Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Update bionemo test case + propose to subdirectories per orchastrator #396

Draft
wants to merge 5 commits into
base: main
Choose a base branch
from

Conversation

KeitaW
Copy link
Collaborator

@KeitaW KeitaW commented Aug 5, 2024

Issue #, if available:

Description of changes:

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

@KeitaW KeitaW self-assigned this Aug 5, 2024
@KeitaW KeitaW added the documentation Improvements or additions to documentation label Aug 5, 2024
RUN git clone -b ${NCCL_VERSION} https://github.com/NVIDIA/nccl.git /opt/nccl \
&& cd /opt/nccl \
&& make -j $(nproc) src.build CUDA_HOME=/usr/local/cuda \
NVCC_GENCODE="-gencode=arch=compute_80,code=sm_80 -gencode=arch=compute_86,code=sm_86 -gencode=arch=compute_89,code=sm_89 -gencode=arch=compute_90,code=sm_90"
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Perhaps add a link or quick description on what each arch is. Not required but good to have.

export DATASET_PATH=/fsx/
```

## 1.4. Pull this github repo
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
## 1.4. Pull this github repo
## 1.4. Clone this github repo

```bash
cd /apps/
git clone https://github.com/aws-samples/awsome-distributed-training.git
cp -r /apps/awsome-distributed-training/3.test_cases/14.bionemo/* ./apps/
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why?

```
# Miniconda is already installed if you are using the DLAMI but needs installation with Base AMI

wget -O miniconda.sh "https://repo.anaconda.com/miniconda/${MINICONDA_INSTALLER}.sh" \
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Python virtual env is enough. No need to introduce conda here.

`3.test_cases/14.nemo-multimodal/0.Dockerfile` and we can build a image like below:

```
docker build -t ${DOCKER_IMAGE_NAME}:${TAG} -f 0.Dockerfile .
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Define variables.

We will use the popular [UniRef50](https://www.uniprot.org/help/uniref) dataset for pretraining. We will use BioNemo's in-built functionality to download and pre-process data. To this end, we provide `prepare_uniref50.py` file to do so. You can edit the above to download and process [UniRef90]((https://www.uniprot.org/help/uniref)). To run the above python code on your slurm cluster in the BioNemo cluster execute the following:

```bash
sbatch 1.uniref50.slurm
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

rename the script
uniref50_download_preprocess.sbatch

Once the above image is pulled, you can run the container on the head node like below. This step could be used for troubleshooting purposes. Here we are running the container just to be able to copy launcher scripts on the host machine. If you need to run the container on the compute nodes, you would need to add `--gpus all` flag to the run command. It is recommended to have the docker run flags like below, as recommended by Nvidia PyTorch containers, otherwise you may potentially run into an error like [this](https://github.com/NVIDIA/Megatron-LM/issues/516)

```
docker run -it nvcr.io/nvidia/clara/bionemo-framework:latest bash
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

pin version.

Copy link
Contributor

@mhuguesaws mhuguesaws left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

left comments

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
documentation Improvements or additions to documentation
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants