Search results

Update Rendezvous etcd backend to use etcd v3 API

Status: Open.

#170707 In pytorch/pytorch;

·siyuanfoundation openedon Dec 17, 2025

[torch.compile] <code>CatchErrorsWrapper' object has no attribute '__closure__</code> when wrapping a fullgraph'd function in fullgraph'd module forward

oncall: pt2

Status: Open.

#170696 In pytorch/pytorch;

·vadimkantorov openedon Dec 17, 2025

Proposal to simplify <code>DeviceThreadHandlePool</code>

Status: Open.

#170694 In pytorch/pytorch;

·lakshayg openedon Dec 17, 2025

[Feature Request] Add StateSpaceModel to torch.nn

Status: Open.

#170691 In pytorch/pytorch;

·vishesh9131 openedon Dec 17, 2025

[BUG][Inconsistency]Numerical mismatch between eager and inductor for aten.__interpolate

oncall: pt2

Status: Open.

#170688 In pytorch/pytorch;

·jikechao openedon Dec 17, 2025

[BUG][Inductor ] Inductor fails on aten._to_sparse_bsc due to incorrect stride assumption for Sparse BSC tensors

module: sparseRelated to torch.sparseRelated to torch.sparse

oncall: pt2

Status: Open.

#170686 In pytorch/pytorch;

·jikechao openedon Dec 17, 2025

[Bug][Inductor] Numerical inconsistency with eager for aten.cosine_similarity

oncall: pt2

Status: Open.

#170684 In pytorch/pytorch;

·jikechao openedon Dec 17, 2025

[BUG][Inconsistency]aten.__upsample_nearest gives different results on Inductor vs Eager with bfloat16

oncall: pt2

Status: Open.

#170682 In pytorch/pytorch;

·jikechao openedon Dec 17, 2025

Error when composing jit.script and jacrev

oncall: jitAdd this issue/PR to JIT oncall triage queueAdd this issue/PR to JIT oncall triage queue

Status: Open.

#170680 In pytorch/pytorch;

·Giodiro openedon Dec 17, 2025

Wrapping a DistributedDataParallel module with another One leads to incorrect behavior.

Status: Open.

#170678 In pytorch/pytorch;

·Gao-HaoYuan openedon Dec 17, 2025

CUBLAS_STATUS_INVALID_VALUE at F.linear((14,4096), (4096,4096)) device='cuda:0', dtype=torch.float16 on 4090 and B300

Status: Open.

#170676 In pytorch/pytorch;

·Triang-jyed-driung openedon Dec 17, 2025

Flex attention cannot support page size/block mask size != 128 for prefill.

Status: Open.

#170674 In pytorch/pytorch;

·man2machine openedon Dec 17, 2025