DeepSomatic fails on HPC with small /tmp

Hi all, Thanks for the tool, super excited to try it. I’m running DeepSomatic v1.9.0 via Singularity on an HPC cluster (SLURM), and I consistently encounter the following error early in the workflow: ``` Error in tempfile() using template /tmp/parXXXXX.df: Could not create temp file /tmp/par0xOHc.df: No space left on device at /usr/bin/parallel line 5414. ``` On our HPC system, the node-local /tmp filesystem is intentionally small (a common cluster configuration), so jobs fail immediately when parallel attempts to create temporary files under /tmp. <B>What I am trying to do</B> I would like to change the temporary directory used by DeepSomatic / GNU parallel to a large, writable filesystem rather than local /tmp. <B>What I have already tried (without success)</B> 1. Setting environment variables on the host and inside the container: a. TMPDIR b. PARALLEL_TMPDIR c. SINGULARITY_TMPDIR d. SINGULARITYENV_TMPDIR, e. SINGULARITYENV_PARALLEL_TMPDIR, etc. 2. Using --scratch and pointing it to a large, writable filesystem Despite these, the log output shows that parallel is invoked and creates files such as: ``` /tmp/parXXXXX.df ``` Command I am using ``` singularity run --scratch "${TMPDIR_HOST}" \ -B /usr/lib/locale/:/usr/lib/locale/ \ -B "${TMPDIR_HOST}:/usr/bin/parallel" \ -B "${TMPDIR_HOST}:${TMPDIR_HOST}" \ -B "${REF_DIR}:${REF_DIR}" \ -B "${TUMOR_DIR}:${TUMOR_DIR}" \ -B "${OUT_DIR}:${OUT_DIR}" \ -B "${TMPDIR_HOST}:/tmp" \ docker://google/deepsomatic:1.9.0 \ run_deepsomatic \ --model_type=FFPE_WGS_TUMOR_ONLY \ --ref=${REF} \ --reads_tumor=${BAM} \ --output_vcf=${OUT_DIR}/${SAMPLE}.deepsomatic.vcf.gz \ --sample_name_tumor=${SAMPLE} \ --num_shards=8 \ --logging_dir=${OUT_DIR}/logs/${SAMPLE} \ --use_default_pon_filtering=true \ --intermediate_results_dir=${OUT_DIR}/logs/${SAMPLE} ``` Is there a supported way to change the temporary directory in DeepSomatic? Best regards, Lipika