Crate candle_coreml

Re-exports

pub use builder::CoreMLModelBuilder;

pub use config::Config;

pub use model::CoreMLModel;

pub use qwen::QwenConfig;

pub use qwen::QwenModel;

pub use state::CoreMLState;

pub use model_downloader::download_model;

pub use model_downloader::download_model_to;

pub use model_downloader::ensure_model_downloaded;

pub use model_downloader::get_cached_model_path;

pub use clean_git_lfs_downloader::download_hf_model_clean;

pub use clean_git_lfs_downloader::verify_download_completeness;

pub use clean_git_lfs_downloader::CleanDownloadConfig;

pub use utils::mask;

pub use utils::multi_component;

pub use utils::sampling;

Modules

builder

CoreML model builder for convenient model loading

clean_git_lfs_downloader

Clean Git2 + HF Hub LFS Downloader

config

Configuration types for CoreML models

conversion

Tensor conversion utilities for CoreML integration

model

Core CoreML model implementation

model_downloader

Unified Model Downloader Utility

qwen

Qwen model integration for candle-coreml

state

CoreML state management for autoregressive inference

utils

Shared utilities for transformer models and multi-component architectures

Functions

get_local_or_remote_file

Helper function to get a file locally first, then download from HuggingFace Hub if needed. Follows the same pattern as quantized-t5 example.