Scribd
Upload
15 views3 pages

Useful Commands

The document provides a series of command-line instructions for using Torch and Deepspeed for training models, as well as Docker commands for managing images and containers. It includes steps for AWS ECR login, pulling and running Docker images, managing GPU resources, and installing the AWS CLI. Additionally, it covers Alembic commands for database migrations and Git commands for version control operations.

Uploaded by

faraz jamal
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
15 views3 pages

Useful Commands

The document provides a series of command-line instructions for using Torch and Deepspeed for training models, as well as Docker commands for managing images and containers. It includes steps for AWS ECR login, pulling and running Docker images, managing GPU resources, and installing the AWS CLI. Additionally, it covers Alembic commands for database migrations and Git commands for version control operations.

Uploaded by

faraz jamal
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 3

Torch run console script

torchrun --nproc_per_node=NUM_GPUS_YOU_HAVE train.py --arg1 value1 --arg2 value2

torchrun --nproc_per_node=8 train.py \

--training_dir /home/ec2-user/SageMaker/9.1_EXP_fairscale/fairscale/train \

--test_dir /home/ec2-user/SageMaker/9.1_EXP_fairscale/fairscale/test \

--model_uri /home/ec2-user/SageMaker/9.1_EXP_fairscale/fairscale/model_uri \

--rank 0 \

--world_size 8

Deepspeed console script


deepspeed src/train.py --num_gpus=8 --deepspeed_config=deepspeed-config.json

deepspeed --num_gpus 8 training_1/src/train_bash.py \

--deepspeed examples/deepspeed/ds_z2_config.json \

--stage sft \

ECR login

aws ecr get-login --no-include-email --region "us-east-1" | awk '{print $6}' | docker login -u AWS --
password-stdin "381492148851.dkr.ecr.us-east-1.amazonaws.com"

Docker pull

docker pull [OPTIONS] IMAGE[:TAG|@DIGEST]

docker pull 381492148851.dkr.ecr.us-east-1.amazonaws.com/sagemaker-training-containers/


isbank:llamafactory

Docker run

docker run [OPTIONS] IMAGE[:TAG|@DIGEST] [COMMAND] [ARG...]

docker run -it --name training_2 381492148851.dkr.ecr.us-east-1.amazonaws.com/sagemaker-


training-containers/isbank:llamafactory

Docker tag

docker tag SOURCE_IMAGE[:TAG] TARGET_IMAGE[:TAG]

docker tag cd91f33067a8 381492148851.dkr.ecr.us-east-1.amazonaws.com/sagemaker-training-


containers/isbank:llamafactory_2.0
Docker push

docker push username/my-image:latest

docker push 381492148851.dkr.ecr.us-east-1.amazonaws.com/sagemaker-training-containers/


isbank:llamafactory_2.0

Docker rmi

docker rmi [OPTIONS] IMAGE [IMAGE...]

docker rmi 381492148851.dkr.ecr.us-east-1.amazonaws.com/sagemaker-training-containers/


isbank:llamafactory

Docker interactive
docker exec -it <container ID> /bin/bash

mount volume for docker

docker run -v /home/ec2-user/SageMaker:/opt/ml -it 381492148851.dkr.ecr.us-east-


1.amazonaws.com/sagemaker-training-containers/isbank:llamafactory_2.0

Docker run all gpus

docker run --gpus all -v /home/ec2-user/SageMaker:/opt/ml -it 381492148851.dkr.ecr.us-east-


1.amazonaws.com/sagemaker-training-containers/isbank:llamafactory_2.0

Docker build

docker-compose build
docker-compose up

Monitor nvidia GPU devices


nvidia-smi

Install aws-cli

sudo apt install unzip


curl "https://awscli.amazonaws.com/awscli-exe-linux-x86_64.zip" -o "awscliv2.zip"
unzip awscliv2.zip
./aws/install --bin-dir ~/bin --install-dir ~/aws-cli –update

Download model s3 to local instance

/root.bin/aws s3 sync <source> <destination>

/root.bin/aws s3 sync s3://ish-bank-data/NewModel/Retriever/ /opt/ml/input/data/model


Cache cleanup

for cache "cd ~/.cache/"

rm -Rf ~/.cache/

Alembic

alembic heads

alembic revision -m ‘message’ ==> generates new migration/revision and assigns the metadata

alembic history ==> see complete log of applied migrations to the database.

alembic current ==> see the current head/revision

alembic upgrade head ==> to migrate the database to the head revision

alembic downgrade -1 ==> moves the database to previous revision

Git Commands

git fetch origin

git checkout main

git pull origin main

git checkout mybranch

git rebase main

alembic history

git push -u origin mybranch –force

git fetch –all

git config --global user.email " "

git config --global user.name " "

git add .

You might also like