fixes to scripts

rapidsai · Jan 8, 2024 · 89ac530 · 89ac530
1 parent 61f30a2
commit 89ac530
Show file tree

Hide file tree

Showing 3 changed files with 2 additions and 15 deletions.
diff --git a/benchmarks/cugraph/standalone/bulk_sampling/README.md b/benchmarks/cugraph/standalone/bulk_sampling/README.md
@@ -143,7 +143,7 @@ You will need to modify the bash scripts to run appopriately for your environmen
 desired training workflow.  The standard sbatch arguments are at the top of the script, such as
 job name, queue, etc.  These will need to be modified for your SLURM cluster.
 
-Next are arguments for the container image (which is currently set to the current DLFW image),
+Next are arguments for the container image (required),
 and directories where the data and outputs are stored.  The directories default to subdirectories
 of the current working directory.  But if there is a high-throughput storage system available,
 using that storage for the samples and datasets is highly recommended.

diff --git a/benchmarks/cugraph/standalone/bulk_sampling/run_sampling.sh b/benchmarks/cugraph/standalone/bulk_sampling/run_sampling.sh
@@ -36,8 +36,6 @@ export CUDF_SPILL=1
 export LIBCUDF_CUFILE_POLICY="OFF"
 export GPUS_PER_NODE=8
 
-PATCH_CUGRAPH=1
-
 export SCHEDULER_FILE=$SCHEDULER_FILE
 export LOGS_DIR=$LOGS_DIR
 
@@ -60,17 +58,6 @@ else
     ${MG_UTILS_DIR}/run-dask-process.sh workers &
 fi
 
-if [[ $PATCH_CUGRAPH == 1 ]]; then
-    mkdir /opt/cugraph-patch
-    git clone https://github.com/alexbarghi-nv/cugraph -b dlfw-patch-24.01 /opt/cugraph-patch
-
-    rm /opt/rapids/cugraph/python/cugraph/cugraph/structure/graph_implementation/simpleDistributedGraph.py
-    cp /opt/cugraph-patch/python/cugraph/cugraph/structure/graph_implementation/simpleDistributedGraph.py /opt/rapids/cugraph/python/cugraph/cugraph/structure/graph_implementation/simpleDistributedGraph.py
-    rm /usr/local/lib/python3.10/dist-packages/cugraph/structure/graph_implementation/simpleDistributedGraph.py
-    cp /opt/cugraph-patch/python/cugraph/cugraph/structure/graph_implementation/simpleDistributedGraph.py /usr/local/lib/python3.10/dist-packages/cugraph/structure/graph_implementation/simpleDistributedGraph.py
-
-fi
-
 echo "properly waiting for workers to connect"
 NUM_GPUS=$(python -c "import os; print(int(os.environ['SLURM_JOB_NUM_NODES'])*int(os.environ['GPUS_PER_NODE']))")
 handleTimeout 120 python ${MG_UTILS_DIR}/wait_for_workers.py \

diff --git a/benchmarks/cugraph/standalone/bulk_sampling/run_train_job.sh b/benchmarks/cugraph/standalone/bulk_sampling/run_train_job.sh
@@ -18,7 +18,7 @@
 #SBATCH -N 1
 #SBATCH -t 00:25:00 
 
-CONTAINER_IMAGE="/lustre/fsw/rapids/abarghi/dlfw_patched.squash"
+CONTAINER_IMAGE=${CONTAINER_IMAGE:="please_specify_container"}
 SCRIPTS_DIR=$(pwd)
 LOGS_DIR=${LOGS_DIR:=$(pwd)"/logs"}
 SAMPLES_DIR=${SAMPLES_DIR:=$(pwd)/samples}