also have to modify the training config file that is within fairsec
so if we clone it, we lose those two
Read the paper - need to create a data processing thing (doing on the train)
10 is not how the paper specified it
K means clustering - X number of different chards
Those shards are similar - didn’t read the paper properly - it’s actually 100
Calvin to redo on 100
Currently working on pre-training step (first stage of pre-training step)
Replicated the paper configuration and currently cycling through
Deliverable being developed. Once it’s passed the first iterations, will be as smooth as possible for developers to get on and have their workflow replicated
Experiment ID: 03038eeb-7106-4a5e-95ec-fc91ae703751
Has checkpointing in-built