The scratch filesystem is currently not mounted on the login nodes.
It is, however, available from the compute nodes, in two ways.

1. Through an interactive session on a compute node

srun -A <project> -p genoa --pty bash -i

This opens a shell on a compute node, from which the scratch filesystem is accessible.
Type exit to release the resources.

2. In your job script, by placing the data copy at the beginning of the file:

#!/bin/bash
#SBATCH -A <project>
#SBATCH -p genoa
#SBATCH -J my_job
#SBATCH -N 1
#SBATCH -n 1
#SBATCH -c 16
#SBATCH -t 12:00:00
#SBATCH -o %x-%j.out

# 1. Create a dedicated working directory on scratch
WORKDIR=/scratch/$USER/$SLURM_JOB_ID
mkdir -p "$WORKDIR"

# 2. Copy input data to scratch
cp -r "$SLURM_SUBMIT_DIR"/inputs/* "$WORKDIR"/
cd "$WORKDIR"

# 3. Run the computation
srun my_program input.dat > output.log

# 4. Copy results back to the submission directory
cp -r "$WORKDIR"/* "$SLURM_SUBMIT_DIR"/results/

Remember to copy your results back at the end of the script (step 4): as long as they remain on scratch,
they cannot be viewed from the login nodes.

Some executables copied from Rheticus fail at startup with a missing shared library error, for example:

lmp_auto: error while loading shared libraries: libmpi_cxx.so.40: cannot open shared object file: No such file or directory

Answer coming soon.