Search results

Channel-reduce logsumexp on Apple Neural Engine has a hard fp16 overflow at x ≈ 7.63 (output drops to 0)

bugUnexpected behaviour that should be corrected (type)Unexpected behaviour that should be corrected (type)

Status: Open.

#2690 In apple/coremltools;

· ChinChangYang opened on May 5, 2026

Softplus on Apple Neural Engine has a hard fp16 discontinuity at x ≈ 10.4 (output drops to 0)

bugUnexpected behaviour that should be corrected (type)Unexpected behaviour that should be corrected (type)

Status: Open.

#2687 In apple/coremltools;

· ChinChangYang opened on May 3, 2026

MLModelBenchmarker.benchmark_predict fails for models with ImageType inputs (unsupported feature type error)

bugUnexpected behaviour that should be corrected (type)Unexpected behaviour that should be corrected (type)

Status: Open.

#2678 In apple/coremltools;

· sisaman opened on Apr 30, 2026

InstanceNorm3d crashes on CoreML GPU path but works on CPU path

bugUnexpected behaviour that should be corrected (type)Unexpected behaviour that should be corrected (type)

Status: Open.

#2667 In apple/coremltools;

· 0xShug0 opened on Apr 2, 2026

CoreML CPU InstanceNorm3d behaves differently depending on track_running_stats

bugUnexpected behaviour that should be corrected (type)Unexpected behaviour that should be corrected (type)

PyTorch (traced)

Status: Open.

#2666 In apple/coremltools;

· 0xShug0 opened on Apr 2, 2026

Support Python 3.14

feature requestFunctionality does not currently exist, would need to be created as a new feature (type)Functionality does not currently exist, would need to be created as a new feature (type)

Status: Open.

#2646 In apple/coremltools;

· cdeil opened on Jan 26, 2026

CoreMLTools PyTorch converter fails on rotary-attention einsum (BeatThis model) with rank/perm mismatch

bugUnexpected behaviour that should be corrected (type)Unexpected behaviour that should be corrected (type)

PyTorch (traced)

Status: Open.

#2644 In apple/coremltools;

· tillt opened on Jan 18, 2026

MIL pow op for 2.0 to 0.0 is 0.0

bugUnexpected behaviour that should be corrected (type)Unexpected behaviour that should be corrected (type)

triagedReviewed and examined, release as been assigned if applicable (status)Reviewed and examined, release as been assigned if applicable (status)

Status: Open.

#2628 In apple/coremltools;

· kentslaney opened on Nov 27, 2025

Large output errors when converting MobileNetV3 in FP16

bugUnexpected behaviour that should be corrected (type)Unexpected behaviour that should be corrected (type)

Status: Open.

#2625 In apple/coremltools;

· johan-sightic opened on Nov 21, 2025

CoreML Conversion hangs with minimal pipeline when 16-bit cast pass is specified

bugUnexpected behaviour that should be corrected (type)Unexpected behaviour that should be corrected (type)

PyTorch (traced)

triagedReviewed and examined, release as been assigned if applicable (status)Reviewed and examined, release as been assigned if applicable (status)

Status: Open.

#2621 In apple/coremltools;

· jgibson2 opened on Nov 11, 2025

Difference between CoreML and Pytorch inference

questionResponse providing clarification needed. Will not be assigned to a release. (type)Response providing clarification needed. Will not be assigned to a release. (type)

Status: Open.

#2620 In apple/coremltools;

· onurtore opened on Nov 9, 2025

List acceptence for scale parameter

questionResponse providing clarification needed. Will not be assigned to a release. (type)Response providing clarification needed. Will not be assigned to a release. (type)

Status: Open.

#2619 In apple/coremltools;

· onurtore opened on Nov 8, 2025