Skip to content

Instantly share code, notes, and snippets.

Show Gist options
  • Save vardaan123/6028f0a5ea7b591870018aa81ea72fd6 to your computer and use it in GitHub Desktop.
Save vardaan123/6028f0a5ea7b591870018aa81ea72fd6 to your computer and use it in GitHub Desktop.
Nvidia docker creation steps
docker pull nvcr.io/nvidia/pytorch:18.01-py3
nvidia-docker run -it --name sai nvcr.io/nvidia/pytorch:18.01-py3 /bin/bash

(Just try to import torch in python shell here). Exit with Ctrl+D.

docker commit -m test -a sai pytorch_sai nvcr.io/mila1234/pytorch_sai:1
docker push nvcr.io/mila1234/pytorch_sai:1
@vardaan123
Copy link
Author

sinter -w mila01 -c 1 --mem=10000 --qos=low
source ~lisa/NGCCLI/bin/activate
ngc dataset list
ngc dataset upload --ace nv-us-west-2 --desc "mnist is awesome" --source c:\dataset\mnist.zip mnist
ngc dataset remove [-h] <dataset id>
ngc result download <job_id>
ngc batch exec <job_id>

docker start $USER_CONTAINER_NAME
docker attach $USER_CONTAINER_NAME

docker commit -m test -a AUTHOR $USER_pCONTAINER_NAME nvcr.io/mila1234/$USER_CONTAINER_NAME:1
docker push nvcr.io/mila1234/$USER_CONTAINER_NAME:1

ngc batch run -n "Cloud-nv-us-west-2-752479" -i "nvidia/pytorch:18.01-py3" --ace nv-us-west-2 -in ngcv1 -c "python /mnt/code/main.py --data_dir /mnt/VQA_2/ --save_dir /results/save/ --save_prefix toy --fp16" --result /results --datasetid 8270:/mnt/code --datasetid 8231:/mnt/VQA_1 --datasetid 8232:/mnt/embed --datasetid 8230:/mnt/VQA_2

@vardaan123
Copy link
Author

ngc batch run -n "Cloud-nv-us-west-2-633840" -i "nvidia/pytorch:18.01-py3" --ace nv-us-west-2 -in ngcv8 -c "python /mnt/code/main.py --train_img --train_CNN --train_img_dir /mnt/VQA_1/train2014/ --val_img_dir /mnt/VQA_1/val2014/ --data_dir /mnt/data/ --show_batch_res --batch_size 16" --result /results --datasetid 8493:/mnt/code --datasetid 8478:/mnt/data --datasetid 8322:/mnt/VQA_1

@vardaan123
Copy link
Author

ngc batch run -n "Cloud-nv-us-west-2-633840" -i "nvidia/pytorch:18.01-py3" --ace nv-us-west-2 -in ngcv8 -c "python /mnt/code/main.py --train_img --train_CNN --train_img_dir /mnt/VQA_1/train2014/ --val_img_dir /mnt/VQA_1/val2014/ --data_dir /mnt/data/ --show_batch_res --batch_size 600 --cnn_type resnext-101-modular --cuda --pretrained_model_file /mnt/pretrain/resnext_101_32x4d.pth" --result /results --datasetid 8556:/mnt/code --datasetid 8478:/mnt/data --datasetid 8322:/mnt/VQA_1 --datasetid 8558:/mnt/pretrain

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment