I am finding issues with preprocessing Criteo Terabyte Dataset for the inference runs for Intel CPU.
When the readme file talks about to prepare the DLRM dataset,
Create a directory (such as /data/mlperf_data/dlrm/) which contain:
day_fea_count.npz
terabyte_processed_test.bin
About how to get the dataset, please refer to
https://github.com/facebookresearch/dlrm
I tried to follow through the instructions in the https://github.com/facebookresearch/dlrm repo. After following through this command, I found issues.
./bench/dlrm_s_criteo_terabyte.sh ["--test-freq=10240 --memory-map --data-sub-sample-rate=0.875"]
What are the best commands to get the dataset preprocessed and get the inference runs started.
I am finding issues with preprocessing Criteo Terabyte Dataset for the inference runs for Intel CPU. When the readme file talks about to prepare the DLRM dataset,
I tried to follow through the instructions in the https://github.com/facebookresearch/dlrm repo. After following through this command, I found issues.
./bench/dlrm_s_criteo_terabyte.sh ["--test-freq=10240 --memory-map --data-sub-sample-rate=0.875"]
What are the best commands to get the dataset preprocessed and get the inference runs started.