Issues

• torch-directml: F.cross_entropy(reduction='none') silently produces zero gradient on backward (0.2.5.dev240914 / torch 2.4.1+cpu)

  microsoft/DirectML#740 · Filter by author Rook-CodeVolt opened on Sep 18, 2026

• torch-directml: F.cross_entropy(reduction='none') silently produces a zero backward gradient (loss value correct, gradient vanishes)

  microsoft/DirectML#739 · Filter by author Rook-CodeVolt opened on Sep 14, 2026

• DX12 x64 WDDM 3.0 WSL DirectML Metacommands Tests need L"C:\\Program Files\\DirectMLTestContent\\bin"

  microsoft/DirectML#738 · Filter by author wak1984 opened on Aug 14, 2026

• torch-directml: Tensor *= bool_tensor silently produces bool dtype instead of following standard type promotion

  microsoft/DirectML#737 · Filter by author gucciwong opened on Aug 1, 2026

• BindTemporaryResource hangs on new IDMLDevice after prior device removal during operator initializer dispatch

  microsoft/DirectML#735 · Filter by author ALWAYSxREADY opened on May 1, 2026
  2

• Training YOLO on AMD GPU in window

  microsoft/DirectML#733 · Filter by author HUYVIESEA opened on Apr 3, 2026

• Request: to open-source the torch_directml native extension now that DirectML is in maintenance mode

  microsoft/DirectML#729 · Filter by author yanite opened on Feb 4, 2026

• "Invalid Argument (80070057)" crash with Transformer-based models (D-FINE/RT-DETR) due to Int64 indices in Gather/Scatter ops

  microsoft/DirectML#727 · Filter by author LeoYang06 opened on Jan 23, 2026
  2

• "Invalid device_id argument supplied" -> typo

  microsoft/DirectML#723 · Filter by author GitMensch opened on Jan 3, 2026

• Access to latest defines

  microsoft/DirectML#722 · Filter by author tgraupmann opened on Dec 20, 2025

• The operator 'aten::upsample_nearest1d.out' is not currently supported on the DML backend and will fall back to run on the CPU

  microsoft/DirectML#718 · Filter by author CARLPC opened on Sep 12, 2025

• does not see the video card

  microsoft/DirectML#716 · Filter by author artemj23 opened on Sep 5, 2025
  1