- Auto device selection: Now supports CPU, GPU, and multi-GPU configurations.
- Enhanced LoRa/QLoRa Support: Added 8-bit and 4-bit support for LoRa and QLoRa devices.
- Bitsandbytes Integration: Added support for bitsandbytes data format.
- Hardware-Agnostic Quantized Models: Implemented GGML for hardware-agnostic quantized models.
- RAG Generation with Pinecone: Added support for RAG generation using Pinecone.
- ChatGPT Finetuning: Enabled finetuning capabilities for ChatGPT models.
- New Models: Added support for Open-Orca models: LlongOrca 7B & 13B, OpenOrca-Platypus2-13B and Mistral-7B-OpenOrca.
- ChatGPT Finetune Model Update: Updated ChatGPT finetune model to gpt-3.5-turbo-0613.
- Knowledge Distillation Update: Generate finetuning dataset using all supported models.
- MacOS and Linux support: Added bash script for setting up on MacOS and Linux.
- Refactored Completions: Code completions have been refactored for better performance.
- Class Name Standardization: Corrected class names to adhere to CamelCase conventions.
- Cleanup: Removed duplicated example files and ensured all necessary requirements are included.