Crate torsh_models 

Pre-trained models and model zoo for ToRSh deep learning framework

This crate provides a comprehensive collection of pre-trained models and utilities for loading, using, and managing deep learning models in ToRSh.

Re-exports

pub use downloader::DownloadProgress;

pub use downloader::ModelDownloader;

pub use lazy_loading::CacheStats;

pub use lazy_loading::LazyModelLoader;

pub use lazy_loading::LazyTensor;

pub use lazy_loading::StreamingModelLoader;

pub use model_merging::LoRAMerger;

pub use model_merging::MergeStrategy;

pub use model_merging::ModelMerger;

pub use model_merging::ModelSoup;

pub use model_sharding::DevicePlacement;

pub use model_sharding::ModelSharder;

pub use model_sharding::ShardingStats;

pub use model_sharding::ShardingStrategy;

pub use registry::ModelHandle;

pub use registry::ModelInfo;

pub use registry::ModelRegistry;

pub use utils::convert_model_format;

pub use utils::convert_pytorch_state_dict;

pub use utils::convert_to_pytorch_state_dict;

pub use utils::load_model_from_file;

pub use utils::load_model_weights;

pub use utils::load_pytorch_checkpoint;

pub use utils::load_safetensors_weights;

pub use utils::load_state_dict;

pub use utils::map_parameter_names;

pub use utils::save_model_to_file;

pub use utils::save_pytorch_checkpoint;

pub use utils::save_tensors_to_safetensors;

pub use utils::ModelFormat;

pub use utils::ModelMetadata;

Modules

architectures

Advanced neural network architectures and building blocks

audio

Audio models for ToRSh deep learning framework

benchmark

Model benchmarking utilities for performance evaluation

builder

Model builders and factories for easy instantiation

common

Common components shared between vision and NLP models

comparison

Model comparison and analysis tools

config

Model configuration system for parameterizing architectures

distillation

Knowledge distillation utilities for model compression and transfer learning

domain

Specialized Domain Models

downloader

Model downloader for fetching pre-trained models

ensembling

Model ensembling utilities for combining multiple models

few_shot

Few-shot learning utilities and meta-learning algorithms

fine_tuning

Fine-tuning utilities for transfer learning and model adaptation

generative

Generative model implementations for ToRSh

gnn

Graph Neural Network implementations for ToRSh

lazy_loading

Lazy loading optimizations for efficient model loading

model_merging

Model merging and fusion utilities

model_sharding

Model sharding for distributed inference and training

multimodal

Multimodal models for ToRSh deep learning framework

nlp

NLP models organized by model family

optimization

Model optimization utilities for improving performance and efficiency

prelude

Prelude module for convenient imports

pruning

Model pruning utilities for reducing model size and improving efficiency

quantization

Model quantization utilities for reducing precision and model size

registry

Model registry for managing pre-trained models

rl

Reinforcement Learning Models

surgery

Model surgery utilities for architecture modification and composition

utils

Utility functions for model loading and saving

validation

Model validation and accuracy testing utilities

video

Video model implementations for ToRSh

vision

Vision models for ToRSh deep learning framework

vision_3d

3D Vision model implementations for ToRSh

Enums

ModelError

ModelType

Concrete model enum to avoid trait object issues

Type Aliases

ModelResult

Result type for model operations