal pad at the front of symmetric memory allocations in this PR. The purpose is to fix the potential signal pad pollution when we reuse the same VA for one block. For example, within one block if we first use that VA for a large tensor and the...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/187631' target='_blank' rel='noopener noreferrer'>Pull Request #187631: [dynamo] Fix segfault in list comprehension graph break</a></td><td>ready for maintainer decision</td><td>70</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, topic: not user facing, module: dynamo, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #178393 #179094 #188029 -&gt; #187631 #189306 Reusing the name of the iterator variable in a list comprehension that graph breaks to store the result causes a segfault (the stack underflows in the generated comprehension function). The current approach using bytecode templates is...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189305' target='_blank' rel='noopener noreferrer'>Pull Request #189305: Reapply &quot;Remove CUDA sync in torch.combinations to improve `torch.compile` support (#186595)&quot;</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, release notes: cuda, module: inductor, ciflow/inductor, ciflow/torchtitan, ciflow/dtensor</td><td><div class='wide'>Reapplies #186595 after Meta internal revert. Let me know if there are any failures that require fixing from my end. /cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @kadeng @muchulee8 @amjames @chauhang @aakhundov @coconutruben @jataylo...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188029' target='_blank' rel='noopener noreferrer'>Pull Request #188029: [dynamo] Add namespace for cell variables that collide with locals</a></td><td>ready for maintainer decision</td><td>45</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, topic: not user facing, module: dynamo, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #178393 #179094 -&gt; #188029 #187631 #189306 Cell variables that share a name with a local variable are stored separately in localsplus in cpython. This distinction is lost in dynamo, where all variables are stored in symbolic_locals. This doesn&#x27;t matter for most purposes, but be...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188521' target='_blank' rel='noopener noreferrer'>Pull Request #188521: Fix OrderingBarrier additional_buffer_deps key mismatch in control_deps lowering</a></td><td>waiting on CI/check fix</td><td>73</td><td><span class='chip-row'><span class='chip'>#188513</span></span></td><td>triaged, open source, topic: not user facing, module: inductor, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Summary Fix a key mismatch in control_deps_op_lowering that caused OrderingBarrier&#x27;s additional_buffer_deps to be silently dropped by the scheduler, breaking stream synchronization ordering. OrderingBarrier inherits from OperationBuffer, which has distinct buffer and operation names (assigned by register_buffer() an...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/183328' target='_blank' rel='noopener noreferrer'>Pull Request #183328: [special] Add arbitrary-order modified Bessel functions</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: cpu, triaged, open source, release notes: cpp, module: inductor, module: dynamo, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Issue Partially addresses #76324. Summary Adds arbitrary-order modified Bessel functions: torch.special.modified_bessel_i(x, nu) torch.special.modified_bessel_k(x, nu) This follows the existing torch.special pointwise operator pattern with: ATen operator registration TensorIterator CPU/CUDA kernels CUDA Jiterator su...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/186927' target='_blank' rel='noopener noreferrer'>Pull Request #186927: [MPS] Gemv kernels</a></td><td>waiting on CI/check fix</td><td>108</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, open source, topic: improvements, release notes: mps, ciflow/mps</td><td><div class='wide'>Gemv kernels. Draft till I put perf gains below on various machines and add tests. TODOs: Add tests Heuristics M2 base Heuristics M3 Ultra Heuristics M4 base Heuristics M5 Pro More heuristics (need help) Add perf benchmarks on M5 Pro Add perf benchmarks on M4 Add perf benchmarks on M3 Ultra Add perf benchmarks on M2...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/179094' target='_blank' rel='noopener noreferrer'>Pull Request #179094: [DO NOT MERGE] Add docker image for python 3.15</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, topic: not user facing, no-runner-experiments, ciflow/docker</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #178393 -&gt; #179094 #188029 #187631 #189306 Authored with Claude.</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/186252' target='_blank' rel='noopener noreferrer'>Pull Request #186252: [inductor] Fix truncdiv off-by-one on CUDA with _div_rn</a></td><td>waiting on CI/check fix</td><td>73</td><td><span class='chip-row'><span class='chip'>#184408</span></span></td><td>triaged, open source, ciflow/trunk, topic: not user facing, module: inductor, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Fixes #184408 Triton&#x27;s approximate fp32 division can produce results slightly below the true quotient, causing trunc(a / b) to be off by one when the quotient is near an integer. Use _div_rn (IEEE round-to-nearest) instead of plain div, matching the existing fix for the floor path. Change: trunc(div(a, b)) -&gt; trunc(...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189313' target='_blank' rel='noopener noreferrer'>Pull Request #189313: Support aten::cumprod ONNX export via CumProd (opset 26)</a></td><td>draft/noise</td><td>61</td><td><span class='chip-row'><span class='chip'>#102948</span></span></td><td>open source, release notes: onnx</td><td><div class='wide'>Fixing an Issue Before submitting, please review: The Ultimate Guide to PyTorch Contributions AI-Assisted Development policy Summary torch.cumprod could not be exported to ONNX (UnsupportedOperatorError: aten::cumprod). ONNX added the CumProd operator in opset 26, so this registers an implementation in the dynamo ex...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189286' target='_blank' rel='noopener noreferrer'>Pull Request #189286: Fix DeviceContext mode leaks from checkpoint recompute and set_default_device restore</a></td><td>ready for maintainer decision</td><td>80</td><td><span class='chip-row'><span class='chip'>#189133</span></span></td><td>release notes: autograd, topic: not user facing, module: inductor, ciflow/inductor, ciflow/torchtitan, mergedog, make tests great again</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #189286 #188573 #188299 #188100 Fixes #189133 This PR addresses flaky tests by fixing some DeviceContext mode leaks that may randomly affect other tests (e.g. the vmap test disabled by #189133). There are 3 issues addressed by this PR: TestAutograd and TestCompiledAutogradOp...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189306' target='_blank' rel='noopener noreferrer'>Pull Request #189306: [dynamo, 3.15] Fix test_bytecode_utils</a></td><td>ready for maintainer decision</td><td>45</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, topic: not user facing, module: dynamo, ciflow/inductor</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #178393 #179094 #188029 #187631 -&gt; #189306 Two changes to update this for 3.15 co_lnotab has been deprecated since 3.10, and was removed in 3.15. It was replaced by co_lines, which we can use instead LOAD_COMMON_CONSTANT can load None in 3.15, so is now used for functions that...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189297' target='_blank' rel='noopener noreferrer'>Pull Request #189297: [CI] Pin quack to torch&#x27;s vendored SHA to fix B200 smoke cutlass import crash</a></td><td>draft/noise</td><td>13</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>ciflow/trunk, topic: not user facing, ciflow/b200</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #189297 The B200 smoke job&#x27;s test_flex_flash::test_hierarchical_indexing_4d crashes at runtime with ModuleNotFoundError: No module named &#x27;cutlass._mlir_helpers&#x27;. The compiled flex-attention CuTeDSL kernel imports flash_attn.cute, which pulls in quack, whose recent releases d...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189319' target='_blank' rel='noopener noreferrer'>Pull Request #189319: Enable flake8-bugbear B007 (unused loop control variable)</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, release notes: quantization, release notes: releng, ciflow/mps, module: inductor, module: dynamo, ciflow/inductor, release notes: distributed (dtensor), module: compiled autograd, ciflow/h100-symm-mem, ciflow/inductor-pallas, ciflow/torchtitan, ciflow/dtensor</td><td><div class='wide'>Summary Enables flake8-bugbear B007 (unused loop control variable), one of the remaining bugbear codes still suppressed in the ignore list of pyproject.toml. Part of #106571. This removes B007 from the ignore list and fixes all 169 violations across 110 files so the lint is enforced going forward. Approach Most fixe...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188600' target='_blank' rel='noopener noreferrer'>Pull Request #188600: [ROCm][inductor] Support grouped GEMM Triton lowering</a></td><td>waiting on contributor</td><td>43</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: rocm, triaged, open source, release notes: rocm, release notes: linalg_frontend, module: inductor, ciflow/inductor, release notes: inductor, ciflow/rocm-mi300, ciflow/h100, ciflow/b200, ciflow/torchtitan</td><td><div class='wide'>Summary Allow Inductor&#x27;s grouped GEMM Triton lowering on ROCm instead of gating it to NVIDIA SM90+ only. Match ROCm ATen extern output layout by using contiguous grouped GEMM output strides on HIP while preserving CUDA&#x27;s TMA-aligned padded strides. Enable existing grouped GEMM compile tests on ROCm. Test plan python...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/181781' target='_blank' rel='noopener noreferrer'>Pull Request #181781: [dynamo] turn canonicalize_output_graph_node_order on by default</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: dynamo, ciflow/inductor, ciflow/torchtitan, ciflow/dtensor</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #181781 #188176 #188175 #181780 cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @kadeng @chauhang @amjames @Lucaskabela @jataylo @azahed98</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188137' target='_blank' rel='noopener noreferrer'>Pull Request #188137: [inductor] Decompose shard_dim_alltoall via regular decomposition</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>triaged, open source, ciflow/trunk, module: inductor, ciflow/inductor, release notes: distributed (dtensor), release notes: inductor, ciflow/torchtitan, ciflow/dtensor</td><td><div class='wide'>Summary: add an opt-in regular Inductor decomposition for _dtensor.shard_dim_alltoall into layout ops plus _c10d_functional.all_to_all_single/wait_tensor decompose only when the dims are valid, the dtype is supported, the process group size is resolvable, and the local shard dim is evenly divisible by the process gr...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189312' target='_blank' rel='noopener noreferrer'>Pull Request #189312: [CUDA][cuBLAS] Change cuBLAS default workspace size for SM 11.0 to 32 MiB</a></td><td>waiting on CI/check fix</td><td>108</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: cuda, module: cublas, open source, topic: not user facing</td><td><div class='wide'>Thor is closest to datacenter blackwell, and current defaults cause CUBLAS_STATUS_NOT_SUPPORTED in e.g., python test/test_linalg.py TestLinalgCUDA.test_matmul_mv_cuda_float32 I assume there are basically internal users of SM 11.0 that would complain? CC @ngimel authored with codex cc @ptrblck @msaroufim @tinglvv @nW...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189301' target='_blank' rel='noopener noreferrer'>Pull Request #189301: [ROCm] Bump AOTriton to 0.13b</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: rocm, open source, topic: not user facing, ciflow/rocm-mi300</td><td><div class='wide'>Notable new features compared with 0.12.1b: Debug printing can be enabled with env var AOTRITON_DEBUG_LEVEL, without re-compiling We use the same logging level as of AMD_LOG_LEVEL CAVEAT: AOTriton itself does not read AMD_LOG_LEVEL Set AOTRITON_DEBUG_LEVEL=5 to print everything BREAKING .aks2 files are packed into u...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189022' target='_blank' rel='noopener noreferrer'>Pull Request #189022: [dynamo] Support itertools.count keyword args and repr</a></td><td>ready for maintainer decision</td><td>45</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>open source, module: dynamo, ciflow/inductor, release notes: dynamo, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #185424 #189053 #189052 -&gt; #189022 #189051 #189021 Two itertools.count object-protocol gaps, both mirroring CPython Modules/itertoolsmodule.c: Construction: the count branch had a not kwargs guard, so the keyword forms count(start=..)/count(step=..) fell through to a graph brea...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188006' target='_blank' rel='noopener noreferrer'>Pull Request #188006: [ROCm] Fix torch.pdist correctness on ROCm; re-enable test_pdist_norm_large</a></td><td>waiting on contributor</td><td>53</td><td><span class='chip-row'><span class='chip'>#168868</span></span></td><td>module: rocm, triaged, open source, release notes: rocm</td><td><div class='wide'>Fixes #168868 Investigation summary Test environment PyTorch version: 2.13.0a0+git75d18bb Hip version: 7.2.53211 GPU 0 name: AMD Instinct MI355X Command to reproduce $ PYTORCH_TEST_WITH_ROCM=1 pytest test/test_torch.py -k &quot;test_pdist_norm_large&quot; -v Default behaviour test/test_torch.py::TestTorchDeviceTypeCPU::test_p...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189092' target='_blank' rel='noopener noreferrer'>Pull Request #189092: [dynamo] reduce call_obj_hasattr overrides</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: dynamo, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #189092 #189091 #187707 #187532 #187531 #187469 #187468 Fix object_generic_getattr step 7 to raise ObservedAttributeError instead of _UnhandledDescriptorError when an attribute genuinely doesn&#x27;t exist after exhaustive MRO search and getattr fallback. This is semantically cor...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/187707' target='_blank' rel='noopener noreferrer'>Pull Request #187707: [dynamo] remove GetAttrVariable</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: dynamo, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189092 #189091 -&gt; #187707 #187532 #187531 #187469 #187468 GetAttrVariable was a catch-all fallback that deferred attribute access by wrapping it in a proxy. It made it impossible to distinguish &quot;attribute not found&quot; from &quot;attribute found but unhandled,&quot; masking real errors (e....</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189091' target='_blank' rel='noopener noreferrer'>Pull Request #189091: [dynamo] guard getattr() builtin constant-fold fallback</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: dynamo, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189092 -&gt; #189091 #187707 #187532 #187531 #187469 #187468 GetAttrBuiltinVariable.call_function has a constant-fold fallback that fires when getattro_impl raises Unsupported and all args are python constants. It was calling VariableTracker.build(tx, result) without a source, so...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188978' target='_blank' rel='noopener noreferrer'>Pull Request #188978: swap isinstance(t, FakeTensor) to is_fake</a></td><td>waiting on CI/check fix</td><td>73</td><td><span class='chip-row'><span class='chip'>#188694</span></span></td><td>ciflow/trunk, module: inductor, module: dynamo, ciflow/inductor, release notes: export, release notes: inductor (aoti), ciflow/torchtitan, ciflow/docker</td><td><div class='wide'>closed #188694 bc need to stack smth on top copying pr summary over... in preparation for C++ FakeTensor migration, i am landing this change first to change isinstance(t, FakeTensor) to is_fake_tensor and adding a check for C++ FakeTensor to the is_fake_tensor function. but since C++ faketensor is not being used yet...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/187531' target='_blank' rel='noopener noreferrer'>Pull Request #187531: [dynamo] enable generic attribute mutation on non-UDOV VTs</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: dynamo, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189092 #189091 #187707 #187532 -&gt; #187531 #187469 #187468 Adds a get_value_for_setattr() opt-in hook on the base VariableTracker that enables attribute mutation (setattr/getattr/hasattr) for VTs beyond UserDefinedObjectVariable. The mechanism: VTs override get_value_for_setatt...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/187532' target='_blank' rel='noopener noreferrer'>Pull Request #187532: [dynamo] fix __getattr__ and __getattribute__ dunder semantics</a></td><td>draft/noise</td><td>51</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: dynamo, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189092 #189091 #187707 -&gt; #187532 #187531 #187469 #187468 Previously, both obj.__getattr__(&quot;x&quot;) and obj.__getattribute__(&quot;x&quot;) in call_method dispatched through getattro_impl, which does full GenericGetAttr including the __getattr__ fallback. This was semantically wrong for bot...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/187468' target='_blank' rel='noopener noreferrer'>Pull Request #187468: [dynamo] add break_graph_if_unsupported to LOAD_ATTR</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>topic: not user facing, module: dynamo, ciflow/inductor, keep-going, ciflow/dynamo, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189092 #189091 #187707 #187532 #187531 #187469 -&gt; #187468 LOAD_ATTR previously relied on the step() fallback for graph breaks, which restarts from the most recent empty-stack checkpoint. This is less precise than other opcodes like STORE_ATTR and CALL that use break_graph_if_u...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/187469' target='_blank' rel='noopener noreferrer'>Pull Request #187469: [dynamo] extract UDOV hooks: lookup_instance_dict and call_getattr_fallback</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>topic: not user facing, module: dynamo, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189092 #189091 #187707 #187532 #187531 -&gt; #187469 #187468 Extract two chunks of UDOV&#x27;s generic_getattr into hook overrides, matching the hook interface on base VariableTracker: lookup_instance_dict (step 3): instance dict lookup including side-effects awareness (INSTANCE_DICT...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/181780' target='_blank' rel='noopener noreferrer'>Pull Request #181780: [dynamo] set canonicalize_output_graph_node_order=True for export tests</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>topic: not user facing, module: dynamo, ciflow/inductor, keep-going, ciflow/torchtitan, ciflow/dtensor</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #181781 #188176 #188175 -&gt; #181780 cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @kadeng @chauhang @amjames @jataylo @azahed98</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188175' target='_blank' rel='noopener noreferrer'>Pull Request #188175: [dynamo] fix autograd cache and split_module tests under canonicalization</a></td><td>ready for maintainer decision</td><td>70</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: inductor, module: dynamo, ciflow/inductor</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #181781 #188176 -&gt; #188175 #181780 Two tests opted out of canonicalize_output_graph_node_order because they were sensitive to placeholder ordering. This PR fixes both so the opt-outs can be removed. test_multi_graph_specialization failed because canonicalization reorders placeh...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189292' target='_blank' rel='noopener noreferrer'>Pull Request #189292: Fix dead ExecuTorch XNNPACK doc link in mobile_optimizer</a></td><td>ready for maintainer decision</td><td>70</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>ciflow/trunk, topic: not user facing</td><td><div class='wide'>Stack from ghstack (oldest at bottom): -&gt; #189292 The nightly Link checks / lint-urls / lint job fails on a genuine 404: the XNNPACK backend doc referenced from mobile_optimizer.md was renamed by ExecuTorch from native-delegates-executorch-xnnpack-delegate.html (now 404) to backends-xnnpack.html (200). Point the lin...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188176' target='_blank' rel='noopener noreferrer'>Pull Request #188176: [dynamo] fix test_perf.py tests under canonicalization</a></td><td>ready for maintainer decision</td><td>45</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>topic: not user facing, module: inductor, ciflow/inductor</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #181781 -&gt; #188176 #188175 #181780 Two tests in test_perf.py opted out of canonicalize_output_graph_node_order unnecessarily. This removes both opt-outs. test_cat_pointwise passes deterministically with canonicalization -- the numel counts are unchanged because canonicalization...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189190' target='_blank' rel='noopener noreferrer'>Pull Request #189190: flex_gemm: extend feed-main to trailing pointwise and axis-1 fragments</a></td><td>waiting on CI/check fix</td><td>63</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: inductor, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189316 #189315 #189314 -&gt; #189190 #189188 #188739 #188112 #188470 #188469 #188468 The feed-main matcher now recurses through trailing shape-preserving pointwise nodes after the un-grouping view, so quant casts like .to(float8_e4m3fn) compose with physical axis-0 feeds. Axis-1...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/189188' target='_blank' rel='noopener noreferrer'>Pull Request #189188: flex_gemm: support scalar captured epilogue args</a></td><td>ready for maintainer decision</td><td>45</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: inductor, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189316 #189315 #189314 #189190 -&gt; #189188 #188739 #188112 #188470 #188469 #188468 Add a fourth captured-epilogue-arg kind &quot;scalar&quot; beside tile/row/col for [1, 1] read-only tensors. This unlocks native float8 tensorwise quant epilogues, e.g. (acc * s).to(torch.float8_e4m3fn) wi...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188739' target='_blank' rel='noopener noreferrer'>Pull Request #188739: flex_gemm: add quant lowerings</a></td><td>draft/noise</td><td>13</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: inductor, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189316 #189315 #189314 #189190 #189188 -&gt; #188739 #188112 #188470 #188469 #188468</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188112' target='_blank' rel='noopener noreferrer'>Pull Request #188112: flex_gemm: feed local reductions into epilogues</a></td><td>draft/noise</td><td>13</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>topic: not user facing, skip-pr-sanity-checks, module: inductor, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189316 #189315 #189314 #189190 #189188 #188739 -&gt; #188112 #188470 #188469 #188468 Let dense aten.mm FlexGEMM epilogues feed a grouped local reduction back into the main output expression (and optionally store it), e.g. acc / acc.view(-1, group, N).sum(1) or centered-mean/varia...</div></td></tr><tr><td><a href='https://github.com/pytorch/pytorch/pull/188470' target='_blank' rel='noopener noreferrer'>Pull Request #188470: flex_gemm: extend local reduce aux groups</a></td><td>draft/noise</td><td>13</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td>module: inductor, ciflow/inductor, ciflow/torchtitan</td><td><div class='wide'>Stack from ghstack (oldest at bottom): #189316 #189315 #189314 #189190 #189188 #188739 #188112 -&gt; #188470 #188469 #188468 Extend compressed local-reduce aux outputs beyond one 32-lane TensorSSA fragment: axis-1 groups above the fragment width (CTA-subtile groups such as 64/128) and axis-0 M-groups now lower through...</div></td></tr></tbody></table></div></section></div>

data/relationship_edges.html

@@ -0,0 +1,75 @@
+
+<style>
+:root { color-scheme: light; }
+.bi-wrap { font-family: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; color: #1f2328 !important; }
+.bi-wrap h1, .bi-wrap h2, .bi-wrap h3, .bi-wrap p, .bi-wrap strong, .bi-wrap td, .bi-wrap th, .bi-wrap div { color: #1f2328 !important; }
+.bi-band { border: 1px solid #d8dee4; border-radius: 8px; background: #ffffff !important; padding: 16px; margin: 0 0 12px; }
+.bi-title { display: flex; gap: 12px; align-items: flex-start; justify-content: space-between; margin-bottom: 12px; }
+.bi-title h1, .bi-title h2 { margin: 0; line-height: 1.15; letter-spacing: 0; }
+.bi-title h1 { font-size: 1.55rem; }
+.bi-title h2 { font-size: 1.12rem; }
+.bi-title p { margin: 4px 0 0; color: #57606a; max-width: 820px; }
+.status-pill { border-radius: 999px; padding: 5px 9px; font-size: .78rem; font-weight: 700; background: #dafbe1; color: #116329 !important; white-space: nowrap; }
+.metric-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(138px, 1fr)); gap: 8px; }
+.metric { border: 1px solid #d0d7de; border-radius: 8px; padding: 10px; background: #f6f8fa !important; min-height: 86px; }
+.metric-value { font-size: 1.45rem; font-weight: 800; line-height: 1.1; color: #0969da !important; }
+.metric-label { margin-top: 4px; font-size: .84rem; font-weight: 700; }
+.metric-note { margin-top: 2px; font-size: .76rem; color: #57606a !important; }
+.callout-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 8px; }
+.callout { border-left: 4px solid #0969da; background: #f6f8fa !important; border-radius: 8px; padding: 10px 12px; }
+.callout:nth-child(2) { border-left-color: #1a7f37; }
+.callout:nth-child(3) { border-left-color: #bf8700; }
+.callout:nth-child(4) { border-left-color: #cf222e; }
+.callout strong { display: block; margin-bottom: 4px; }
+.group-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(310px, 1fr)); gap: 10px; }
+.group-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.group-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.meta-row { display: flex; flex-wrap: wrap; gap: 6px; margin: 6px 0; }
+.chip-row { display: inline-flex; flex-wrap: wrap; gap: 5px; align-items: center; vertical-align: middle; }
+.chip { display: inline-flex; align-items: center; border: 1px solid #d0d7de; border-radius: 999px; padding: 2px 7px; font-size: .78rem; background: #f6f8fa !important; color: #24292f !important; }
+.chip.bucket { border-color: #9a6700; background: #fff8c5 !important; }
+.chip.kind { border-color: #0969da; background: #ddf4ff !important; }
+.workflow-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(260px, 1fr)); gap: 10px; }
+.workflow-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.workflow-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.workflow-card .move-count { font-weight: 800; color: #0969da !important; }
+.move-list { display: grid; gap: 10px; }
+.move-row { border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+.move-row summary { list-style: none; cursor: pointer; padding: 12px; }
+.move-row summary::-webkit-details-marker { display: none; }
+.move-summary { display: grid; grid-template-columns: minmax(190px, 1fr) minmax(260px, 1.45fr) minmax(260px, 1.45fr) minmax(150px, .8fr) minmax(250px, 1.35fr) auto; gap: 16px; align-items: start; }
+.move-title, .move-field { min-width: 0; }
+.move-title h3 { margin: 0 0 4px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.move-field span { display: block; color: #57606a !important; font-size: .72rem; font-weight: 800; text-transform: uppercase; letter-spacing: 0; }
+.move-field p { margin: 3px 0 0; font-size: .84rem; line-height: 1.35; overflow-wrap: anywhere; }
+.move-field .chip, .move-title { overflow-wrap: anywhere; }
+.view-bundles { align-self: center; justify-self: end; border: 1px solid #000000; border-radius: 999px; color: #000000 !important; background: #ffffff !important; padding: 4px 9px; font-size: .78rem; font-weight: 800; white-space: nowrap; }
+.move-row[open] .view-bundles { background: #f6f8fa !important; }
+.move-expanded { border-top: 1px solid #d8dee4; padding: 0 12px 12px; }
+.bundle-row { border-top: 1px solid #eef1f4; padding: 12px 0; }
+.bundle-row:first-child { border-top: none; }
+.bundle-row h4 { margin: 0 0 6px; font-size: .94rem; line-height: 1.25; letter-spacing: 0; }
+.bundle-scope { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 7px; margin-top: 8px; }
+.evidence-list { margin: 8px 0 0 18px; padding: 0; color: #57606a !important; font-size: .82rem; line-height: 1.35; }
+.evidence-list li { margin: 3px 0; }
+.muted { color: #57606a !important; }
+.evidence { font-size: .84rem; color: #57606a !important; }
+.scroll-table { overflow-x: auto; border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+table.bi-table { width: 100%; border-collapse: collapse; font-size: .86rem; }
+.bi-table th { text-align: left; background: #f6f8fa !important; color: #57606a !important; font-size: .78rem; border-bottom: 1px solid #d8dee4; padding: 8px; white-space: nowrap; }
+.bi-table td { border-bottom: 1px solid #eef1f4; padding: 8px; vertical-align: top; min-width: 86px; }
+.bi-table tr:last-child td { border-bottom: none; }
+.wide { min-width: 280px; }
+.tight { max-width: 180px; }
+a { color: #0969da !important; text-decoration: none; }
+a:hover { text-decoration: underline; }
+.bi-wrap, .bi-wrap * { color: #000000 !important; }
+@media (max-width: 720px) {
+  .bi-title { display: block; }
+  .status-pill { display: inline-block; margin-top: 8px; }
+  .group-grid { grid-template-columns: 1fr; }
+  .move-summary { grid-template-columns: 1fr; }
+  .view-bundles { justify-self: start; }
+}
+</style>
+<div class='bi-wrap'><section class='bi-band'><div class='bi-title'><div><h2>Relationship Edges</h2><p>Why items were connected. Excerpts are short and hashed; raw API payloads stay local.</p></div></div><div class='scroll-table'><table class='bi-table'><thead><tr><th>Kind</th><th>Source</th><th>Target</th><th>Confidence</th><th>Evidence</th><th>Evidence Hash</th></tr></thead><tbody><tr><td>references</td><td>issue #141287</td><td>issue #77764</td><td>medium</td><td><div class='wide'>lace to list and track work on adding support to new ops for the MPS backend. Most requested ops (extracted from comments to this issue and #77764) PyTorch MPS Ops Project : Project to track all the ops for MPS backend. There are a very large number of operators in pytorch and...</div></td><td>27287ca4e6e3b6b4ece6528bc77edeef25116a884123cb7ee2d31af4ad80d401</td></tr><tr><td>references</td><td>issue #189287</td><td>issue #188492</td><td>medium</td><td><div class='wide'>ibution needs confirmation (it is by failing-since alignment, not root-caused). Same #188251 bump has produced other confirmed regressions: #188492 (swin inductor accuracy) and #188841 (compile + float8 training numerics). Filed by pt2-oncall triage from HUD signal. cc @ptrblc...</div></td><td>9698400f2da6521759a9575e45c7f88418e35787b03ed55b7631eb8e3bc3e359</td></tr><tr><td>references</td><td>issue #189287</td><td>issue #188841</td><td>medium</td><td><div class='wide'>failing-since alignment, not root-caused). Same #188251 bump has produced other confirmed regressions: #188492 (swin inductor accuracy) and #188841 (compile + float8 training numerics). Filed by pt2-oncall triage from HUD signal. cc @ptrblck @msaroufim @eqy @tinglvv @nWEIdia @...</div></td><td>25eed4d5223772761e053cdbd2af3b7cce89b4147c534b321bd6700d9a93f1d2</td></tr><tr><td>references</td><td>issue #188756</td><td>issue #84520</td><td>medium</td><td><div class='wide'>09 notes flip is the only remaining MPSGraph user in Indexing.mm, but other files still lower gather/scatter-like ops to MPSGraph. Related: #84520 (MPS 32-bit limits umbrella — but here nothing asserts, it corrupts silently), #144824 (OOB indexing returns 0), #154235 (missing...</div></td><td>5ddb5dfa430472e79585e537d3eb57af3316c88b726367c07226c054df366a82</td></tr><tr><td>references</td><td>issue #188756</td><td>issue #144824</td><td>medium</td><td><div class='wide'>slicing — the device copy is intact; only the gather is broken. All indices involved are valid; this is not the OOB-index case of #154235 / #144824, though the failure surface is the same &quot;OOB reads return 0&quot; behavior. Root cause (as far as I can tell) In 2.12.0, index_select_...</div></td><td>f092cb9517d78cce4120db3a3afe47aaf2128377257db42ad51791c110aabff3</td></tr><tr><td>references</td><td>issue #189270</td><td>pr #187328</td><td>medium</td><td><div class='wide'>Summary kernel_information.json (#160540, extended in #183952) and the profiler-timeline provenance (#186230, #187328) map each generated kernel to a set of pre/post-grad node names, stack traces, and now extern I/O shapes. What is still missing is the phys</div></td><td>cf1d27af3d74794a50a5e5b334efdcaf82727003595ee820cb256914508b80a0</td></tr><tr><td>references</td><td>issue #129131</td><td>issue #97894</td><td>medium</td><td><div class='wide'>/_inductor/fx_passes/pad_mm.py#L636 https://github.com/pytorch/pytorch/blob/main/torch/_dynamo/convert_frame.py#L172 (Logic refers to issue #97894) Minified repro import torch @torch.compile def some_fn(x): return torch.pow(x, 0.1) / 2 with torch.device(&quot;cpu&quot;): x = torch.ones(...</div></td><td>0580bf147fcef88bd580d61c801894d53b3a463de97ba138037606fe24aed172</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #182109</td><td>medium</td><td><div class='wide'>s; OpInfo / common_methods_invocations.py remains the broad semantic owner. Revisit LayerNorm backward after the active reduce/loss stacks. #182109 already has feedback asking for clearer rationale, benchmark script, and before/after results. If revived, split it into smaller...</div></td><td>9f2d6dee8fd1f5c80583f3267a977d73119c9166fd00f0623328c3eafee36435</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #182731</td><td>medium</td><td><div class='wide'>tern used by the leaves. No user-visible behavior change; this is the cross-op infrastructure reviewers should see first. Scalar read leaf: #182731. Rebase on the UMA base helper. Scope: _local_scalar_dense / .item() / scalar truthiness direct CPU read. MPS-&gt;CPU copy leaf: #18...</div></td><td>931ba6c553643a760ccb443d494823a0c03e5d2d705a5ab71b93f5a266fa0aa4</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #182736</td><td>medium</td><td><div class='wide'>leaf: #182731. Rebase on the UMA base helper. Scope: _local_scalar_dense / .item() / scalar truthiness direct CPU read. MPS-&gt;CPU copy leaf: #182736. Rebase on the UMA base helper. Scope: contiguous same-dtype copy_from_mps_ / .to(&quot;cpu&quot;) / .numpy() direct read. CPU-&gt;MPS copy le...</div></td><td>dd905b22f5c769d19280e2f72feb508eebe5b0c978eae6dcd77a6820a259b9d2</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #182749</td><td>medium</td><td><div class='wide'>182736. Rebase on the UMA base helper. Scope: contiguous same-dtype copy_from_mps_ / .to(&quot;cpu&quot;) / .numpy() direct read. CPU-&gt;MPS copy leaf: #182749. Rebase on the UMA base helper. Scope: contiguous same-dtype CPU-&gt;MPS direct write, with non-blocking/non-contiguous/dtype-cast f...</div></td><td>39ad423213d6c30d7edcd3619f8789b5914d4a0303b1c24a70bc34c5c6107928</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #182791</td><td>medium</td><td><div class='wide'>elper. Scope: contiguous same-dtype CPU-&gt;MPS direct write, with non-blocking/non-contiguous/dtype-cast fallbacks preserved. Fill/zero leaf: #182791. Rebase after the write-side copy leaf. Scope: dense zero_/fill_ via memset/std::fill_n, with negative-zero and unsupported dtype...</div></td><td>8ce2999c8c1c71211ab727a25ada447ebba73c20e3518765eae8f316d6715822</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #183688</td><td>medium</td><td><div class='wide'>182791 draft/pending rework Rebase onto the UMA base/write-side direction; do not review as a standalone final PR. MPS capture/replay stack #183688 is much larger than a reviewable PyTorch PR and should be treated as a reference branch to split, not a final PR. Order: Capture...</div></td><td>9df4bd150ddf31bc316ce476fd7f829ad13277ecb7bace954fc85b13c855a01c</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #187456</td><td>medium</td><td><div class='wide'>lit at the last-dim vs non-last-dim boundary and stacked as cumulative-leaf forks on main (external contributor, no ghstack): softmax core (#187456, 52a3223): native last-dim _softmax/_softmax_backward (single-row + 8-wide-half + looped online-softmax); non-last-dim and huge-r...</div></td><td>345ffb2dce60cb2b34d124fed94231118ec9415c1281c040878c8ed45ac7f037</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #187457</td><td>medium</td><td><div class='wide'>-dim-only gate. Residual dim=0-backward tail is documented (MPSGraph&#x27;s fused reduce still wins there; escape hatch covers it). log_softmax (#187457, 3d6f962): native log_softmax stacked on the softmax kernels. cross_entropy (#187458, ee60969): fused 2D cross-entropy Metal kern...</div></td><td>4693297ad47a1ed0ffaf041a3a3b91720068b11cbbadf04b36ed80bb5e106daa</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #187458</td><td>medium</td><td><div class='wide'>antiations each), not chained on each other. Independent loss-style work: #187559 (nll_loss, review comments addressed at fc26ad2e55e3) and #187458 (cross_entropy) do not fit the same elementwise fused-loss template. Softmax / log_softmax / cross_entropy: refreshed and opened...</div></td><td>2d0d3aa86c56419e390e88ebb84b2aef1fe2f403d789ea9a8a900d66e98ac76a</td></tr><tr><td>competes with</td><td>issue #187455</td><td>pr #187553</td><td>medium</td><td><div class='wide'>into the existing TestMetalLibrary.test_reduction_utils owner surface instead of adding a standalone regression. Land the loss-stack base: #187553 (mse_loss). Scope: shared fused-loss infrastructure (FusedLossParams, fused_loss_pass1, fused_loss_bwd, fused_loss_reduce, TensorI...</div></td><td>5ead698bdfb6499f55eff3d9865a7aca19a56d5d71a0e0c3761c5d982d27dc30</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #187556</td><td>medium</td><td><div class='wide'>ases); OpInfo consistency (including the f16 grad test) remains the broad semantic owner. Stack: rebuild #187557 (binary_cross_entropy) and #187556 (smooth_l1_loss/huber_loss) as sibling leaves on the final #187553 (small op functor + instantiations each), not chained on each...</div></td><td>919f58fd26fa22c48388ef0ce0cc76ab80ab406ed90785dc78ba0b4a212ea983</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #187557</td><td>medium</td><td><div class='wide'>d, mixed dtypes, out-variant edge cases); OpInfo consistency (including the f16 grad test) remains the broad semantic owner. Stack: rebuild #187557 (binary_cross_entropy) and #187556 (smooth_l1_loss/huber_loss) as sibling leaves on the final #187553 (small op functor + instant...</div></td><td>4b4c4b9ff1073e29e28405a99aff70564283a96f27ef4b0c1d4d4b4699486118</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #187559</td><td>medium</td><td><div class='wide'>s) as sibling leaves on the final #187553 (small op functor + instantiations each), not chained on each other. Independent loss-style work: #187559 (nll_loss, review comments addressed at fc26ad2e55e3) and #187458 (cross_entropy) do not fit the same elementwise fused-loss temp...</div></td><td>c2e9079116af0b2664b8d9cf1704ea257b819b8f57dd55fd0c116be7e28a284d</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #187787</td><td>medium</td><td><div class='wide'>guard; prod perf: outer/inner shape specializations, benchmark, perf table. Review var/std after the reduce base/prod direction is settled: #187787. Current state: rebuilt and pushed at cd01523561027; ready as the Welford follow-up, but should be read after #188156/#188333 bec...</div></td><td>952d13051c9d44a446a2be73e86248a8d47422d25eeff451dfe01383173f6430</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #187976</td><td>medium</td><td><div class='wide'>anup: complex enablement, integer rejection, duplicate-dim behavior, and large-output errors if those remain separate review concerns. Keep #187976 independent. #187976 fixes an int64 amax/amin partial-simdgroup bug in the existing value-reduction helper path. It is a sibling...</div></td><td>a6a44834bd4df619c0ecf08fbcf7fc9d812c5e0bf2cd91c7401bf348805177df</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #188156</td><td>medium</td><td><div class='wide'>A standalone test_mps.py case needs a clear reason why no existing surface owns the route. Order of operations Land the reduce-family base: #188156. Scope: collapse generic sum/mean/nansum/count_nonzero onto the shared value_reduction&lt;Op, Load, TA&gt; path already used by min/max...</div></td><td>6d4f2eed50ae483231d5f1291cd1d08b298ba1ff098a4eed5c37fae987ed7a54</td></tr><tr><td>supersedes</td><td>issue #187455</td><td>pr #188333</td><td>medium</td><td><div class='wide'>route for non-contiguous multi-dim sum-family reductions and load policies, not just a private helper. Land prod on top of the reduce base: #188333. Scope: migrate prod to value_reduction&lt;ProdOp&gt; and remove its shape-dependent MPSGraph route. This supersedes the closed standal...</div></td><td>64907f0296db66e3f73e9c6406bcd1aa15b12816066489d2f98f9630b55a7190</td></tr><tr><td>references</td><td>issue #187455</td><td>pr #189146</td><td>medium</td><td><div class='wide'>ned MPSGraph fallback, so it is correct standalone and strictly reduces the softmax graph-cache for the common last-dim case. softmax perf (#189146, 8fa0937): adds the huge-row two-pass and non-last-dim (coalesced/tiled/blocked) native kernels on top of core and removes the la...</div></td><td>ff04e385d39996a9012222e032a6741e7118b0823c5d3a9b59ce73aeaa1c8b26</td></tr><tr><td>references</td><td>issue #187175</td><td>issue #186241</td><td>medium</td><td><div class='wide'>to it unconditionally. There are several recently filed issues that highlight the TMA path for pointwise/reduction is not yet fully mature: #186241 #185513 #185222 #184714 #184563 Phase 2 — Once we are confident the TMA path is correct and mature: Enabling autotuning alone (wi...</div></td><td>447ca33d4dbd72febb21f81154fa1ee7db554dfabb534b423eff70ad1d01e8d1</td></tr><tr><td>references</td><td>issue #187810</td><td>issue #187811</td><td>medium</td><td><div class='wide'>Hi @chunhuanMeng , please help handle this issue and #187811</div></td><td>84c3b4bdab0afe48a014091b2f1016135305c1871a6dcf86c9eb8939aff8f862</td></tr><tr><td>references</td><td>issue #189138</td><td>issue #189135</td><td>medium</td><td><div class='wide'>is replaced by _exposes_streams(device), which checks whether iface.Stream is not DeviceInterface.Stream. Please refer to the dedicated RFC #189135 for detailed design elaboration. Dynamo Registry Splicing Dynamo answers two questions for every value/call during symbolic execu...</div></td><td>e8b0f2becfc15424643138f062faee984ee45d6b1e5b8479bfcd04876886be85</td></tr><tr><td>references</td><td>issue #189138</td><td>issue #189136</td><td>medium</td><td><div class='wide'>e-specialized tensor types — is handled by one optional slot dynamo_tensor_classes (default empty, safe). Please refer to the dedicated RFC #189136 for detailed design elaboration. Inductor Codegen Registration Interfaces Inductor&#x27;s codegen path has four things that lack regis...</div></td><td>175dc64f062f304b809a5c88277b99d8dcf7740ecb1dafd45a75009868f8126c</td></tr><tr><td>references</td><td>issue #189138</td><td>issue #189137</td><td>medium</td><td><div class='wide'>E_TO_ATEN fallback. No-c-shim backends will still fail AOTI at link time. This needs a separate proposal. Please refer to the dedicated RFC #189137 for detailed design elaboration. For a more comprehensive and detailed understanding of each component&#x27;s design, we highly encour...</div></td><td>75a60e2adbd3fd72c533e18e7d53249ad6bf4b74cb2ed1705175cb815516f709</td></tr><tr><td>references</td><td>issue #144965</td><td>issue #69078</td><td>medium</td><td><div class='wide'>der some specific conditions, raise a RuntimeError with message: &quot;Global alloc not supported yet.&quot;. I think this is linked to an old issue: #69078, however, I managed to consistently reproduce this error. The code that reproduces the bug is quite long and needs some explanatio...</div></td><td>af0a8b92d21995f8ec4b0b5a5d44613aaaaab920e32715959df72818a2c6066d</td></tr><tr><td>references</td><td>issue #188970</td><td>issue #180397</td><td>medium</td><td><div class='wide'>not collect Expected behavior Compile succeeds, or the unsupported/unsafe option is ignored with a warning — not a native SIGSEGV. Related #180397 proposes a capture/replay API for MPS and notes MPS currently has no CUDA-Graphs equivalent — possibly useful context, since trito...</div></td><td>8dc4c11946ae99660774b1637ca1d88153f73a9a68462b4a70a8207182287fd5</td></tr><tr><td>references</td><td>issue #187093</td><td>pr #189127</td><td>medium</td><td><div class='wide'>Opened a draft PR with the scoped post-grad rewrite: #189127 The patch extends the existing addmm bias-unfuse pattern to baddbmm: it only applies on GPU when the input is bias-like/broadcast and all b</div></td><td>56c41fc8de4001e0f38282047b6bfb8a47f1e098c860eb723fb2cd2a8c320d4c</td></tr><tr><td>references</td><td>issue #189121</td><td>pr #189124</td><td>medium</td><td><div class='wide'>Opened a draft PR with the focused cache-gate patch: #189124 The patch keeps the ordinary single-config fast path unchanged, but lets DSR-eligible single-config reductions create/load an autotune cach</div></td><td>d2b79c1ccd9aea3c11d6213c11b639a5cc50550df8650ad729bda49e2271fd54</td></tr><tr><td>references</td><td>issue #188602</td><td>pr #186245</td><td>medium</td><td><div class='wide'>used to optimize the binary. The current plan is that this needs the following high level steps: Integrate LLVM-BOLT into the build system #186245 Add a profile staleness metric script which guides profile updates Enable profile collection in PyTorch CI cc @malfet @ptrblck @ms...</div></td><td>eafa00736571e47a2117cc23a189b6b089ae0d1f7ef2932593feb7b174ca8374</td></tr><tr><td>references</td><td>issue #189078</td><td>pr #186245</td><td>medium</td><td><div class='wide'>add this so we can keep track of how various changes impact build times. The main motivation for this comes from the desire to ensure that #186245 does not introduce major build time overhead but metrics should be helpful in general. Alternatives No response Additional context...</div></td><td>15a0ad69b44471a4d861bf8220eee35293248bd9f1f3227256c083a28432d825</td></tr><tr><td>references</td><td>issue #189098</td><td>pr #189097</td><td>medium</td><td><div class='wide'>o Kineto could utilize the export_using_protocol method in order to access them in Pytorch. A draft PR of the changes has been opened here: #189097 Alternatives No response Additional context No response cc @robieta @chaekit @guotuofeng @guyang3532 @dzhulgakov @davidberard98 @...</div></td><td>02871cf4cce200404cad179b35736868e013efe2530612dc49c5e2b89b1af758</td></tr><tr><td>references</td><td>issue #189077</td><td>issue #188557</td><td>medium</td><td><div class='wide'>d-environment linux-jammy-cuda13.0-py3.12-gcc11-sm100, single unsharded (1, 1) job on one Blackwell GPU. This is the NVIDIA/b200 sibling of #188557 (which tracks the same failure mode on ROCm gfx950). Same root cause and fix levers; different infra/owners, so filing separately...</div></td><td>a245fc0a78e2348b1ce75282062a5e55c27c5db63737321c27f17e4ea8a3485b</td></tr><tr><td>references</td><td>issue #189075</td><td>issue #140556</td><td>medium</td><td><div class='wide'>never completes. Genuine per-test failures in the same job (e.g. the test_backward_prod_cuda_float32 mem-leak-check, tracked separately in #140556) are a distinct problem from this timeout. Fix levers (CI-side) Increase the shard count (currently 8) so each shard&#x27;s slice fits...</div></td><td>682fa9067cfe772f7378d44a92d3c58c9d10c338662ca6b4b9b83579609f2bde</td></tr><tr><td>references</td><td>issue #189065</td><td>issue #154297</td><td>medium</td><td><div class='wide'>0 runner pool: investigate GPU state on linux.dgx.b200.8 (leaked processes, compute mode, or reimage). Nothing to fix in test code. Related #154297 - Hangs and timeouts on dist.reduce_scatter on B200 GPU (user-reported; possibly shared underlying B200 cause). #162178 - [CUDA]...</div></td><td>e7c80387a19e22fcfe21acebc2f3514911d5bab4d80513a29d95491c258a1062</td></tr><tr><td>references</td><td>issue #189065</td><td>issue #162178</td><td>medium</td><td><div class='wide'>test code. Related #154297 - Hangs and timeouts on dist.reduce_scatter on B200 GPU (user-reported; possibly shared underlying B200 cause). #162178 - [CUDA] Umbrella Issues/Failures on B200 Runner (stale umbrella). cc @ezyang @gchanan @kadeng @msaroufim @awgu @wanchaol @fegin @...</div></td><td>b36586005896bfd37626f9eb243032b2e5752e9be2914e7be637b68a084a6805</td></tr><tr><td>references</td><td>issue #188733</td><td>issue #45111</td><td>medium</td><td><div class='wide'>al MLA / attention changes that alter KV-cache layout/strides and Sm90 backend selection for the MLA path -- most notably: vllm-project/vllm#45111 - &quot;Re-enable cross-layer KV cache layout for MLA via stride-aware kernels&quot; (changes MLA KV-cache strides/layout) -- primary suspec...</div></td><td>1d558cb9461e859db727514da4a760bfc7ca0df7fb0de4238ebc15a6b331d43a</td></tr><tr><td>references</td><td>issue #189034</td><td>issue #173059</td><td>medium</td><td><div class='wide'>iPy to a NumPy-2-compatible build (SciPy &gt;=1.13 supports NumPy 2). Whoever owns the H100 image / bumped numpy 2.2.6 should confirm. Related #173059 -- same class (NumPy binary/ABI incompatibility breaking test imports), different instance (pandas + test_dataloader/test_datapip...</div></td><td>4ff8aa912ca76a90e33f07b086d70223c3964f06cf9565c1b1a97a7b8e85df01</td></tr><tr><td>references</td><td>issue #188900</td><td>issue #158861</td><td>medium</td><td><div class='wide'>s-is, so only the slice at index 0 along the broadcast dimension gets values and the rest stay zero. This looks like the same root cause as #158861, where `@nikitaved` noted that explicit broadcasting via `sparse_broadcast_to` followed by coalesce is missing. Proposed fix: bef...</div></td><td>ea7c404fac461685b6138c3782e0a7aa2a90ce8b26aa1c804b94bd8e9cf24563</td></tr><tr><td>references</td><td>issue #188792</td><td>pr #188791</td><td>medium</td><td><div class='wide'>each worker do a serial 1-element torch.sqrt first; this drops the rate to 0 (5,200 first-calls), which is the basis of the fix proposed in #188791. mkl_vml_race_repro.py &quot;&quot;&quot;Minimal repro: MKL VML first-call dispatch race in torch.sqrt. Spawns waves of worker processes on an o...</div></td><td>7a83aacf57b04dde3d1b2448cf0f4270ff7c363003b1b84fc081ebdb5c886f78</td></tr><tr><td>references</td><td>issue #188564</td><td>issue #152595</td><td>medium</td><td><div class='wide'>Here the build is fine, the GPU is found, tests run and pass, and the job is killed purely by the 270-min wall-clock. (Other gfx1100 issues #152595 / #165141 / #169857 / #177834 are unrelated.) Suggested fix Unblock now (CI-side): increase shard count for this job (it&#x27;s num_sh...</div></td><td>7f98c780b875fbc5917ea441da987ef5c78b69b8ed2330a6f9d3b6d21d088cde</td></tr><tr><td>references</td><td>issue #188564</td><td>issue #165040</td><td>medium</td><td><div class='wide'>s confirmed for the current runs but not individually proven as the culprit for all 388 red runs. This is NOT covered by the existing issue #165040 &quot;DISABLED rocm / linux-jammy-rocm-py3_10-gfx1100 / test (default)&quot; (OPEN; module: rocm, module: ci) is adjacent but distinct: it&#x27;...</div></td><td>1727021f32469529a36c72b5fc4f013ba7de4ed2f4b810df8c96ebecb1fc8be8</td></tr><tr><td>references</td><td>issue #188564</td><td>issue #165141</td><td>medium</td><td><div class='wide'>uild is fine, the GPU is found, tests run and pass, and the job is killed purely by the 270-min wall-clock. (Other gfx1100 issues #152595 / #165141 / #169857 / #177834 are unrelated.) Suggested fix Unblock now (CI-side): increase shard count for this job (it&#x27;s num_shards=2 tod...</div></td><td>60d9d4eb9c8753752450ee1167bd539202a190c38396b057e8ef0c3262a396e7</td></tr><tr><td>references</td><td>issue #188564</td><td>issue #169857</td><td>medium</td><td><div class='wide'>ne, the GPU is found, tests run and pass, and the job is killed purely by the 270-min wall-clock. (Other gfx1100 issues #152595 / #165141 / #169857 / #177834 are unrelated.) Suggested fix Unblock now (CI-side): increase shard count for this job (it&#x27;s num_shards=2 today) so eac...</div></td><td>1f4d60a6b9f14e9096dc8a717bf935fce550e4de05919055bdbbf7467eb50770</td></tr><tr><td>references</td><td>issue #188564</td><td>issue #177834</td><td>medium</td><td><div class='wide'>U is found, tests run and pass, and the job is killed purely by the 270-min wall-clock. (Other gfx1100 issues #152595 / #165141 / #169857 / #177834 are unrelated.) Suggested fix Unblock now (CI-side): increase shard count for this job (it&#x27;s num_shards=2 today) so each Test ste...</div></td><td>75777d9e62901abcd78bcf4fcf56ad11f56310d0079b192934bc543a14640ce5</td></tr><tr><td>references</td><td>issue #188545</td><td>issue #187332</td><td>medium</td><td><div class='wide'>e returns [nan, nan, nan], but Inductor returns [nan, 0., nan]. This changes the isnan result for the +inf element. This appears related to #187332 , but it affects torch.special.bessel_y0. Repro import torch def test_bessel_y0_mismatch(): x = torch.tensor([float(&#x27;nan&#x27;), float...</div></td><td>6ccedbc5c7e5104c094940d20c35e1caf868f0491c04b80f0ac66cfa51bc06db</td></tr><tr><td>references</td><td>issue #188008</td><td>pr #187955</td><td>medium</td><td><div class='wide'>ing unwinding → std::terminate. This is a pre-existing latent bug (the per-segment unmap loop had the same defect); it is not introduced by #187955, but that PR&#x27;s unmapHandles rework touches the same code, so filing as follow-up. Mechanism map() / fromShared() set handles_[beg...</div></td><td>a0b4f7b49dff0e32705ba8c38ef4f342544be554b1648ec1a1fed54210ffcf10</td></tr><tr><td>references</td><td>issue #188008</td><td>pr #187955</td><td>medium</td><td><div class='wide'>Being fixed in #187955: mapAndSetAccess is now strongly exception-safe — on a partial map/setAccess failure it unmaps the mapped prefix and releases the range&#x27;s h</div></td><td>aaa82942a131c603561fa9378f65c45b3a3d5cc7cd2000202884d9ffee5f006f</td></tr><tr><td>references</td><td>issue #140718</td><td>issue #77764</td><td>medium</td><td><div class='wide'>y implemented for the MPS device. If you want this op to be added in priority during the prototype phase of this feature, please comment on #77764. As a temporary fix, you can set the environment variable PYTORCH_ENABLE_MPS_FALLBACK=1 to use the CPU as a fallback for this op....</div></td><td>913bb5cf94a50de8ea0c409f50d5e078de09feb8a75dfcd334449ef7e68ce925</td></tr><tr><td>references</td><td>issue #188953</td><td>issue #179891</td><td>medium</td><td><div class='wide'>xpu: GPU detected, data transfer works, but any non-trivial computation (matmul) fails with could not create an engine. Related to existing #179891 (get_device_properties segfault on BMG). Intel compute-runtime issue cross-reference: intel/compute-runtime#947 Environment Compo...</div></td><td>edf3792aa9ff53d09b457d938437da2f12830a88948c595bbf819c7e2ba2b893</td></tr><tr><td>references</td><td>issue #188718</td><td>pr #188719</td><td>medium</td><td><div class='wide'>Thanks @Sahil170595. The PR is #188719</div></td><td>d8de2cb15f5aab0c45ed4e1a918095e627d417e6a759e0bb722011f90c8d4967</td></tr><tr><td>references</td><td>issue #175891</td><td>issue #62530</td><td>medium</td><td><div class='wide'>itional nodes, switch can do the same in CUDA &gt;= 12.8. Alternatives No response Additional context This has previously been requested here: #62530 cc @mcarilli @ezyang @eellison @penguinwu @BoyuanFeng @chauhang @ydwu4 @bdhirsh @bobrenjc93 @aorenste</div></td><td>9a00af6f1c18741467c70dc5876e37327ef8e01a8789fa2f2fe567b1877b247f</td></tr><tr><td>references</td><td>issue #182277</td><td>issue #177827</td><td>medium</td><td><div class='wide'>aise a clean Python exception such as OverflowError, ValueError, or a normal CUDA OOM RuntimeError, not SystemError. This may be related to #177827, but it does not seem to be the same issue. #177827 is about negative allocation sizes, while this issue is triggered by a positi...</div></td><td>0c280b4f4ac95b5cd7b9eb9d77e20e1830114ba4cb6350cfd1897c45adbc30d9</td></tr><tr><td>references</td><td>issue #188812</td><td>pr #188801</td><td>medium</td><td><div class='wide'>MetalShaderLibrary&amp; MetalShaderLibrary::getBundledLibrary() { static BundledShaderLibrary* l = new BundledShaderLibrary(); return *l; } PR: #188801 (will be scoped down to just this leak). cc @malfet @aditvenk @kulinseth @DenisVieriu97 @jhavukainen @Isalia20 Versions PyTorch v...</div></td><td>c2134ecc4dba4b1da73c0d76b5a01693ba1c84a1c535b16b58927411e651d4df</td></tr><tr><td>references</td><td>issue #188319</td><td>pr #188951</td><td>medium</td><td><div class='wide'>Implemented the core pipeline changes in #188951, covers matrix generation, Makefile, Dockerfile, and workflow env vars. Would appreciate any feedback on the approach.</div></td><td>2efa83a6f79f702c8ae605facf76fdd636076faba758ac5e93db3089c80368d1</td></tr><tr><td>references</td><td>issue #177116</td><td>issue #117826</td><td>medium</td><td><div class='wide'>shaders, fixed for torch.mm/torch.bmm via tiling in PR #117549 but not all code paths), #122045 (F.linear wrong results for large inputs), #117826 (incorrect einsum gradient on MPS). Minimal reproduction The model is Embedding → 2 residual Linear blocks → output Linear, traine...</div></td><td>8ed85e7ca715f1fab7ccca1a190b0fd1593e47912a1109682dbc672e09b66fad</td></tr><tr><td>references</td><td>issue #184575</td><td>pr #188066</td><td>medium</td><td><div class='wide'>~1e12 in backward (not NaN). This is a silent gradient corruption — users see finite gradients when they should see NaN. Submitted test PR: #188066 Detected via NeuralDBG — causal diagnostic engine for PyTorch training.</div></td><td>9d5bc7624ee970ffb047679d26c36ed9cf9a8dab9b283dc84652743252a3ca4a</td></tr><tr><td>references</td><td>issue #188891</td><td>issue #187988</td><td>medium</td><td><div class='wide'>1-D weight correctly, which is why eager rank &gt;= 3 works. Context Found while adding 1-D-weight OpInfo samples to sample_inputs_linear for #187988 / PR #187989 (an unrelated MPS backward crash with 1-D weights): a sample with a scalar bias turns 14 functorch OpInfo tests red o...</div></td><td>5c0571484e233f01f4babb49eac24cc12f29f0aa634da0332f085e0c7df35515</td></tr><tr><td>references</td><td>issue #188891</td><td>pr #187989</td><td>medium</td><td><div class='wide'>orrectly, which is why eager rank &gt;= 3 works. Context Found while adding 1-D-weight OpInfo samples to sample_inputs_linear for #187988 / PR #187989 (an unrelated MPS backward crash with 1-D weights): a sample with a scalar bias turns 14 functorch OpInfo tests red on all backen...</div></td><td>c614743a0c8262a4303764eb0e730a4fdd43b87def99992865382c70940be5f6</td></tr><tr><td>references</td><td>issue #188670</td><td>pr #169694</td><td>medium</td><td><div class='wide'>nce / ROCm follow-up #182028 ROCm MI200 — flex vs SDPA assert_close tolerances #186889 XPU — relaxed assertEqual vs eager (atomic ordering) #169694 Transformer tests — golden precision / reference choice #183691 Test infra — tolerance overrides (helps symptoms; root cause is s...</div></td><td>198e53966f626e02bbdf17b398887cd138b4318c2c9a462aa2ae6ccc1ae15874</td></tr><tr><td>references</td><td>issue #188670</td><td>pr #183691</td><td>medium</td><td><div class='wide'>ose tolerances #186889 XPU — relaxed assertEqual vs eager (atomic ordering) #169694 Transformer tests — golden precision / reference choice #183691 Test infra — tolerance overrides (helps symptoms; root cause is still design in individual tests) 4. Conclusion &amp; proposed next s...</div></td><td>e77c5c394c4db350dad1287de44cbd34ca3016b7a29375880c9e86389e7b7a03</td></tr><tr><td>references</td><td>issue #188670</td><td>pr #184531</td><td>medium</td><td><div class='wide'>oesn&#x27;t solve the root of the issue. Tight same() / eager-fp16 equality is stricter than the bug needs and varies by hardware (e.g. #179958, #184531) — false failures even when the grid fix is sound. 3. Other examples (same pattern family) Link Note #183630 test_bmm_large_batch...</div></td><td>bd51bf5985bf862772fa908fe0af098cbe2480688a50b0a83f66cafb860b7ba2</td></tr><tr><td>references</td><td>issue #188670</td><td>pr #187050</td><td>medium</td><td><div class='wide'>Here is a PR that implements tests for bmm issue in a way that I think is better than current implementation: #187050 It shouldn&#x27;t be used to close the issue, just serve as an example of the direction that I propose to move towards to.</div></td><td>a6427297580fb8d1214da154523bc28d92d379f1130f2178df72aaef3e3c91e8</td></tr><tr><td>references</td><td>issue #188727</td><td>pr #188500</td><td>medium</td><td><div class='wide'>ike GraphPP and Eager Torchtitan PP. We are changing PP to no longer handle microbatch splitting and let the user supply microbatches here: #188500 Torchtitan will now produce microbatches right at source i.e. dataloader: pytorch/torchtitan#3856 So no changes are required from...</div></td><td>fa1d6ca45bc0918ba4f584c1be38f60155ad873dd4d1789d3473e40dc5f8e10a</td></tr><tr><td>references</td><td>issue #188772</td><td>pr #188698</td><td>medium</td><td><div class='wide'>tem.py (which is brittle) and implement their own workarounds for asynchronous saving over fsspec. Additional context This is drafted in PR #188698 and discussed offline with @LucasLLC and @meetv18. Opening this issue to take it further with the process cc @awgu @wanchaol @feg...</div></td><td>f672141928f220635f80b6689a8a4bf16d8e328f2cb1f56d4c4965a7c26729e1</td></tr><tr><td>references</td><td>issue #188711</td><td>issue #188675</td><td>medium</td><td><div class='wide'>fast_accum_True_scaling_block_sizes2_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @ptrblck @msaroufim @eqy @tinglvv @nWEIdia</div></td><td>380e8b3b0e3d23c78bc388f0c39b430e7091ba46c1b318dee418bc47871250aa</td></tr><tr><td>references</td><td>issue #188711</td><td>issue #188704</td><td>medium</td><td><div class='wide'>m_True_scaling_block_sizes2_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @ptrblck @msaroufim @eqy @tinglvv @nWEIdia</div></td><td>5d870351811498879b4882276cdb3b8ad227981535b4d9babb5dbe37ee5349d0</td></tr><tr><td>references</td><td>issue #188707</td><td>issue #188675</td><td>medium</td><td><div class='wide'>ast_accum_False_scaling_block_sizes1_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @ptrblck @msaroufim @eqy @tinglvv @nWEIdia</div></td><td>086d0df5d173b95613cf8cc09a713ec31938bbe00b72919802796fb04158cce8</td></tr><tr><td>references</td><td>issue #188707</td><td>issue #188704</td><td>medium</td><td><div class='wide'>_False_scaling_block_sizes1_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @ptrblck @msaroufim @eqy @tinglvv @nWEIdia</div></td><td>8aad4c9a3346ff569d46f2e73ceaea770185441378d8365818c617c77dc759ca</td></tr><tr><td>references</td><td>issue #188706</td><td>issue #188675</td><td>medium</td><td><div class='wide'>ast_accum_False_scaling_block_sizes0_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @ptrblck @msaroufim @eqy @tinglvv @nWEIdia</div></td><td>800422d2dbb726ce7dc6d5d2b41483d112e6bfa3de8f3cbb186fd73a351fbc2f</td></tr><tr><td>references</td><td>issue #188706</td><td>issue #188704</td><td>medium</td><td><div class='wide'>_False_scaling_block_sizes0_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @ptrblck @msaroufim @eqy @tinglvv @nWEIdia</div></td><td>e945e5c0dec8f603f8a1226536e8ebb9ceec1f90de8152f9072435d0f78e09f7</td></tr><tr><td>references</td><td>issue #102207</td><td>issue #108798</td><td>medium</td><td><div class='wide'>_.from needs to be renamed&quot; &#x27;test_advancedindex_mixed_cpu_devices&#x27; - &quot;FIXME&quot; &#x27;test_advancedindex_mixed_devices_error&#x27; - &quot;FIXME&quot; #108181 and #108798 &#x27;test_typed_storage_deprecation_warning&#x27; - &quot;FIXME&quot; &#x27;test_pin_memory&#x27; - &quot;pin_memory isn&#x27;t yet supported in TorchInductor&quot; &#x27;test_me...</div></td><td>9eac0f3787e9f7be4bae833fe135aa372e815033066e68e5669d452d2ff375ce</td></tr><tr><td>references</td><td>issue #188721</td><td>issue #188477</td><td>medium</td><td><div class='wide'>: module &#x27;cutlass.cute.arch&#x27; has no attribute &#x27;ProxyKind&#x27; (unpinned cutlass_api vs pinned nvidia-cutlass-dsl==4.5.2). Root cause tracked in #188477. Unstable until the cutlass_api / cutlass-dsl version skew is resolved. cc @malfet @pytorch/pytorch-dev-infra</div></td><td>10df704ed063fc92983779ef79ab6ae6c3f17cb633156e78fca63c99cc6dbec4</td></tr><tr><td>references</td><td>issue #188714</td><td>issue #188675</td><td>medium</td><td><div class='wide'>fast_accum_True_scaling_block_sizes1_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @ptrblck @msaroufim @eqy @tinglvv @nWEIdia @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-...</div></td><td>c0ffd4ffb8e61edc1a7c08df3df4169354695b25febc034bde3c4b3e5565bf21</td></tr><tr><td>references</td><td>issue #188714</td><td>issue #188704</td><td>medium</td><td><div class='wide'>m_True_scaling_block_sizes1_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @ptrblck @msaroufim @eqy @tinglvv @nWEIdia @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @Xia...</div></td><td>f6642f1800f107c17de69c746216551c020e432c2aafb41e60b78c092f123237</td></tr><tr><td>references</td><td>issue #188715</td><td>issue #188675</td><td>medium</td><td><div class='wide'>fast_accum_True_scaling_block_sizes2_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @we...</div></td><td>5a37ccad30a9eecfe83cb39ab9baab004caf4c3a94d0e77bb4de32f34f0a85b7</td></tr><tr><td>references</td><td>issue #188715</td><td>issue #188704</td><td>medium</td><td><div class='wide'>m_True_scaling_block_sizes2_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv...</div></td><td>eabbe0bea09182cafeb8da6c348ee12821462fde0eea5a628c7d38bee1b51061</td></tr><tr><td>references</td><td>issue #188712</td><td>issue #188675</td><td>medium</td><td><div class='wide'>ast_accum_False_scaling_block_sizes1_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @ptrblck @msaroufim @eqy @tinglvv @nWEIdia @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-...</div></td><td>f290676f1ab2d12c4f511352083abff8c8d2dfde47772cd0e89040927553fa49</td></tr><tr><td>references</td><td>issue #188712</td><td>issue #188704</td><td>medium</td><td><div class='wide'>_False_scaling_block_sizes1_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @ptrblck @msaroufim @eqy @tinglvv @nWEIdia @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @Xia...</div></td><td>4c5f171e6f5735599314786e8f0bbcedce80c4d199357c5c956644a673e21452</td></tr><tr><td>references</td><td>issue #188709</td><td>issue #188675</td><td>medium</td><td><div class='wide'>fast_accum_True_scaling_block_sizes0_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @we...</div></td><td>970a255e54311b85311aeac6f6fb42b2e9aea2c6e665b9f70ddeefe26ce11610</td></tr><tr><td>references</td><td>issue #188709</td><td>issue #188704</td><td>medium</td><td><div class='wide'>m_True_scaling_block_sizes0_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv...</div></td><td>7f22e90fe7a4f84c421309be4f27ee8ed5f0f2fdaddfe5504b01287f7d528fd6</td></tr><tr><td>references</td><td>issue #188713</td><td>issue #188675</td><td>medium</td><td><div class='wide'>ast_accum_False_scaling_block_sizes2_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @ptrblck @msaroufim @eqy @tinglvv @nWEIdia @mruberry @jianyuh @nikitaved @walterddr @lezcano @chauhang @penguinwu @v...</div></td><td>506ac453723407ac03c04d17c5f9ed7f2effbe2e149b13e3d8cd1224c3b808c5</td></tr><tr><td>references</td><td>issue #188713</td><td>issue #188704</td><td>medium</td><td><div class='wide'>_False_scaling_block_sizes2_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @ptrblck @msaroufim @eqy @tinglvv @nWEIdia @mruberry @jianyuh @nikitaved @walterddr @lezcano @chauhang @penguinwu @voznesensk...</div></td><td>faa883fd8e576c81e29b73be484507092dc97967ef43815ea59cb7f1df56deb7</td></tr><tr><td>references</td><td>issue #188717</td><td>issue #188675</td><td>medium</td><td><div class='wide'>ast_accum_False_scaling_block_sizes2_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @we...</div></td><td>9d6aa963f9e1b3659683551432bb0e30abe511d2cb5a6e2f058abbd3e5d11022</td></tr><tr><td>references</td><td>issue #188717</td><td>issue #188704</td><td>medium</td><td><div class='wide'>_False_scaling_block_sizes2_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv...</div></td><td>64eee4d83aca66f94e0f6b4b11c4304546de08d3f2a7916b64e4a00d94bb2bc6</td></tr><tr><td>references</td><td>issue #188708</td><td>issue #188675</td><td>medium</td><td><div class='wide'>ast_accum_False_scaling_block_sizes2_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @we...</div></td><td>1e7aa4b708515aed1abe502d66f11ba976d81326e4cff82fddbbe69d393fb8f0</td></tr><tr><td>references</td><td>issue #188708</td><td>issue #188704</td><td>medium</td><td><div class='wide'>_False_scaling_block_sizes2_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv...</div></td><td>a63ac04f3aa92924d143b5f3c76de537074f8941a76015bd963bb28801d1a9b8</td></tr><tr><td>references</td><td>issue #188716</td><td>issue #188675</td><td>medium</td><td><div class='wide'>ast_accum_False_scaling_block_sizes1_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @we...</div></td><td>0ea007c8d36e2ffeef844c1311a78327436c618f2324070d1743f416d8c89c4e</td></tr><tr><td>references</td><td>issue #188716</td><td>issue #188704</td><td>medium</td><td><div class='wide'>_False_scaling_block_sizes1_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv...</div></td><td>3533f21e3e7b72b453b8187347400979a070d26555cfd388a18ddb2c8e08d22a</td></tr><tr><td>references</td><td>issue #188710</td><td>issue #188675</td><td>medium</td><td><div class='wide'>fast_accum_True_scaling_block_sizes1_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @we...</div></td><td>4ae6d492ad2b3c3935086523dc68b6b689c360a25dac91b1ca3984b599f5c4d2</td></tr><tr><td>references</td><td>issue #188710</td><td>issue #188704</td><td>medium</td><td><div class='wide'>m_True_scaling_block_sizes1_cuda One of ~14 failing test_main_loop_scaling parametrizations (w = BlockWise1x128 recipes). See also #188675, #188704. cc @mruberry @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv...</div></td><td>fdc73e68154a961c66e385be08098f7660819a75001ad588ab6a8e9a4480183e</td></tr><tr><td>references</td><td>issue #188652</td><td>pr #188263</td><td>medium</td><td><div class='wide'>he system, it downloads its own python-six sources files from the web (pypi to be exact) This makes pytorch not buildable from sources only #188263 shows the changes needed for allowing air-gapped builds To reproduce the bug simply try to build pytorch on a machine without int...</div></td><td>c64925d08a1e2b495fad1e9153b7c9abe99e2fda39eaab3f5468c1ed5467b161</td></tr><tr><td>competes with</td><td>issue #188425</td><td>issue #188420</td><td>medium</td><td><div class='wide'>@Blooming-Tree I noticed that #188420, #188430, and #188432 appear to have the same methodology issue. They compare eager FP64 against compiled FP32 rather than eager FP32 again</div></td><td>6030c43457e684c489aa1189093263292c95b210873f3bdec9848039f82a9e41</td></tr><tr><td>competes with</td><td>issue #188425</td><td>issue #188430</td><td>medium</td><td><div class='wide'>@Blooming-Tree I noticed that #188420, #188430, and #188432 appear to have the same methodology issue. They compare eager FP64 against compiled FP32 rather than eager FP32 against compil</div></td><td>dfc3ea4565e9d75e9bd4e2d1a354936446792c6c243acbf699c6d0b7b85e6376</td></tr><tr><td>competes with</td><td>issue #188425</td><td>issue #188432</td><td>medium</td><td><div class='wide'>@Blooming-Tree I noticed that #188420, #188430, and #188432 appear to have the same methodology issue. They compare eager FP64 against compiled FP32 rather than eager FP32 against compiled FP32, so t</div></td><td>25952773a33260bddb92f3d86fd71d0c8614fcce65ca99bf69a67ad8950ae3e5</td></tr><tr><td>references</td><td>issue #132845</td><td>issue #15963</td><td>medium</td><td><div class='wide'>ue, please check the box on this page and assign the issue to yourself. Thanks! C10D NCCL #131203 #112815 #94003 #58856 GLOO #115740 #16295 #15963 Tcpstore #61671 #50701 MultiProcessing #65559 MultiThreadedTestCase #107425 #107424 #100459 #100448 RPC #125365 #74180 #70753 #705...</div></td><td>6c9111a02c5a0197606cd4c4228cdde6e8327e9346c654d4895eb3d1e5c8c6f5</td></tr><tr><td>references</td><td>issue #132845</td><td>issue #16295</td><td>medium</td><td><div class='wide'>an issue, please check the box on this page and assign the issue to yourself. Thanks! C10D NCCL #131203 #112815 #94003 #58856 GLOO #115740 #16295 #15963 Tcpstore #61671 #50701 MultiProcessing #65559 MultiThreadedTestCase #107425 #107424 #100459 #100448 RPC #125365 #74180 #7075...</div></td><td>3f5de47a55f7828e5b27e907165fc3be2fdd9844956ed32d2d4924eb21942134</td></tr><tr><td>references</td><td>issue #132845</td><td>issue #51062</td><td>medium</td><td><div class='wide'>1 #50701 MultiProcessing #65559 MultiThreadedTestCase #107425 #107424 #100459 #100448 RPC #125365 #74180 #70753 #70546 #69179 #69178 #64093 #51062 DeviceMesh and DTensor DeviceMesh #132833 #123906 #122525 #122664 #121951 #121776 #121372 #120706 #120692 #110026 #98875 #98059 #9...</div></td><td>ee4a99abde9564c99b6d193c26838b234f9ec91787e275ee2a2591f21c900c0e</td></tr><tr><td>references</td><td>issue #132845</td><td>issue #58856</td><td>medium</td><td><div class='wide'>u would like to take an issue, please check the box on this page and assign the issue to yourself. Thanks! C10D NCCL #131203 #112815 #94003 #58856 GLOO #115740 #16295 #15963 Tcpstore #61671 #50701 MultiProcessing #65559 MultiThreadedTestCase #107425 #107424 #100459 #100448 RPC...</div></td><td>2fcc1403a4a723e7d1a8c76cddfc97094f497485ee4cc71d2ba7a0b7ff58fc5c</td></tr><tr><td>references</td><td>issue #132845</td><td>issue #64093</td><td>medium</td><td><div class='wide'>e #61671 #50701 MultiProcessing #65559 MultiThreadedTestCase #107425 #107424 #100459 #100448 RPC #125365 #74180 #70753 #70546 #69179 #69178 #64093 #51062 DeviceMesh and DTensor DeviceMesh #132833 #123906 #122525 #122664 #121951 #121776 #121372 #120706 #120692 #110026 #98875 #9...</div></td><td>6de948c3fa1a17c69f726908fad72c292a277d6af77e747ec94e4c1149179992</td></tr><tr><td>references</td><td>issue #132845</td><td>issue #69178</td><td>medium</td><td><div class='wide'>Tcpstore #61671 #50701 MultiProcessing #65559 MultiThreadedTestCase #107425 #107424 #100459 #100448 RPC #125365 #74180 #70753 #70546 #69179 #69178 #64093 #51062 DeviceMesh and DTensor DeviceMesh #132833 #123906 #122525 #122664 #121951 #121776 #121372 #120706 #120692 #110026 #9...</div></td><td>ce44fb6f9ea09942a5f6cff4f68ee347d11719250e1d08e63c138b3c854a64a6</td></tr><tr><td>references</td><td>issue #132845</td><td>issue #69179</td><td>medium</td><td><div class='wide'>#15963 Tcpstore #61671 #50701 MultiProcessing #65559 MultiThreadedTestCase #107425 #107424 #100459 #100448 RPC #125365 #74180 #70753 #70546 #69179 #69178 #64093 #51062 DeviceMesh and DTensor DeviceMesh #132833 #123906 #122525 #122664 #121951 #121776 #121372 #120706 #120692 #11...</div></td><td>fcc954b1e21d7dfdc0b1f09f4a8123017402c67814372fb459ebc2b642715795</td></tr><tr><td>references</td><td>issue #132845</td><td>issue #94003</td><td>medium</td><td><div class='wide'>. If you would like to take an issue, please check the box on this page and assign the issue to yourself. Thanks! C10D NCCL #131203 #112815 #94003 #58856 GLOO #115740 #16295 #15963 Tcpstore #61671 #50701 MultiProcessing #65559 MultiThreadedTestCase #107425 #107424 #100459 #100...</div></td><td>04a20f55ce13ee93eb9c8f836de1361967940dc10cc7073e4b90f29a0233183a</td></tr><tr><td>references</td><td>issue #132845</td><td>issue #103625</td><td>medium</td><td><div class='wide'>14 #132447 #131446 #127373 #126868 #126852 #126493 #123294 DDP, FSDP, PiPPy FSDP1 #123726 #105024 FSDP2 #120961 DDP #106361 #104328 #104011 #103625 #77342 #77317 pipeline #121234 Other DCP #113937 #113936 Functional Collectives #107278 ShardedTensor #78068 cc @XilunWu @H-Huang...</div></td><td>7104f447a89af157a51d4b595d5cde1609425c7884a153dfe0b77f4d707f13a3</td></tr><tr><td>references</td><td>issue #132845</td><td>issue #104011</td><td>medium</td><td><div class='wide'>in #132114 #132447 #131446 #127373 #126868 #126852 #126493 #123294 DDP, FSDP, PiPPy FSDP1 #123726 #105024 FSDP2 #120961 DDP #106361 #104328 #104011 #103625 #77342 #77317 pipeline #121234 Other DCP #113937 #113936 Functional Collectives #107278 ShardedTensor #78068 cc @XilunWu...</div></td><td>6a719ae638ac282e95ba9b87e203085b4f5bc6c23cb2aeadeffcde444daac0de</td></tr><tr><td>references</td><td>issue #187473</td><td>issue #73764</td><td>medium</td><td><div class='wide'>ributed 500/shm_broadcast hang) and 4 (CPU-MM 3 timeout) are silent on build #72925 (2026-06-18) — both passed; never filed upstream. Build #73764 status (2026-06-23) Still reproducing (fixes in pytorch main, not yet in test-channel wheel): #187484 (Inductor FxGraphCache guard...</div></td><td>f2ab20b1cd7894d89eca1ab09af87d149a93a12aedd167abdac443781319eb50</td></tr><tr><td>references</td><td>issue #187473</td><td>pr #184193</td><td>medium</td><td><div class='wide'>eam. #187484 — [Inductor] w8a8 block-fp8 BackendCompilerFailed: inductor (DeepGEMM H100, FusedMoE H100/B200, Kernels MoE Test 1–4) · Cause: #184193 · Fix: revert #187666 (release/2.13, bisected — all 75 failures pass reverted) #187735 — [CPU multimodal] qwen2_vl multi-image to...</div></td><td>f90706cfb2736e37955d14ed1b6cce7c4c26e149d7614ae98ce25afef6ad29e3</td></tr><tr><td>references</td><td>issue #158968</td><td>issue #158371</td><td>medium</td><td><div class='wide'>ture, motivation and pitch These enhancements are to have a better UX when using foreach_map, suggested in a few places, but most recently, #158371 (comment) These enhancements should allow easier compiler-first custom optimizer implementations. Alternatives No response Additi...</div></td><td>a06136a5d639c01df4166d0ecc0fe59446cd3972cf8b65cf24fae0d7baf8aa69</td></tr><tr><td>references</td><td>issue #77764</td><td>issue #141287</td><td>medium</td><td><div class='wide'>lace to list and track work on adding support to new ops for the MPS backend. Most requested ops (extracted from comments to this issue and #141287) PyTorch MPS Ops Project : Project to track all the ops for MPS backend. There are a very large number of operators in pytorch an...</div></td><td>43524886f1e21123a7453a8d5ab0713bb84880c7928fbbaa614e5646b57bf808</td></tr><tr><td>references</td><td>issue #139792</td><td>issue #101699</td><td>medium</td><td><div class='wide'>speeding up pickling itself, it would be great to have core support for arrays of strings represented as tensors (and simple data frames): #101699 If there is a million-sized list of python strings, pickle probably would be slower compared to a single tensor of concatenated ut...</div></td><td>00c944a126a20273d4aac658a2c66174e32c8e948480ea861b96d06cad8000e8</td></tr><tr><td>references</td><td>issue #188113</td><td>pr #188114</td><td>medium</td><td><div class='wide'>ses manual attention. It works on both AOTriton v0.11.2 and CK: a, w = attn(x, x, x, need_weights=True) # OK on both backends Updated PR at #188114 with all four files.</div></td><td>91dd135026cbb2814acedea56df2a1908c59fe3c7d459e78130daf7898a0c19e</td></tr><tr><td>references</td><td>issue #184229</td><td>issue #175467</td><td>medium</td><td><div class='wide'>uted.device_mesh import init_device_mesh from torch.distributed.tensor.parallel import ColwiseParallel, parallelize_module # Workaround for #175467 (DTensorSpec not registered as pytree constant). torch.utils._pytree.register_constant( torch.distributed.tensor._dtensor_spec.DT...</div></td><td>5551942bc14ca08353529e9362d5cbd688dbe26d12bff87952fa9e12d33b43f3</td></tr><tr><td>references</td><td>issue #184229</td><td>pr #184508</td><td>medium</td><td><div class='wide'>There is an AI generated fix here: #184508 Which is waiting for review.</div></td><td>92976a5224c4d6aa9a3b36bac0bede9cf2ba2600b8f0bf6608f71e57d8370b41</td></tr><tr><td>references</td><td>issue #187953</td><td>issue #135859</td><td>medium</td><td><div class='wide'>recompile under torch.compile on every new input shape, while the functional variants specialize only once. Reported originally as part of #135859 for linalg.norm and cholesky; the max / min / topk cases (structured ops) are fixed in #187952, but the composite ops need a diffe...</div></td><td>ecfef6ec584732aa9ed14763fda420b2313435bb050dd5ee5a3a838487499aeb</td></tr><tr><td>references</td><td>issue #187953</td><td>pr #187952</td><td>medium</td><td><div class='wide'>ze only once. Reported originally as part of #135859 for linalg.norm and cholesky; the max / min / topk cases (structured ops) are fixed in #187952, but the composite ops need a different fix. Minimal repro: import torch torch._logging.set_logs(recompiles=True) def f(x): out =...</div></td><td>03517d2896bd196f1b2189c6ae7890ac461dccc3b686339326c7962bc719747c</td></tr></tbody></table></div></section></div>

data/work_threads.html

@@ -0,0 +1,75 @@
+
+<style>
+:root { color-scheme: light; }
+.bi-wrap { font-family: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; color: #1f2328 !important; }
+.bi-wrap h1, .bi-wrap h2, .bi-wrap h3, .bi-wrap p, .bi-wrap strong, .bi-wrap td, .bi-wrap th, .bi-wrap div { color: #1f2328 !important; }
+.bi-band { border: 1px solid #d8dee4; border-radius: 8px; background: #ffffff !important; padding: 16px; margin: 0 0 12px; }
+.bi-title { display: flex; gap: 12px; align-items: flex-start; justify-content: space-between; margin-bottom: 12px; }
+.bi-title h1, .bi-title h2 { margin: 0; line-height: 1.15; letter-spacing: 0; }
+.bi-title h1 { font-size: 1.55rem; }
+.bi-title h2 { font-size: 1.12rem; }
+.bi-title p { margin: 4px 0 0; color: #57606a; max-width: 820px; }
+.status-pill { border-radius: 999px; padding: 5px 9px; font-size: .78rem; font-weight: 700; background: #dafbe1; color: #116329 !important; white-space: nowrap; }
+.metric-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(138px, 1fr)); gap: 8px; }
+.metric { border: 1px solid #d0d7de; border-radius: 8px; padding: 10px; background: #f6f8fa !important; min-height: 86px; }
+.metric-value { font-size: 1.45rem; font-weight: 800; line-height: 1.1; color: #0969da !important; }
+.metric-label { margin-top: 4px; font-size: .84rem; font-weight: 700; }
+.metric-note { margin-top: 2px; font-size: .76rem; color: #57606a !important; }
+.callout-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 8px; }
+.callout { border-left: 4px solid #0969da; background: #f6f8fa !important; border-radius: 8px; padding: 10px 12px; }
+.callout:nth-child(2) { border-left-color: #1a7f37; }
+.callout:nth-child(3) { border-left-color: #bf8700; }
+.callout:nth-child(4) { border-left-color: #cf222e; }
+.callout strong { display: block; margin-bottom: 4px; }
+.group-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(310px, 1fr)); gap: 10px; }
+.group-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.group-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.meta-row { display: flex; flex-wrap: wrap; gap: 6px; margin: 6px 0; }
+.chip-row { display: inline-flex; flex-wrap: wrap; gap: 5px; align-items: center; vertical-align: middle; }
+.chip { display: inline-flex; align-items: center; border: 1px solid #d0d7de; border-radius: 999px; padding: 2px 7px; font-size: .78rem; background: #f6f8fa !important; color: #24292f !important; }
+.chip.bucket { border-color: #9a6700; background: #fff8c5 !important; }
+.chip.kind { border-color: #0969da; background: #ddf4ff !important; }
+.workflow-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(260px, 1fr)); gap: 10px; }
+.workflow-card { border: 1px solid #d8dee4; border-radius: 8px; padding: 12px; background: #fff !important; }
+.workflow-card h3 { margin: 0 0 7px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.workflow-card .move-count { font-weight: 800; color: #0969da !important; }
+.move-list { display: grid; gap: 10px; }
+.move-row { border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+.move-row summary { list-style: none; cursor: pointer; padding: 12px; }
+.move-row summary::-webkit-details-marker { display: none; }
+.move-summary { display: grid; grid-template-columns: minmax(190px, 1fr) minmax(260px, 1.45fr) minmax(260px, 1.45fr) minmax(150px, .8fr) minmax(250px, 1.35fr) auto; gap: 16px; align-items: start; }
+.move-title, .move-field { min-width: 0; }
+.move-title h3 { margin: 0 0 4px; font-size: 1rem; line-height: 1.25; letter-spacing: 0; }
+.move-field span { display: block; color: #57606a !important; font-size: .72rem; font-weight: 800; text-transform: uppercase; letter-spacing: 0; }
+.move-field p { margin: 3px 0 0; font-size: .84rem; line-height: 1.35; overflow-wrap: anywhere; }
+.move-field .chip, .move-title { overflow-wrap: anywhere; }
+.view-bundles { align-self: center; justify-self: end; border: 1px solid #000000; border-radius: 999px; color: #000000 !important; background: #ffffff !important; padding: 4px 9px; font-size: .78rem; font-weight: 800; white-space: nowrap; }
+.move-row[open] .view-bundles { background: #f6f8fa !important; }
+.move-expanded { border-top: 1px solid #d8dee4; padding: 0 12px 12px; }
+.bundle-row { border-top: 1px solid #eef1f4; padding: 12px 0; }
+.bundle-row:first-child { border-top: none; }
+.bundle-row h4 { margin: 0 0 6px; font-size: .94rem; line-height: 1.25; letter-spacing: 0; }
+.bundle-scope { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 7px; margin-top: 8px; }
+.evidence-list { margin: 8px 0 0 18px; padding: 0; color: #57606a !important; font-size: .82rem; line-height: 1.35; }
+.evidence-list li { margin: 3px 0; }
+.muted { color: #57606a !important; }
+.evidence { font-size: .84rem; color: #57606a !important; }
+.scroll-table { overflow-x: auto; border: 1px solid #d8dee4; border-radius: 8px; background: #fff !important; }
+table.bi-table { width: 100%; border-collapse: collapse; font-size: .86rem; }
+.bi-table th { text-align: left; background: #f6f8fa !important; color: #57606a !important; font-size: .78rem; border-bottom: 1px solid #d8dee4; padding: 8px; white-space: nowrap; }
+.bi-table td { border-bottom: 1px solid #eef1f4; padding: 8px; vertical-align: top; min-width: 86px; }
+.bi-table tr:last-child td { border-bottom: none; }
+.wide { min-width: 280px; }
+.tight { max-width: 180px; }
+a { color: #0969da !important; text-decoration: none; }
+a:hover { text-decoration: underline; }
+.bi-wrap, .bi-wrap * { color: #000000 !important; }
+@media (max-width: 720px) {
+  .bi-title { display: block; }
+  .status-pill { display: inline-block; margin-top: 8px; }
+  .group-grid { grid-template-columns: 1fr; }
+  .move-summary { grid-template-columns: 1fr; }
+  .view-bundles { justify-self: start; }
+}
+</style>
+<div class='bi-wrap'><section class='bi-band'><div class='bi-title'><div><h2>Work Threads</h2><p>Graph-connected issue/PR units with referenced context and review or CI gates attached.</p></div></div><div class='scroll-table'><table class='bi-table'><thead><tr><th>Thread</th><th>Score</th><th>Issues</th><th>PRs</th><th>Next Action</th><th>Edges</th></tr></thead><tbody><tr><td><strong>PR #188801 - [MPS] Leak MetalShaderLibrary bundled singleton to avoid exit-time destructor crashes</strong><div class='muted'>waiting on contributor</div></td><td>151</td><td><span class='chip-row'><span class='chip'>#188812</span></span></td><td><span class='chip-row'><span class='chip'>PR #188801</span></span></td><td><div class='wide'>Wait for contributor update on PR #188801; keep related issue/PR context attached.</div></td><td>1</td></tr><tr><td><strong>PR #189122 - Fix sparse-dense mul dropping data when broadcasting a size-1 sparse dim</strong><div class='muted'>ready for maintainer decision</div></td><td>139</td><td><span class='chip-row'><span class='chip'>#158861</span><span class='chip'>#188900</span></span></td><td><span class='chip-row'><span class='chip'>PR #189122</span></span></td><td><div class='wide'>Review active PRs and linked issues as one maintainer work thread.</div></td><td>4</td></tr><tr><td><strong>Issue #186535 - Windows, gloo: Access violation (0xC0000005) in ProcessGroupGloo::enqueue when calling allreduce on CUDA tensors — GlooAllreduceRegistry has no kCUDA creator</strong><div class='muted'>needs design decision</div></td><td>139</td><td><span class='chip-row'><span class='chip'>#186535</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #188323 - [Inductor][CPU] dynamic=True convolution lowering crashes with ValueError: Exponent must be non-negative</strong><div class='muted'>needs design decision</div></td><td>139</td><td><span class='chip-row'><span class='chip'>#188323</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #157668 - NCCL error caused due to use of NVLS in torch 2.7.1-cu128 on aarch64 gb200 cluster</strong><div class='muted'>needs design decision</div></td><td>127</td><td><span class='chip-row'><span class='chip'>#157668</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #187912 - [CPU] Concurrent `cpublas::brgemm` calls can crash in the AMX path when the underlying oneDNN ukernel is shared</strong><div class='muted'>needs design decision</div></td><td>127</td><td><span class='chip-row'><span class='chip'>#187912</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>PR #185730 - Fix dynamic shapes for variadic kwargs</strong><div class='muted'>waiting on CI/check fix</div></td><td>121</td><td><span class='chip-row'><span class='chip'>#150022</span><span class='chip'>#150371</span></span></td><td><span class='chip-row'><span class='chip'>PR #185730</span></span></td><td><div class='wide'>Review active PRs and linked issues as one maintainer work thread.</div></td><td>5</td></tr><tr><td><strong>Issue #116254 - C++ API `at::quantized_max_pool2d`: Heap-buffer-overflow</strong><div class='muted'>needs design decision</div></td><td>121</td><td><span class='chip-row'><span class='chip'>#116254</span><span class='chip'>#162476</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>1</td></tr><tr><td><strong>Issue #162422 - Runtime failure when running torch.compile() &amp; using GCC 11.5.0 on Neoverse V1</strong><div class='muted'>needs design decision</div></td><td>121</td><td><span class='chip-row'><span class='chip'>#162422</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #154297 - Hangs and timeouts on dist.reduce_scatter on B200 GPU</strong><div class='muted'>needs maintainer decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#154297</span><span class='chip'>#162178</span><span class='chip'>#162745</span><span class='chip'>#162748</span><span class='chip'>#162820</span><span class='chip'>#162871</span><span class='chip'>#162897</span><span class='chip'>#162917</span><span class='chip'>#162940</span><span class='chip'>#163429</span><span class='chip'>#165170</span><span class='chip'>#165685</span><span class='chip'>#187158</span><span class='chip'>#189065</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs maintainer decision</div></td><td>21</td></tr><tr><td><strong>Issue #55655 - Performance debugging / warning mode</strong><div class='muted'>stale/low urgency</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#55655</span><span class='chip'>#57118</span><span class='chip'>#68768</span><span class='chip'>#72948</span><span class='chip'>#75725</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>stale/low urgency</div></td><td>4</td></tr><tr><td><strong>Issue #66504 - BatchNorm runtimeError: one of the variables needed for gradient computation has been modified by an inplace operation</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#66504</span><span class='chip'>#68407</span><span class='chip'>#73332</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>3</td></tr><tr><td><strong>Issue #144965 - RuntimeError &quot;global alloc not supported yet&quot; when using TorchScript optimization.</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#69078</span><span class='chip'>#144965</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>1</td></tr><tr><td><strong>Issue #106164 - distributed.batch_isend_irecv() crash when send/recv refers to itself</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#106164</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #116423 - PyTorch Distributed Elastic Launch Segmentation Fault with Python 3.12</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#116423</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #119845 - Segmentation fault in dataloader worker sub-process</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#119845</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #145610 - mmap fails on 64k page aarch64 systems for AOTI model loading</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#145610</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #153517 - [CI][CUDA][Distributed] test_non_blocking_with_eager_init timeout</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#153517</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #162731 - DTensor cached op propagation results can be mutated when propagating other ops</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#162731</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #167693 - DISABLED test_side_stream_backward_overlap_cuda (__main__.TestAutogradStreamSynchronizationCUDA)</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#167693</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #170003 - batch_isend_irecv with nccl causes illegal memory access depending on P2P ordering</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#170003</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #173921 - [Inductor] Significant numerical divergence (4.7% relative error) in Conv2d on CPU with torch.compile</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#173921</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #174288 - [distributed] Batched isend/irecv with NCCL backend hangs on high load</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#174288</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #179502 - [DTensor] sharded view incorrectly passes when redistribution is needed</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#179502</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #180088 - [DTensor] RNG tracker does not advance state for CPU tensors on CUDA mesh, causing trunc_normal_ infinite loop</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#180088</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #183459 - Regression in torch.distributed._functional_collectives.all_to_all_single in 2.9.1 -&gt; 2.11</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#183459</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #189126 - Report a bug, as requested by the error message</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#189126</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #189144 - [XPU][B580] `flex-attn-causal` performance drop with `last_level_cache_size` cache clear between runs</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#189144</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #189150 - SIGSEGV / cudaErrorIllegalAddress when replaying a CUDA graph that captured multiple training iterations of two trainers over deeply unrolled recurrent modules</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#189150</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #189194 - [ROCm] torch 2.13 wheel: &quot;Can&#x27;t detect vectorized ISA for CPU&quot; in torch.compile smoke test on non-ROCm image (regression vs 2.12.1)</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#189194</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #189239 - [Inductor]Unable to apply layout optimization on convolution backward causes training performance regression.</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#189239</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #189281 - [CPU] operator_benchmark: embedding / embeddingbag ~70-130x slower since ~Nov 2025 (x86 + aarch64; possible measurement issue)</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#189281</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #189282 - [CPU] operator_benchmark: backward-pass (add, + batchnorm on aarch64) ~200-500x slower since ~Nov 2025 (x86 + aarch64; possible measurement artifact)</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#189282</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #71303 - [RFC] Cross-Process Performance Analysis: Straggler Detection</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#71303</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #76449 - Enhance _verify_param_shape_across_processes</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#76449</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #81684 - Message exchange failure when perform alltoallv (cpus)</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#81684</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>Issue #85088 - reentrant torch.utils.checkpoint does not work with NamedTuple outputs</strong><div class='muted'>needs design decision</div></td><td>115</td><td><span class='chip-row'><span class='chip'>#85088</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>0</td></tr><tr><td><strong>PR #186461 - Fix SymBool equality handling in symbolic shapes</strong><div class='muted'>ready for maintainer decision</div></td><td>112</td><td><span class='chip-row'><span class='chip'>#124110</span></span></td><td><span class='chip-row'><span class='chip'>PR #186461</span></span></td><td><div class='wide'>Review active PRs and linked issues as one maintainer work thread.</div></td><td>1</td></tr><tr><td><strong>PR #186429 - Fix ONNX export of pad_sequence with symbolic split lengths</strong><div class='muted'>ready for maintainer decision</div></td><td>112</td><td><span class='chip-row'><span class='chip'>#127153</span></span></td><td><span class='chip-row'><span class='chip'>PR #186429</span></span></td><td><div class='wide'>Review active PRs and linked issues as one maintainer work thread.</div></td><td>1</td></tr><tr><td><strong>PR #185691 - Avoid UB in float to signed integer casts</strong><div class='muted'>ready for maintainer decision</div></td><td>112</td><td><span class='chip-row'><span class='chip'>#151510</span></span></td><td><span class='chip-row'><span class='chip'>PR #185691</span></span></td><td><div class='wide'>Review active PRs and linked issues as one maintainer work thread.</div></td><td>1</td></tr><tr><td><strong>PR #184716 - Fix conv_transpose2d meta output padding validation</strong><div class='muted'>ready for maintainer decision</div></td><td>112</td><td><span class='chip-row'><span class='chip'>#178127</span></span></td><td><span class='chip-row'><span class='chip'>PR #184716</span></span></td><td><div class='wide'>Review active PRs and linked issues as one maintainer work thread.</div></td><td>1</td></tr><tr><td><strong>PR #188632 - Fix Dynamo opaque object staticmethod guards</strong><div class='muted'>ready for maintainer decision</div></td><td>112</td><td><span class='chip-row'><span class='chip'>#188544</span></span></td><td><span class='chip-row'><span class='chip'>PR #188632</span></span></td><td><div class='wide'>Review active PRs and linked issues as one maintainer work thread.</div></td><td>3</td></tr><tr><td><strong>PR #184939 - Support SymInt steps for linspace/logspace export</strong><div class='muted'>ready for maintainer decision</div></td><td>112</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><span class='chip-row'><span class='chip'>PR #184939</span></span></td><td><div class='wide'>ready for maintainer decision</div></td><td>0</td></tr><tr><td><strong>PR #184648 - [Test] Add has_sufficient_memory() hook to DeviceTypeTestBase</strong><div class='muted'>waiting on CI/check fix</div></td><td>108</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><span class='chip-row'><span class='chip'>PR #184648</span></span></td><td><div class='wide'>PR blocked</div></td><td>1</td></tr><tr><td><strong>PR #185882 - [PrivateUse1] Return `False` instead of `None` when PU1 backend is not available</strong><div class='muted'>waiting on CI/check fix</div></td><td>108</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><span class='chip-row'><span class='chip'>PR #185882</span></span></td><td><div class='wide'>PR blocked</div></td><td>0</td></tr><tr><td><strong>PR #186927 - [MPS] Gemv kernels</strong><div class='muted'>waiting on CI/check fix</div></td><td>108</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><span class='chip-row'><span class='chip'>PR #186927</span></span></td><td><div class='wide'>PR blocked</div></td><td>0</td></tr><tr><td><strong>PR #187501 - [Stable C Shim] Use error message retrieval shim if available at runtime</strong><div class='muted'>waiting on CI/check fix</div></td><td>108</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><span class='chip-row'><span class='chip'>PR #187501</span></span></td><td><div class='wide'>PR blocked</div></td><td>0</td></tr><tr><td><strong>PR #188237 - Optimize CSR SpMM CPU grain size and inner accumulation</strong><div class='muted'>waiting on CI/check fix</div></td><td>108</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><span class='chip-row'><span class='chip'>PR #188237</span></span></td><td><div class='wide'>PR blocked</div></td><td>1</td></tr><tr><td><strong>PR #188301 - [inductor] Fix inductor dropping ordering dep between effectful ops with different kernel types</strong><div class='muted'>waiting on CI/check fix</div></td><td>108</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><span class='chip-row'><span class='chip'>PR #188301</span></span></td><td><div class='wide'>PR blocked</div></td><td>2</td></tr><tr><td><strong>PR #189307 - [CI/CD] Copy newer CUPTI headers into manywheel binary-build images</strong><div class='muted'>waiting on CI/check fix</div></td><td>108</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><span class='chip-row'><span class='chip'>PR #189307</span></span></td><td><div class='wide'>PR blocked</div></td><td>0</td></tr><tr><td><strong>PR #189312 - [CUDA][cuBLAS] Change cuBLAS default workspace size for SM 11.0 to 32 MiB</strong><div class='muted'>waiting on CI/check fix</div></td><td>108</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><span class='chip-row'><span class='chip'>PR #189312</span></span></td><td><div class='wide'>PR blocked</div></td><td>0</td></tr><tr><td><strong>PR #188948 - Fix triangular_solve for sparse CPU tensors on non-MKL platforms</strong><div class='muted'>waiting on contributor</div></td><td>107</td><td><span class='chip-row'><span class='chip'>#153410</span></span></td><td><span class='chip-row'><span class='chip'>PR #188948</span></span></td><td><div class='wide'>Wait for contributor update on PR #188948; keep related issue/PR context attached.</div></td><td>2</td></tr><tr><td><strong>PR #181720 - [MPS] Make pin_memory return CPU-aliased storage backed by a unified MTLBuffer</strong><div class='muted'>has linked issue</div></td><td>106</td><td><span class='chip-row'><span class='chip'>#180397</span><span class='chip'>#181374</span><span class='chip'>#188970</span></span></td><td><span class='chip-row'><span class='chip'>PR #181720</span><span class='chip'>PR #189256</span></span></td><td><div class='wide'>Review active PRs and linked issues as one maintainer work thread.</div></td><td>9</td></tr><tr><td><strong>PR #189043 - Preload full bundled cuDNN set with RTLD_GLOBAL to prevent sublibrary version mismatch</strong><div class='muted'>waiting on contributor</div></td><td>106</td><td><span class='chip-row'><span class='chip'>#188892</span></span></td><td><span class='chip-row'><span class='chip'>PR #189043</span></span></td><td><div class='wide'>Wait for contributor update on PR #189043; keep related issue/PR context attached.</div></td><td>2</td></tr><tr><td><strong>PR #135631 - [scan] Autograd</strong><div class='muted'>waiting on CI/check fix</div></td><td>100</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><span class='chip-row'><span class='chip'>PR #135631</span></span></td><td><div class='wide'>PR blocked</div></td><td>0</td></tr><tr><td><strong>PR #188115 - [Inductor][TEST] Align `test_main_loop_scaling` with H100 support surface</strong><div class='muted'>waiting on CI/check fix</div></td><td>98</td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><span class='chip-row'><span class='chip'>PR #188115</span></span></td><td><div class='wide'>PR blocked</div></td><td>0</td></tr><tr><td><strong>Issue #103625 - DISABLED test_backward_ddp_outside (__main__.TensorPipeDdpUnderDistAutogradTest)</strong><div class='muted'>needs maintainer decision</div></td><td>97</td><td><span class='chip-row'><span class='chip'>#15963</span><span class='chip'>#16295</span><span class='chip'>#51062</span><span class='chip'>#58856</span><span class='chip'>#64093</span><span class='chip'>#69178</span><span class='chip'>#69179</span><span class='chip'>#94003</span><span class='chip'>#103625</span><span class='chip'>#104011</span><span class='chip'>#132845</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs maintainer decision</div></td><td>10</td></tr><tr><td><strong>Issue #104506 - TImeout in NCCL doesn&#x27;t work</strong><div class='muted'>stale/low urgency</div></td><td>97</td><td><span class='chip-row'><span class='chip'>#88103</span><span class='chip'>#104506</span><span class='chip'>#124950</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>stale/low urgency</div></td><td>2</td></tr><tr><td><strong>Issue #42258 - Cannot load certain function from dumped Torchscript file</strong><div class='muted'>needs design decision</div></td><td>97</td><td><span class='chip-row'><span class='chip'>#42258</span><span class='chip'>#46944</span><span class='chip'>#46946</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>2</td></tr><tr><td><strong>Issue #58111 - Static graph training fails if forward is called multiple times before backward</strong><div class='muted'>needs design decision</div></td><td>97</td><td><span class='chip-row'><span class='chip'>#58111</span><span class='chip'>#62998</span><span class='chip'>#70865</span></span></td><td><span class='chip-row'><span class='muted'>none</span></span></td><td><div class='wide'>needs triage</div></td><td>2</td></tr></tbody></table></div></section></div>

space_release_manifest.json

@@ -0,0 +1,243 @@
+{
+  "manifest_version": "github_backlog_hf_space_v1",
+  "generated_utc": "2026-07-09T04:37:59+00:00",
+  "repo_id": "cjc0013/pytorch-backlog-intelligence",
+  "space_url": "https://huggingface.co/spaces/cjc0013/pytorch-backlog-intelligence",
+  "space_visibility": "public",
+  "source_report_version": "github_backlog_report_v0",
+  "source_packet_hash": "bb2ad53aa6630f3381c2a30f2e4225d98ab5e28172ceb254137002bd27dd863c",
+  "source_repo": "pytorch/pytorch",
+  "artifact_count": 32,
+  "artifacts": [
+    {
+      "artifact_key": ".gitattributes",
+      "relative_path": ".gitattributes",
+      "exists": true,
+      "sha256": "fa428368c44c8737e2874156afc31aa7cfafae921bf4cdd29da7feb75464b990",
+      "byte_count": 36
+    },
+    {
+      "artifact_key": "data/attention_queue.csv",
+      "relative_path": "data/attention_queue.csv",
+      "exists": true,
+      "sha256": "bc3ec99fb711314bfe4fc2481834a2d67668c2cb405cc14cce0c9fd99ca1a86b",
+      "byte_count": 5583108
+    },
+    {
+      "artifact_key": "data/attention_queue.html",
+      "relative_path": "data/attention_queue.html",
+      "exists": true,
+      "sha256": "4f5861217011f1b8d8ba6a4c84103f5f99063fcc4ef6f2b34083041f2e280497",
+      "byte_count": 34904
+    },
+    {
+      "artifact_key": "data/backlog_report.md",
+      "relative_path": "data/backlog_report.md",
+      "exists": true,
+      "sha256": "dfe8793706a9e2406c3cad655e3c9a5b6e9b6f2e662a8a87505758598554240e",
+      "byte_count": 20419
+    },
+    {
+      "artifact_key": "data/cluster_cards.csv",
+      "relative_path": "data/cluster_cards.csv",
+      "exists": true,
+      "sha256": "6c1a5be790e8628c02a6f3ad6f345d09d4bdb958c862485d053ed444863c030f",
+      "byte_count": 137713
+    },
+    {
+      "artifact_key": "data/cluster_cards.html",
+      "relative_path": "data/cluster_cards.html",
+      "exists": true,
+      "sha256": "dadf793b5730cd3021888ea8366bcadd64a2933fd286a8013b3d1bfe3029d39d",
+      "byte_count": 27166
+    },
+    {
+      "artifact_key": "data/dashboard.html",
+      "relative_path": "data/dashboard.html",
+      "exists": true,
+      "sha256": "2c528676fb43d6575779cbc74ec1ee70cdb5e50a0baf7a5bbdf62fefab4e2adc",
+      "byte_count": 11815
+    },
+    {
+      "artifact_key": "data/diagnostics.html",
+      "relative_path": "data/diagnostics.html",
+      "exists": true,
+      "sha256": "7dd648c77da00b483de7ca53662ff321c765cba718046b16c243c5a2353a0bdc",
+      "byte_count": 131595
+    },
+    {
+      "artifact_key": "data/downloads.html",
+      "relative_path": "data/downloads.html",
+      "exists": true,
+      "sha256": "ec4a503253c547c99bf8f70d071c82328d9c00c613d7bb2a14e1fcf7685e3a1d",
+      "byte_count": 7664
+    },
+    {
+      "artifact_key": "data/dry_run_proof.html",
+      "relative_path": "data/dry_run_proof.html",
+      "exists": true,
+      "sha256": "83c5a7096f367b87c128c3512a613b96fcae45e42cbf926e13ae31edc99dacf4",
+      "byte_count": 90356
+    },
+    {
+      "artifact_key": "data/dry_run_proof_actions.csv",
+      "relative_path": "data/dry_run_proof_actions.csv",
+      "exists": true,
+      "sha256": "11bb3014ceea96ac5cabc75a3d77cacf8f44f8ada1ebcd57265f25c4f4e8d02c",
+      "byte_count": 12258994
+    },
+    {
+      "artifact_key": "data/evidence_bundles.csv",
+      "relative_path": "data/evidence_bundles.csv",
+      "exists": true,
+      "sha256": "1c029575a4657a3eadb63cee311e56744dba2938476f5ff373c9ca48a2ecd166",
+      "byte_count": 2070899
+    },
+    {
+      "artifact_key": "data/evidence_bundles.html",
+      "relative_path": "data/evidence_bundles.html",
+      "exists": true,
+      "sha256": "f43bb71a0d71cf978c2ef6db64cdeb18491c97fb30b94bc56762b1798677d014",
+      "byte_count": 128986
+    },
+    {
+      "artifact_key": "data/issue_collapse_groups.csv",
+      "relative_path": "data/issue_collapse_groups.csv",
+      "exists": true,
+      "sha256": "0d78225e5add603589e8c4ad2ab1cefc7ab9c4b823f6f6e32d7ec9c4250eb3f2",
+      "byte_count": 2625620
+    },
+    {
+      "artifact_key": "data/issue_collapse_groups.html",
+      "relative_path": "data/issue_collapse_groups.html",
+      "exists": true,
+      "sha256": "51bb4919d17e5d2bc0d9d09f77b217b6f1d3ce59c96ac8d292dbe8871124b527",
+      "byte_count": 1552758
+    },
+    {
+      "artifact_key": "data/issues.csv",
+      "relative_path": "data/issues.csv",
+      "exists": true,
+      "sha256": "8bb02ea7cd2de5290825dddd91b839bd43ae21c8d70c3462da77c53e5f1f2054",
+      "byte_count": 12173666
+    },
+    {
+      "artifact_key": "data/issues.html",
+      "relative_path": "data/issues.html",
+      "exists": true,
+      "sha256": "4889f51c66da62589347e099c497d9b8453f9a4fb8853d9e62934b1090edbcfe",
+      "byte_count": 83399
+    },
+    {
+      "artifact_key": "data/maintainer_moves.csv",
+      "relative_path": "data/maintainer_moves.csv",
+      "exists": true,
+      "sha256": "d8754fb08d4c32a578cf4765f862e682161f4d7b17ff0d67634b10e640f6f4ad",
+      "byte_count": 358873
+    },
+    {
+      "artifact_key": "data/maintainer_moves.html",
+      "relative_path": "data/maintainer_moves.html",
+      "exists": true,
+      "sha256": "1db031a6fc9bbe4cc6ebe5b26072048a4f674c55b8bdc51547d038fcae96e6ff",
+      "byte_count": 171633
+    },
+    {
+      "artifact_key": "data/manifest.html",
+      "relative_path": "data/manifest.html",
+      "exists": true,
+      "sha256": "edae3c4bc1bddd034897848912a7e394cfb0235a2bebf244bd26f7e7f08072d6",
+      "byte_count": 6814
+    },
+    {
+      "artifact_key": "data/pull_requests.csv",
+      "relative_path": "data/pull_requests.csv",
+      "exists": true,
+      "sha256": "244e75ceba0b3346e4d4d27ff9c883e2c5c00e65543603325da1beae03c86efa",
+      "byte_count": 2186769
+    },
+    {
+      "artifact_key": "data/pull_requests.html",
+      "relative_path": "data/pull_requests.html",
+      "exists": true,
+      "sha256": "2ddd1131a9d392419fd0732141b60edfeded20a67804e9f97c045805148fc6e8",
+      "byte_count": 80850
+    },
+    {
+      "artifact_key": "data/relationship_edges.csv",
+      "relative_path": "data/relationship_edges.csv",
+      "exists": true,
+      "sha256": "e104e75e93ff5d704d9d111f5a6b935c327113897875a01905de26a531225a1b",
+      "byte_count": 1502516
+    },
+    {
+      "artifact_key": "data/relationship_edges.html",
+      "relative_path": "data/relationship_edges.html",
+      "exists": true,
+      "sha256": "fee481df0145d2c61612bbdf382d2b300684bfc8251e82df601e90529d254c33",
+      "byte_count": 61289
+    },
+    {
+      "artifact_key": "data/simulated_resolution_packages.html",
+      "relative_path": "data/simulated_resolution_packages.html",
+      "exists": true,
+      "sha256": "428f1fa97663140e3758a7c7438130bd314910ce236c7d6509406841301a04b6",
+      "byte_count": 141310
+    },
+    {
+      "artifact_key": "data/summary_metrics.json",
+      "relative_path": "data/summary_metrics.json",
+      "exists": true,
+      "sha256": "093ea5e3fdec9b5905a4ec7243d2a638d9127dcd54b0c83a2e30aa847a5d32dd",
+      "byte_count": 110310
+    },
+    {
+      "artifact_key": "data/work_threads.csv",
+      "relative_path": "data/work_threads.csv",
+      "exists": true,
+      "sha256": "791d9b935a3945f7275b9a05844ec4b7a86696a2b0da48ef624e46f355d5a1eb",
+      "byte_count": 3499011
+    },
+    {
+      "artifact_key": "data/work_threads.html",
+      "relative_path": "data/work_threads.html",
+      "exists": true,
+      "sha256": "1744e7e2edfa2dc305951a4fbe33129fd818e78ea2316ac14fc6b9c6d5745ba2",
+      "byte_count": 31445
+    },
+    {
+      "artifact_key": "data/workflow_lanes.csv",
+      "relative_path": "data/workflow_lanes.csv",
+      "exists": true,
+      "sha256": "1435d5033dce20e11715e82276265210ef3a69f35d4a7e9018e2b4024361e37e",
+      "byte_count": 283175
+    },
+    {
+      "artifact_key": "data/workflow_lanes.html",
+      "relative_path": "data/workflow_lanes.html",
+      "exists": true,
+      "sha256": "1b2a23958170db33bc5382d3d4544e12451652797899f8100f1744955cfabb89",
+      "byte_count": 207729
+    },
+    {
+      "artifact_key": "index.html",
+      "relative_path": "index.html",
+      "exists": true,
+      "sha256": "5b365045b950276814a052cd92ebed65d02cf073697e22ebe1a1bb486cf4553a",
+      "byte_count": 2551382
+    },
+    {
+      "artifact_key": "README.md",
+      "relative_path": "README.md",
+      "exists": true,
+      "sha256": "b9ab9331b09f1cbe5a6afa0afad439f79e65854b5d0b00a38c623df74c9d9637",
+      "byte_count": 2467
+    }
+  ],
+  "github_writes": "none",
+  "bundle_mode": "public_static",
+  "promotion_ready": true,
+  "raw_api_payloads_included": false,
+  "local_paths_public": false,
+  "manifest_hash": "7f9781c33f0116ed6e7722a30b7dae3b82f736a0919e0d643c6b4ff6439e64c1"
+}

style.css

@@ -1,28 +0,0 @@
-body {
-	padding: 2rem;
-	font-family: -apple-system, BlinkMacSystemFont, "Arial", sans-serif;
-}
-
-h1 {
-	font-size: 16px;
-	margin-top: 0;
-}
-
-p {
-	color: rgb(107, 114, 128);
-	font-size: 15px;
-	margin-bottom: 10px;
-	margin-top: 5px;
-}
-
-.card {
-	max-width: 620px;
-	margin: 0 auto;
-	padding: 16px;
-	border: 1px solid lightgray;
-	border-radius: 16px;
-}
-
-.card p:last-child {
-	margin-bottom: 0;
-}