Fine-tuning LLMs with PEFT and LoRA
0
As LLMs get larger fine-tuning to the full extent can become difficult to train on consumer hardware. Storing and deploying these tuned models can also be quite expensive and difficult to store. With PEFT (parameter -efficent fine tuning), it approaches fine-tune on a smaller scale of model parameters while freezing most parameters of the pretrained LLMs. Basically it is providing full performance that which is similar if not better than full fine tuning while only having a small number of trainable parameters. This source explains that as well as going over LORA diagrams and a code walk through.
GPU Computing Workshop Series for the Earth Science Community
0
GPU training series for scientists, software engineers, and students, with emphasis on Earth science applications.
The content of this course is coordinated with the 6 month series of GPU Training sessions starting in Februrary 2022. The NVIDIA High Performance Computing Software Development Kit (NVHPC SDK) and CUDA Toolkit will be the primary software requirements for this training which will be already available on NCAR's HPC clusters as modules you may load. This software is free to download from NVIDIA by navigating to the NVHPC SDK Current Release Downloads page and the CUDA Toolkit downloads page. Any provided code is written specifically to build and run on NCAR's Casper HPC system but may be adapted to other systems or personal machines. Material will be updated as appropriate for the future deployment of NCAR's Derecho cluster and as technology progresses.