/yolov8-determinedai

Proof of Concept on how to integrate YoloV8 with DeterminedAI

Primary LanguageJupyter NotebookMIT LicenseMIT

YoloV8 with DeterminedAI (WIP, Experimental)

Proof of Concept on how to integrate YoloV8(https://github.com/ultralytics/ultralytics) with DeterminedAI

Author: Andrew Mendez

Install Dependencies

bash install-dep.sh

Current Limitations

Currently the code only supports single node multi-gpu training. More work is needed to support multiple node. If you want to train with multiple gpus, only edit the: device config setting. You should edit where you specify all the cuda devices you want to use. Examples are follows:

  • Train on single GPU: device: 0
  • Train on 4 GPUs: device: 0,1,2,3
  • Train on 8 GPUs: device: 0,1,2,3,4,5,6,7

Spin up Determined Cluster (No GPU cluster)

det deploy local cluster-up --no-gpu --master-port 8081

Spin up Determined Cluster (GPU cluster)

det deploy local cluster-up --master-port 8081

Run Training if using M1 Mac (Arm64)

bash run_exp_arm64.sh

Run Training if using Linux/Mac (Intel-based, Amd64)

bash run_exp.sh

Steps to run Vanilla Training without Determined

python train.py

Questions

Any Questions, please reach out to me at andrew.mendez@hpe.com