JH's repositories
Language:HTML000
Retriever-Reader-Model
Retriever-Reader Models in natural language processing combine a fast, efficient Retriever for narrowing down relevant texts from large datasets with sophisticated Reader models like BERT or GPT for deep text analysis and response generation.
Language:Jupyter NotebookMIT000
Language:Jupyter NotebookApache-2.0000
flexible-LLM-sharding
This repository is designed for individuals who want to run LLM locally with small vRAM (GPU memory) and RAM. It facilitates the efficient execution of unquantized LLM's (such as Llama2-70B) with vRAM>=6GB and RAM>=8GB, i.e. when standard offloading does not work. Of course, larger the vRAM, faster the model runs.