Open clort81 opened 1 week ago
Are the labels on the solid blue and yellow lines swapped?
The dashed lines appear to show a lower loss for MatMul-Free LM, while the solid lines show a higher loss.
Hi,
actually the solid line is the scaling law, which is a fitted line with the real loss, and dash line is the real loss. it seems correct.
Are the labels on the solid blue and yellow lines swapped?
The dashed lines appear to show a lower loss for MatMul-Free LM, while the solid lines show a higher loss.