bert

Running the bert test on TensorFlow

This is currently blocked as a dependent python package needs to be built with a nightly version of the Rust compiler.

Ensure the virtual environment is prepared as described in MLPerf Inference

If on a machine with sufficient free storage space

Activate the virtual environment first and then

python3 -m pip install transformers==4.14.1
cd ~/src/inference/language/bert
make setup
python3 run.py

Or alternatively if you are on a machine in the Cambridge colo.

Mount the NFS drive with the already downloaded data Cambridge Colo data server

python3 -m pip install transformers==4.14.1
cd ~/src/inference/language/bert
make init_submodule
mkdir build
cp ../../mlperf.conf build
ln -s /mnt/datasets/data/bert build/data
python3 run.py

TensorFlow

bert

Analytics

Related content