Skip to content

Changelog

All notable changes to MindNLP are documented here.

Version 0.6.x (Current)

MindSpore: >=2.7.1 | Python: 3.10-3.11

Highlights

  • Full HuggingFace Transformers compatibility via patching mechanism
  • Full HuggingFace Diffusers compatibility
  • Support for latest model architectures (Qwen3, Llama3, etc.)
  • Enhanced mindtorch layer for PyTorch API compatibility
  • Improved device management and heterogeneous computing support

New Features

  • Automatic patching of transformers and diffusers libraries
  • Support for ms_dtype parameter in model loading
  • Enhanced device_map support for multi-device inference
  • Improved tensor serialization and checkpoint handling

Version 0.5.x

MindSpore: 2.5.0-2.7.0 | Python: 3.10-3.11

Highlights

  • Major API refactoring for better HuggingFace compatibility
  • Introduction of mindtorch compatibility layer
  • Support for new model families (Gemma, Phi-3, etc.)

New Features

  • mindnlp.core module providing PyTorch-compatible APIs
  • Enhanced AutoModel classes for various tasks
  • Improved tokenizer support
  • PEFT/LoRA integration for parameter-efficient fine-tuning

Version 0.4.x

MindSpore: 2.2.x-2.5.0 | Python: 3.9-3.11

Highlights

  • Expanded model support
  • Improved training stability
  • Enhanced Trainer API

New Features

  • Support for Qwen2, Mistral, Mixtral models
  • Enhanced gradient checkpointing
  • Improved distributed training support
  • Better memory management for large models

Version 0.3.x

MindSpore: 2.1.0-2.3.1 | Python: 3.8-3.9

Highlights

  • Stable release with comprehensive model coverage
  • Improved documentation and examples

New Features

  • Support for Llama, Llama2 models
  • ChatGLM series support (ChatGLM, ChatGLM2, ChatGLM3)
  • Enhanced dataset loading utilities
  • Improved model serialization

Version 0.2.x

MindSpore: >=2.1.0 | Python: 3.8-3.9

Highlights

  • Major architecture improvements
  • Better alignment with HuggingFace APIs

New Features

  • Refactored model architecture
  • Improved tokenizer implementations
  • Enhanced training engine
  • Better error messages and debugging

Version 0.1.x

MindSpore: 1.8.1-2.0.0 | Python: 3.7.5-3.9

Highlights

  • Initial release of MindNLP
  • Core transformer model support

New Features

  • Basic transformer models (BERT, GPT-2, T5, etc.)
  • Tokenizer support
  • Dataset loading utilities
  • Basic training loop implementation

For detailed release notes, see GitHub Releases.