[RMP] Refine Multi-GPU Data Parallel training for Tensorflow in Merlin Models

[bschifferer]

Problem:

In (536)[https://github.com//issues/536], we provided functionality of horovod with Merlin Models and added features which automates the process on the Merlin Models side. However, the current feature is not 100% user friendly and there are still open questions how a user can use multi-GPU data parallel training.

Goal:

• Improve the user experience to user multi-GPU data parallel training
• Test multi-GPU data parallel training: AUC? Scale-Up Performance?

Constraints:

• I am not sure, if the issues with unequal batch size for the data loader is solved: [BUG] Data parallel training freezes due to different number of batches dataloader#75:
  -- if the solution is about how the data is generated, correctly - how das that work?
  -- How are we ensure it with NVTabular?
  -- How about users who do NOT use NVTabular?
• The unittest is written that each worker runs through the FULL dataset per epoch. That is incorrect. If we have 1M data points and 2 GPUs, each GPU should run only through 500k data points. I wrote the example that NVTabular produces distinct files per worker. Is that the proposed workflow for a user?

Starting Point:

• Analyze scaling factor by using multiple GPUs: If we go from 1x GPU -> 2x GPUs -> 4x GPUs -> 8 GPUx - how much higher is the throughput?
• Provide performance metrics (accuracy / AUC / etc) to show that there is no negative effect in the model performance
• Provide guidance how to set global batch size, batch size per GPU and learning rate when scaling

[bschifferer]

@EvenOldridge @viswa-nvidia - as we discussed - I created a follow up roadmap ticket for multi-data parallel GPU feature