intel/ipex-llm - Accelerate local LLM inference and finetuning (LLaMA, Mistral, ChatGLM, Qwen, DeepSeek, Mixtral, Gemma, Phi, MiniCPM, Qwen-VL, MiniCPM-V, etc.) on Intel XPU (e.g., local PC with iGPU and NPU, discrete GPU such as Arc, Flex and Max); seamlessly integrate with llama.cpp, Ollama, HuggingFace, LangChain, LlamaIndex, vLLM, DeepSpeed, Axolotl, etc.

Yao Zhang 6548dcd66e Support layerwise, weight-bias separated and multiple sorts (L1, L2 and others) of regularization (#714) ... * make linear layerwise regurization work * add regularizer option to layers with weights fix a bug * new version * include accRegularization in accGradparameters * meet code review * add layerwise regualrization performance tests * fix to meet code review * make Regularizer.isRegularized private and optimize L1 regularization implementation * fix jenkins tests failed * change l1 implementation * improve layerwise regularization performance test * improve layerwise regularization performance test * delete layerwise regularization perform test		2017-05-08 16:26:52 +08:00