Acceptable range for grad_norm_t plot

[znavidi]

Hi, Could you pls share your insight about what would be the acceptable range for grad_norm_t plot? I am fine-tuning your model and the (overall) accuracy and loss improve normally, and the range of grad_norm_t falls in the range of 0.1 - 0.5 on average. Does this fall within a normal or expected range during fine-tuning?

Would appreciate your guidance.

[znavidi]

Would appreciate any insight about this.