Search results

SD3: add T5-XXL text encoder support

Status: Open.

#104 In apple/coreai-models;

· stikves opened on Jul 15, 2026

Proposal: Add configurable VLM image preprocessing strategy (stretch / center_crop)

Status: Open.

#100 In apple/coreai-models;

· tmorales2000 opened on Jul 14, 2026

Unit Tests for CoreAISpeech

Status: Open.

#98 In apple/coreai-models;

· kevchengcodes opened on Jul 13, 2026

PyPI wheel coreai-models==0.1.0 has incorrect requires_python>=3.14 (source says >=3.11)

Status: Open.

#96 In apple/coreai-models;

· tmorales2000 opened on Jul 11, 2026

Add a README for the benchmark tool

Status: Open.

#92 In apple/coreai-models;

· stikves opened on Jul 9, 2026

Polish current multiturn tests to catch regression

Status: Open.

#88 In apple/coreai-models;

· carinapeng opened on Jul 6, 2026

Monolithic stateful dynamic prefill becomes nondeterministic above 16 tokens

Status: Open.

#84 In apple/coreai-models;

· kylejfrost opened on Jul 4, 2026

Gemma4 E-series staged export should PAD-mask image-token rows for per-layer embedding inputs

Status: Open.

#83 In apple/coreai-models;

· kylejfrost opened on Jul 4, 2026

VLM: qwen3-vl exports CLIP normalization stats, but Qwen3-VL checkpoints specify image_mean = image_std = [0.5, 0.5, 0.5]

Status: Open.

#82 In apple/coreai-models;

· john-rocky opened on Jul 3, 2026

Flux sometimes fails spuriously before encodeText and then kills my test iPad entirely until restart.

bugSomething isn't workingSomething isn't working

Status: Open.

#77 In apple/coreai-models;

· Bersaelor opened on Jul 2, 2026

Composite RoPE partial-rotary mode pairs contiguously — incompatible with transformers partial/proportional rotary (partial_rotary_factor models)

Status: Open.

#66 In apple/coreai-models;

· kylejfrost opened on Jun 27, 2026

Model request: MiniCPM-V 4.6 (compact vision-language model) for macOS and iOS

Status: Open.

#59 In apple/coreai-models;

· ZMXJJ opened on Jun 20, 2026