All issues

`RuntimeError: Attempting to use a closed pool` on DeepSeek-v3-16b torchtitan on Blackwell + torch nightly + CUDA 13.4

#3944

· syed-ahmed opened

on Jul 17, 2026

[rl] Flex-attention generator crashes with "Dynamo recompile limit exceeded (recompile_limit=8)"

module: torch.compile

#3898

· wwwjn opened

on Jul 10, 2026

Add the required_grads of parameter to register_parameter

#3885

· qyqc731 opened

on Jul 8, 2026

[RL/DeepEP] DeepEP v2 combine deadlocks at 8-host EP=64 (scale-up wait timeout -> CUDA launch failure)

#3877

· yichuan-w opened

on Jul 7, 2026

[graph_trainer] Skip all test_eager_self_deterministic: hardcoded eager goldens unmaintainable vs moving torch nightly

#3876

· SherlockNoMad opened

on Jul 7, 2026

[graph_trainer] Flaky: test_graph_pp_moe_dsv3_aot_fx_trace_vs_eager (DSv3 MoE EP all-to-all nondeterminism under --debug.deterministic)

#3874

· SherlockNoMad opened

on Jul 6, 2026

[RL/DeepEP] Generator OOMs in vLLM cudagraph capture on memory-tight MoE (expandable_segments incompatible with DeepEP; gpu_memory_utilization is a no-op)

#3865

· yichuan-w opened

on Jul 6, 2026

DeepEP v2 doesn't work with per-layer compile

#3830

· wwwjn opened

on Jun 30, 2026

RFC CUDAGraphable GraphPP

#3820

· sanketpurandare opened

on Jun 29, 2026

TP + MXFP8 crashing with "mat2 must be col_major, got stride (4096, 1)" error

#3817

· vallet-alexis-col01-sbint opened

on Jun 29, 2026

GraphTrainer DSv3-16B: healthy single-node, MI355 FSDP EP=2 parity gap

module: torch.compile

#3816

· rishisinhanj opened

on Jun 29, 2026

8 GPU Model Tests red on main: gpt-oss PP+FSDP+EP+SACOP MoE _grouped_mm index OOB (CUDA scatter assert == ROCm HSA 0x1016)

#3761

· rishisinhanj opened

on Jun 24, 2026