Issues

• [Bug]: Keras layer tests never run in precommit because of duplicate test names

  openvinotoolkit/openvino#38997 · Filter by author Daksha1611 opened on Oct 19, 2026

• [Bug]: ErfInv returns inaccurate results on CPU (sign typo in a Giles coefficient)

  bugsupport_request
  openvinotoolkit/openvino#38994 · Filter by author lequangdung2005 opened on Oct 19, 2026

• [Bug]: NPU returns wrong output for single channel Interpolate with align_corners

  openvinotoolkit/openvino#38989 · Filter by author onuralpszr opened on Oct 19, 2026

• [Bug]: NPU returns wrong output for Clamp followed by Exp

  openvinotoolkit/openvino#38988 · Filter by author onuralpszr opened on Oct 19, 2026

• [aw] CI Failure Doctor — Merge Queue produced no safe outputs

  agentic-workflows
  openvinotoolkit/openvino#38981 · Filter by author github-actions opened on Oct 19, 2026

• [Bug]: 量化后的onnx使用openvino推理无结果

  bugsupport_request
  openvinotoolkit/openvino#38972 · Filter by author elee-work opened on Oct 19, 2026
  2

• [Bug]: torchvision preprocessing tracks the wrong image size after a two-value Pad

  openvinotoolkit/openvino#38968 · Filter by author Daksha1611 opened on Oct 19, 2026

• [Bug] ONNX FE: com.microsoft.RotaryEmbedding ignores the start-offset form of position_ids (shape [1])

  openvinotoolkit/openvino#38966 · Filter by author karanshukla opened on Oct 19, 2026

• [Bug]: GPU plugin returns NaN from token 1024 in NNCF-int8 SDPA with a sliding-window mask (decomposed into gemm_mmad_int8_slm)

  bugsupport_request
  openvinotoolkit/openvino#38964 · Filter by author Gabriellgpc opened on Oct 18, 2026

• [Bug]: [NPU] MiniCPM-V 4.7 1B fails to compile on NPU (vpux FeasibleAllocation: Scheduler failure, no buffer to spill)

  openvinotoolkit/openvino#38962 · Filter by author Tariya321 opened on Oct 18, 2026

• [Bug]: ConvertRaises.test_failed_extension is defined twice, so the ConversionExtension failure check never runs

  openvinotoolkit/openvino#38958 · Filter by author Daksha1611 opened on Oct 18, 2026

• [Bug]: [PyOV] float32 input to a bf16 model input gives wrong values or raises

  openvinotoolkit/openvino#38950 · Filter by author Somay-kousis opened on Oct 17, 2026

• [Bug]: test_softmax and test_inverse PyTorch layer tests are shadowed by duplicate definitions

  openvinotoolkit/openvino#38947 · Filter by author Daksha1611 opened on Oct 17, 2026

• [Bug]: TensorFlow FE: FloorDiv and TruncateDiv are one off for exact multiples of the divisor (41 // 41 gives 0 in float32)

  openvinotoolkit/openvino#38945 · Filter by author usmanmateen opened on Oct 17, 2026

• [Bug]: [PyOV] from_torchvision fails with a torch.Tensor input_example on NumPy 2

  openvinotoolkit/openvino#38937 · Filter by author Somay-kousis opened on Oct 17, 2026

• [Bug]: select_scatter and slice_scatter PyTorch layer tests are never collected

  openvinotoolkit/openvino#38935 · Filter by author Daksha1611 opened on Oct 17, 2026
  2

• [Bug]: High memory usage when using OFFLOAD_RATIO (regression)

  bugsupport_request
  openvinotoolkit/openvino#38930 · Filter by author WizardlyBump17 opened on Oct 16, 2026

• [Bug]: CPU Mod and FloorMod return 0 instead of nearly the divisor for float32 multiples (fmod(4.2, 1.4) gives 0)

  openvinotoolkit/openvino#38929 · Filter by author usmanmateen opened on Oct 16, 2026

• [aw] Failed jobs: CI Failure Doctor

  agentic-workflows
  openvinotoolkit/openvino#38911 · Filter by author github-actions opened on Oct 16, 2026

• [aw] CI Failure Doctor failed

  agentic-workflows
  openvinotoolkit/openvino#38910 · Filter by author github-actions opened on Oct 16, 2026

• [Bug]: [PyOV] from_torchvision CenterCrop drops the last row/column when the crop equals the image size

  openvinotoolkit/openvino#38883 · Filter by author Somay-kousis opened on Oct 15, 2026

• [Feature Request]: [GPU] MoE offload: check the weights file is the one the model was read from

  enhancementfeature
  openvinotoolkit/openvino#38864 · Filter by author blairducrayoppat opened on Oct 15, 2026

• [Bug]: CPU plugin's SigAltStackSetup restores an unmapped sigaltstack when Cores are destroyed out of order (SIGSEGV on the next signal)

  openvinotoolkit/openvino#38850 · Filter by author inkwan00 opened on Oct 14, 2026
  1

• [Bug]: PyTorch FE: x // d by a constant is one too small for exact multiples (41 // 41 gives 0 in float32)

  openvinotoolkit/openvino#38832 · Filter by author usmanmateen opened on Oct 13, 2026

• [Bug]: [ARM][CPU] INT8 models with a dynamic batch return unscaled outputs or ±inf: oneDNN FullyConnected fallback drops per-channel dequantization scales

  bugsupport_request
  openvinotoolkit/openvino#38831 · Filter by author useribraim opened on Oct 13, 2026