itwinai 0.2
No due date
94% complete
- Pipeline representation for use cases
- Integrated use cases:
- MNIST (TF, torch, lightning)
- CERN (TF, lightning)
- TorchDistributedTrainer (data-parallel, multi-GPU)
- TensorFlowDistributedTrainer (data-parallel, multi-GPU or multi-worker)
- Containerized environment:
- Run on HPC with SLURM
- (optional) Auto-build images via GH actions, for each use case
- Verti…
- Pipeline representation for use cases
- Integrated use cases:
- MNIST (TF, torch, lightning)
- CERN (TF, lightning)
- TorchDistributedTrainer (data-parallel, multi-GPU)
- TensorFlowDistributedTrainer (data-parallel, multi-GPU or multi-worker)
- Containerized environment:
- Run on HPC with SLURM
- (optional) Auto-build images via GH actions, for each use case
- Vertical integration with infrastructure:
- CERN use case (lightning) distributed on multi-GPU, executed from a containerized environment
- Execute on K8s + SLURM (WP5's interLink)
- Simulate new data available and prove integration with OSCAR