mergekit
mergekit copied to clipboard
Evolutionary Merging out of memory
When attempting to merge three 14B LLMs on a custom task using the mergekit-evolve method, I ran into memory overflow issues on 8 A100 GPUs, each with 80G of memory. Could you please let me know if 4-bit quantized model merging is supported, or if there are any effective solutions to address this memory overflow issue?