openvino · Issues· 881 open

[Good First Issue] [RISCV64]: Implement CPU plugin just-in-time emitter for Swish operation

[Bug]: LFM2.5 350M and 1.2B silently generate garbage on NPU DEVICE_ARCHITECTURE=4000 (Lunar Lake) — correct on 3720, CPU and GPU

[Bug]: bf16 MatMul/GEMM returns corrupted output (cos 0.28-0.65, 5-135x magnitude) - f16 and f32 controls are correct

[Bug]: Qwen3.8-27B with MTP cannot run across 2x Arc Pro B70: PIPELINE_PARALLEL crashes in igdrcl64.dll, TENSOR_PARALLEL unsupported

[Bug]: HETERO:GPU.0,GPU.1 + PIPELINE_PARALLEL silently places whole model on GPU.0 when it fits one card (2x Arc Pro B70, Qwen3.8-27B)

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

[aw] Detection Runs

[Bug]: CPU and GPU plugins ignore Divide's m_pythondiv for integer inputs: -5 / 2 gives -2 instead of -3

[Bug]: [PyTorch Frontend] aten::narrow with a negative start returns an empty tensor

[Bug]: ROI tensor keeps its owner's strides after set_shape() expands the rank — ITensor::copy_to() then reads out of bounds (SIGSEGV in production)

[Bug]: OpenVINO's GPU plugin segfaults on intel 12th gen cpu

[Performance]: [GPU] PagedAttention with 2-4 query tokens per sequence (speculative decoding verify) is ~8x slower than single-token decode at the same context

[Bug]: PyTorch frontend computes `sum((x-y)^p)^(1/p)` for `aten::pairwise_distance` / `aten::cdist` — the `Abs()` is missing, so every `p != 2` is wrong or NaN

[Bug]: [PyTorch Frontend] Incorrect translation for aten::slice when step is negative and start/end bounds are omitted

[Bug]: [GPU] StridedSlice/Slice with non-constant strides is skipped as identity when the output shape equals the input shape (e.g. reversal with step -1)