I have downloaded and unzipped the 4GB dataset. It consists of 3 files, readme.txt, train.txt, and test.txt. It is stored in a folder called dataset, which is a sibling folder to my cloned dlrm folder.
I believe this is the command needed to preprocess, as implied in the README
world size: 1, current rank: 0, local rank: 0
Using CPU...
time/loss/accuracy (if enabled):
Finished training it 1/1 of epoch 0, -1.00 ms/it, loss 0.083850
I have seen another issue, #274 where someone posted lines that should happen when preprocessing occurs.
Reading raw data=/my_raw_data_path/train.txt
Additionally, there is no .npz file(s) in my input directory.
Is it because there are some other required flags?
I have downloaded and unzipped the 4GB dataset. It consists of 3 files, readme.txt, train.txt, and test.txt. It is stored in a folder called dataset, which is a sibling folder to my cloned dlrm folder.
I believe this is the command needed to preprocess, as implied in the README
python dlrm_s_pytorch.py --raw-data-file=../dataset/train.txt
however, the output of this is
world size: 1, current rank: 0, local rank: 0 Using CPU... time/loss/accuracy (if enabled): Finished training it 1/1 of epoch 0, -1.00 ms/it, loss 0.083850
I have seen another issue, #274 where someone posted lines that should happen when preprocessing occurs. Reading raw data=/my_raw_data_path/train.txt
Additionally, there is no .npz file(s) in my input directory.
Is it because there are some other required flags?