Fine-turn pretrained model

I m using google colab pro for training arcface (mxnet) model with 'emore' dataset

I have tried several time for training because my resource have been disconnected every 20 hours.

also my model would not completely trained with this time


also every time while saving checkpoint it will be saved to the same file './models/r50-arcface-emore/model-0001.params'

if someone let me know how to save every check point to the different file or any other solution for return to the last status and iteration after reconnect to my resources.

Inspect the ArcFace MXNet training setup and its checkpoint configuration, starting with how the emore run writes model-0001.params. Verify the behavior across a reconnect or interrupted run; done means checkpoints are retained distinctly and training can resume from the latest saved iteration.

Written by the indexing model from the issue text.