installation problems - package inconsistencies et al

[rtalasilawfb]

Hi,

Thank you for putting together the code on github.

However I encountered some installation problems. Here are a couple:

1. There are inconsistencies in the package requirements. The relso_env.yml file needs to be edited as the packages with specific labels are not available on conda. For example: "python=3.10.4=hdfd78df_0".
2. the torch version in relso_env.yml specifies 1.12.0, but the requirements.txt requires "torch=1.7.1".
3. The "pip install -e ." gives an error " error: Multiple top-level packages discovered in a flat-layout: ['data', 'relso']."

I am trying the install on a fresh computing instance on GCP (4x T4 GPU, 30 GB RAM, 500 GB disk). Wanted to check if the installation & setup was successfully completed on a clean machine before.

[rtalasilawfb]

Nope. I couldn’t install it.

On Jun 19, 2023, at 2:49 PM, luis-sribeiro @.***> wrote:

Hi,

Thank you for putting together the code on github.

However I encountered some installation problems. Here are a couple:

There are inconsistencies in the package requirements. The relso_env.yml file needs to be edited as the packages with specific labels are not available on conda. For example: "python=3.10.4=hdfd78df_0". the torch version in relso_env.yml specifies 1.12.0, but the requirements.txt requires "torch=1.7.1". The "pip install -e ." gives an error " error: Multiple top-level packages discovered in a flat-layout: ['data', 'relso']." I am trying the install on a fresh computing instance on GCP (4x T4 GPU, 30 GB RAM, 500 GB disk). Wanted to check if the installation & setup was successfully completed on a clean machine before.

Hi, were you able to install it?

— Reply to this email directly, view it on GitHub https://github.com/KrishnaswamyLab/ReLSO-Guided-Generative-Protein-Design-using-Regularized-Transformers/issues/3#issuecomment-1597798709, or unsubscribe https://github.com/notifications/unsubscribe-auth/AZJV4H6R5DWOX6CRJC3XEETXMDCOTANCNFSM6AAAAAAVPYSYHA. You are receiving this because you authored the thread.

[ec1340]

Hi thanks for raising this issue and for your patience!

Hopefully this issue is addressed with the latest merge #4 . The updated installation instructions can be found in the README.md.

Reposted here,

# make conda environment
conda create --name relsoenv python=3.9
conda activate relsoenv

# install pytorch
# GPU (linux)
pip3 install torch torchvision torchaudio

# CPU only (linux)
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu

# CPU only (mac)
pip3 install torch torchvision torchaudio

# install other dependencies
python -m pip install networkx pytorch-lightning==1.9 wandb scikit-learn pandas matplotlib gdown phate

# install relso
pip install -e .   

[rtalasilawfb]

The install went ahead without a problem this time.

There is a small missing step before step 4 in the install instructions. Before doing ‘pip install -e .’, one needs to do ‘cd ReLSO-Guided-Generative-Protein-Design-using-Regularized-Transformers’. This is not a big issue as folks can easily figure out the need to position in the correct directory before running pip.

However I found the below problems while doing the training on GPU:

"python train_relso.py --data gifford —n_gpus=4” gave the following error

File "/home/ubuntu/miniconda3/envs/relsoenv/lib/python3.9/site-packages/torch/nn/functional.py", line 2416, in _verify_batch_size raise ValueError("Expected more than 1 value per channel when training, got input size {}".format(size)) ValueError: Expected more than 1 value per channel when training, got input size torch.Size([1, 64])

I then tried running on CPU and I got the following error:

training complete!

now beginning evaluations...

train sequences raw shape: torch.Size([57603, 20]) valid sequences raw shape: torch.Size([10166, 20]) test sequences raw shape: torch.Size([22690, 20]) —— —

model has fitness predictions shape of train outputs: torch.Size([57603, 22, 20]), torch.Size([57603, 1]) Traceback (most recent call last): File "/home/ubuntu/data/ReLSO-Guided-Generative-Protein-Design-using-Regularized-Transformers/train_relso.py", line 268, in seqd_list = [data.train_split_seqd, data.valid_split_seqd, data.test_split_seqd] AttributeError: 'EnsGradData' object has no attribute 'train_split_seqd'(base)

On Jul 5, 2023, at 5:30 PM, Egbert Castro @.***> wrote:

Hi thanks for raising this issue and for your patience!

Hopefully this issue is addressed with the latest merge #4 https://github.com/KrishnaswamyLab/ReLSO-Guided-Generative-Protein-Design-using-Regularized-Transformers/pull/4 . The updated installation instructions can be found in the README.md.

Reposted here,

make conda environment

conda create --name relsoenv python=3.9 conda activate relsoenv

install pytorch

GPU (linux)

pip3 install torch torchvision torchaudio

CPU only (linux)

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu

CPU only (mac)

pip3 install torch torchvision torchaudio

install other dependencies

python -m pip install networkx pytorch-lightning==1.9 wandb scikit-learn pandas matplotlib gdown phate

install relso

pip install -e .
— Reply to this email directly, view it on GitHub https://github.com/KrishnaswamyLab/ReLSO-Guided-Generative-Protein-Design-using-Regularized-Transformers/issues/3#issuecomment-1622732843, or unsubscribe https://github.com/notifications/unsubscribe-auth/AZJV4HZEQB44N76K7G64LYLXOYBJPANCNFSM6AAAAAAVPYSYHA. You are receiving this because you authored the thread.