Low-Rank Pruning of Llama2
mobiusml.github.io
mobiusml.github.io
In a nutshell, we've managed to reduce the model's parameter count by up to 50%, double the training speed, and increase inference speed by 1.25 times.
For those interested in the technical details or looking to replicate our results, the code is openly available for community use and contributions
Also, can y'all dumb it down for a simple end user like me? Is this actually distilling the model down to a smaller parameter count, or is it just reducing VRAM/compute during training and during inference with a lora? Or something else?
It is not distilling the model, it is reducing the model weights on the fly and uses LoRA for training/fine-tuning. After the training phase, we explain how to merge the LoRA weights with the pruned weights to achieve faster inference speed