Berzelius Container Modules

Berzelius Container Modules are lightweight wrappers that make it possible to transparently use Apptainer containers as environment modules.

It has the following benefits:

• Use familiar environment module commands, ensuring a minimal learning curve or change to existing workflows

• Leverage all the benefits of containers, including portability and reproducibility

Please use container modules on compute nodes where you have access to apptainer.

We save all containers at /software/sse/containers on Berzelius.

We currently have

• AlphaFold
• PyTorch
• TensorFlow

AlphaFold

Loading the Module

module load AlphaFold/2.3.1-apptainer

Setting Paths

We have a copy of AlphaFold database on Berzelius at /proj/nsc/users/alphafold_db_2.3.1 which should only be used for test purposes. Please download your own copy for your research.

We set paths for the database and the results folder.

ALPHAFOLD_DB=/proj/nsc/users/alphafold_db_2.3.1
ALPHAFOLD_RESULTS=/proj/nsc_testing/xuan/alphafold_results_2.3.1

Downloading genetic databases

We dowonload the database. It will take ~2.6 T hard disk space.

bash scripts/download_all_data.sh ${ALPHAFOLD_DB}

Running AlphaFold jobs

To make the best utilisations of the GPU resources on Berzelius, we strongly suggest to separate the CPU and GPU parts when running AlphaFold jobs. You should run the CPU part job on Tetralith or on your local computer, and then run the GPU part job on Berzelius.

bash run_alphafold.sh \
  -d ${ALPHAFOLD_DB} \
  -o ${ALPHAFOLD_RESULTS}/output \
  -f ${ALPHAFOLD_RESULTS}/input/T1050.fasta \
  -t 2021-11-01 \
  -g true \
  -P 3 \
  -F false

PyTorch

Loading the Module

module load PyTorch/2.0.0-apptainer

Running Python scripts/commands

The container will be trigered by python ... commands.

For example, we can check the available GPU by:

python -c "import torch; print('GPU Name: ' + torch.cuda.get_device_name(0))"

Running an MNIST example

module load PyTorch/2.0.0-apptainer
cd /tmp
git clone https://github.com/pytorch/examples
cd examples/mnist
python main.py

Results would be

...
Train Epoch: 14 [56960/60000 (95%)] Loss: 0.019554
Train Epoch: 14 [57600/60000 (96%)] Loss: 0.001885
Train Epoch: 14 [58240/60000 (97%)] Loss: 0.011090
Train Epoch: 14 [58880/60000 (98%)] Loss: 0.002160
Train Epoch: 14 [59520/60000 (99%)] Loss: 0.001996

Test set: Average loss: 0.0274, Accuracy: 9915/10000 (99%)

Check if some package(s) installed

python -c "import os; os.system('mamba list numpy')"

TensorFlow

Loading the Module

module load TensorFlow/2.11.1-apptainer

Running Python scripts/commands

The container will be trigered by python ... commands.

For example, we can check the GPU info by:

python -c "import tensorflow as tf; print(tf.test.gpu_device_name())"

Running an MNIST example

module load TensorFlow/2.11.1-apptainer
cd /tmp
wget -O tensorflow_mnist_example.py "https://liuonline-my.sharepoint.com/:u:/g/personal/xuagu37_liu_se/EV416fjpNHVBujn25mbDguYBzxWlBCTdMaluxgesyiCDKw?e=8VH53p&download=1"
python tensorflow_mnist_example.py

Results would be

...
Epoch 1/6
469/469 [==============================] - 7s 2ms/step - loss: 0.3538 - sparse_categorical_accuracy: 0.9017 - val_loss: 0.2000 - val_sparse_categorical_accuracy: 0.9411
Epoch 2/6
469/469 [==============================] - 1s 1ms/step - loss: 0.1641 - sparse_categorical_accuracy: 0.9529 - val_loss: 0.1384 - val_sparse_categorical_accuracy: 0.9603
Epoch 3/6
469/469 [==============================] - 1s 1ms/step - loss: 0.1181 - sparse_categorical_accuracy: 0.9664 - val_loss: 0.1127 - val_sparse_categorical_accuracy: 0.9668
Epoch 4/6
469/469 [==============================] - 1s 1ms/step - loss: 0.0925 - sparse_categorical_accuracy: 0.9733 - val_loss: 0.0969 - val_sparse_categorical_accuracy: 0.9706
Epoch 5/6
469/469 [==============================] - 1s 1ms/step - loss: 0.0747 - sparse_categorical_accuracy: 0.9780 - val_loss: 0.0886 - val_sparse_categorical_accuracy: 0.9725
Epoch 6/6
469/469 [==============================] - 1s 1ms/step - loss: 0.0622 - sparse_categorical_accuracy: 0.9818 - val_loss: 0.0874 - val_sparse_categorical_accuracy: 0.9737

Check if some package(s) installed

python -c "import os; os.system('mamba list numpy')"