first pass of gpu smoke test

scripts/test-config-cuda.sh

@@ -0,0 +1,72 @@
+#!/bin/bash
+set -e
+
+# require one command line argument
+if [ "$#" -ne 1 ]
+then
+  echo "Error: one argument for log location required (e.g. ./gpu-test.log)"
+  echo "Usage: $0 log-location"
+  exit 1
+fi
+
+# set log location from command invokation
+LOG_LOC=$1
+TEST_FAIL=false
+
+# driver
+PROC_DRIVER_FILE=/proc/driver/nvidia/version
+if [ ! -f "$PROC_DRIVER_FILE" ]
+then
+  echo "$PROC_DRIVER_FILE doesn't exist" | tee -a $LOG_LOC
+  echo "WARNING: CUDA driver may not be correctly installed." | tee -a $LOG_LOC
+  TEST_FAIL=true
+else
+  while read line; do
+    IFS=' ' read -ra tmp_array <<< $line
+    if [ ${tmp_array[0]} = "NVRM" ] && [ ${tmp_array[1]} = "version:" ]
+    then
+      VERSION_DRIVER=${tmp_array[7]}
+    fi
+  done < $PROC_DRIVER_FILE
+fi
+
+echo $VERSION_DRIVER | tee -a $LOG_LOC
+
+# toolkit
+if ! TOOLKIT_CHECK_OUTPUT=$(nvcc -V 2>&1);
+then
+  echo "Failed to run 'nvcc -V' with error message: $TOOLKIT_CHECK_OUTPUT" | tee -a $LOG_LOC
+  echo "WARNING: CUDA toolkit may not be correctly installed." | tee -a $LOG_LOC
+  TEST_FAIL=true
+else
+  # parse output to get version number
+  while IFS= read -r line
+  do
+    IFS=' ' read -ra tmp_array <<< $line
+    if [ "${tmp_array[3]}" = "release" ]
+    then
+      VERSION_TOOLKIT=${tmp_array[5]}
+    fi
+  done <<< $TOOLKIT_CHECK_OUTPUT
+fi
+
+echo $VERSION_TOOLKIT | tee -a $LOG_LOC
+
+# tensorflow
+if ! VERSION_TF_OUTPUT=$(python -c 'import tensorflow as tf; print(tf.__version__)' 2>&1);
+then
+  echo "Error: trying to get tensorflow version: $TF_VERSION"
+else
+  while IFS= read -r line
+  do
+    VERSION_TF=$line
+  done <<< $VERSION_TF_OUTPUT
+fi
+
+echo $VERSION_TF | tee -a $LOG_LOC
+
+if [ "$TEST_FAIL" = true ]
+then
+  echo "WARNING: at least one of the GPU functionality tests has failed." | tee -a $LOG_LOC
+  echo "Please run rocker-versioned2/tests/gpu/test-gpu.sh script for more detailed information." | tee -a $LOG_LOC
+fi

tests/gpu/misc/examples_tf.R

@@ -0,0 +1,36 @@
+
+## Tensorflow:
+install.packages('keras', repos='http://cran.us.r-project.org')
+library(keras)
+mnist <- dataset_mnist()
+x_train <- mnist$train$x
+y_train <- mnist$train$y
+x_test <- mnist$test$x
+y_test <- mnist$test$y
+# reshape
+x_train <- array_reshape(x_train, c(nrow(x_train), 784))
+x_test <- array_reshape(x_test, c(nrow(x_test), 784))
+# rescale
+x_train <- x_train / 255
+x_test <- x_test / 255
+y_train <- to_categorical(y_train, 10)
+y_test <- to_categorical(y_test, 10)
+model <- keras_model_sequential()
+model %>%
+  layer_dense(units = 256, activation = 'relu', input_shape = c(784)) %>%
+  layer_dropout(rate = 0.4) %>%
+  layer_dense(units = 128, activation = 'relu') %>%
+  layer_dropout(rate = 0.3) %>%
+  layer_dense(units = 10, activation = 'softmax')
+
+  model %>% compile(
+    loss = 'categorical_crossentropy',
+    optimizer = optimizer_rmsprop(),
+    metrics = c('accuracy')
+  )
+  history <- model %>% fit(
+    x_train, y_train,
+    epochs = 30, batch_size = 128,
+    validation_split = 0.2
+  )
+model %>% evaluate(x_test, y_test)

tests/gpu/misc/nvblas.R

@@ -0,0 +1 @@
+../../ml/nvblas.R

tests/ml/nvblas.R

@@ -0,0 +1,11 @@
+install.packages("callr")
+
+
+callr::r(function(){
+  system.time({
+  N <- 2^14
+  M <- matrix(rnorm(N*N), nrow=N, ncol=N)
+  M %*% M
+  })
+  }, env = c(LD_PRELOAD="libnvblas.so")
+)