Losses and checkpoints

[borgr]

In your blog you bring many runs and iterations and scaling laws. Could you share the training losses and parameters from which those graphs are made for further research and analysis on them? Checkpoints may also be helpful for many I presume.