Search results

ATOL and RTOL should probably be configurable per-test. Ideally, we also make the defaults as small as possible and override it on tests that need it.

Status: Open.

#18424 In pytorch/executorch;

· psiddh opened on Mar 23, 2026

ETRecord generation fails for mv3 with cortex_m.minimum tensor constants

Status: Open.

#18422 In pytorch/executorch;

· psiddh opened on Mar 23, 2026

[ARM] Ethos-U backend wheel or alternative versioned package (to avoid dev installs from source)

module: armIssues related to arm backendIssues related to arm backend

module: quantizationIssues related to quantizationIssues related to quantization

partner: armFor backend delegation, kernels, demo, etc. from the 3rd-party partner, ArmFor backend delegation, kernels, demo, etc. from the 3rd-party partner, Arm

Status: Open.

Feature

#18415 In pytorch/executorch;

· jonasdaugalas opened on Mar 23, 2026

·  

1.3.0

[QNN] InternVL3-1B fails to load on SA8295 (V68) due to insufficient HTP PD memory

module: qnnIssues related to Qualcomm's QNN delegate and code under backends/qualcomm/Issues related to Qualcomm's QNN delegate and code under backends/qualcomm/

Status: Open.

Bug

#18410 In pytorch/executorch;

· fgsiveone opened on Mar 23, 2026

Resolve smoke test timeouts on Windows

Status: Open.

#18381 In pytorch/executorch;

· GregoryComer opened on Mar 20, 2026

·  

1.2.0

Arm ToTosaMemoryFormatPass collapses duplicated output slots when equal placeholders are fused

partner: armFor backend delegation, kernels, demo, etc. from the 3rd-party partner, ArmFor backend delegation, kernels, demo, etc. from the 3rd-party partner, Arm

Status: Open.

#18320 In pytorch/executorch;

· Rob-Hughes-Arm opened on Mar 19, 2026

High inference time for Softmax decomposition for Ethos-U backend

partner: armFor backend delegation, kernels, demo, etc. from the 3rd-party partner, ArmFor backend delegation, kernels, demo, etc. from the 3rd-party partner, Arm

Status: Open.

#18319 In pytorch/executorch;

· eladc4 opened on Mar 19, 2026

Inefficient LayerNorm decomposition in Ethos-U quantization

partner: armFor backend delegation, kernels, demo, etc. from the 3rd-party partner, ArmFor backend delegation, kernels, demo, etc. from the 3rd-party partner, Arm

Status: Open.

#18316 In pytorch/executorch;

· eladc4 opened on Mar 19, 2026

XNNPACK export of Parakeet TDT model produces different output (mismatch vs NeMo / portable backend)

module: xnnpackIssues related to xnnpack delegation and the code under backends/xnnpack/Issues related to xnnpack delegation and the code under backends/xnnpack/

Status: Open.

#18281 In pytorch/executorch;

· sehun1kim-gleo opened on Mar 18, 2026

DINO v2 v3 support

Status: Open.

#18242 In pytorch/executorch;

· Gasoonjia opened on Mar 17, 2026

[ARM/TOSA] Dependency cycle after Q/DQ de-tagging in TOSAPartitioner breaks models with attention blocks

partner: armFor backend delegation, kernels, demo, etc. from the 3rd-party partner, ArmFor backend delegation, kernels, demo, etc. from the 3rd-party partner, Arm

Status: Open.

#18190 In pytorch/executorch;

· beomwookang opened on Mar 16, 2026

Building failed for Cadence target

module: cadenceIssues related to the Cadence/Xtensa backendIssues related to the Cadence/Xtensa backend

Status: Open.

#18181 In pytorch/executorch;

· hegel-wang opened on Mar 14, 2026