Estimations for memory and compute requirements for training large language models.
No changes between version 1 and version 1
