AI MosaicML Proposes Modifying Chinchilla Scaling Laws to Account for Inference Costs when Determining Optimal LLM Size
