Polyaxon deployment best practicesDeploying Polyaxon can be hard, not only because because it requires using Kubernetes a tool that is not yet fully used by several teams, but also because it is a stateful application.
Productive iterations with templatesDeveloping machine learning models requires several iterations, Polyaxon users can create several templates to easily update the node scheduling or parameters.
Polyaxon uses Helm to install and manage a deployment on Kubernetes.
Discussion about autoscaling of preemptible GPU resourcesDiscussion about how to use a pool of preemptible GPUs that are automatically used whenever experiments or hyper parameters search is created in Polyaxon.
CartPole game by Reinforcement Learning, a journey from training to inferenceThis project is intended to play with CartPole game using Reinforcement Learning and to know how we may train a different model experiments with enough observability (metrics/monitoring). The model is divided basically in three parts: Neural network model, QLearning algorithm and application runner.
Polyaxon, Argo and Seldon for model training, package and deployment in KubernetesThe ultimate combination of open-source frameworks for model management in Kubernetes?
Open Source Model Management Roundup: Polyaxon, Argo, and SeldonExploring model management using open source tools to streamline the data science pipeline.
Using Polyaxon on GKE with an NFS serverDeploying Polyaxon on GKE with an NFS servers.
Using Language Models with Approximate Outputs to pre-train spaCy using PolyaxonThis repository contains experiments on spaCy's new pretrain command, which uses a ULMFit/Elmo/BERT/etc-like process for pre-training.
Training cifar10 on PolyaxonCIFAR-10 classification is a common benchmark problem in machine learning. The problem is to classify RGB 32x32 pixel images across 10 categories.