Merge branch 'emma/correct_comment_datapath' into 'main'

Update some comments and datapath

See merge request dl/hugectr/hugectr!1532

Author: EmmaQiaoCh

CMakeLists.txt

@@ -122,7 +122,7 @@ if (DISABLE_A2A_WARMUP)
 endif()
 
 option(DISABLE_CUDF "Disable cudf: disable parquet format related features" OFF)
-# this manual definition is a WAR and RMM team will fix it in the future
+# Add this manual definition
 add_compile_definitions(LIBCUDACXX_ENABLE_EXPERIMENTAL_MEMORY_RESOURCE)
 option(USE_CUDART_STATIC "Setup clangformat target" OFF)
 if(USE_CUDART_STATIC)

HugeCTR/src/hps/CMakeLists.txt

@@ -30,7 +30,7 @@ list(APPEND huge_ctr_hps_src
   "../io/gcs_filesystem.cpp"
 )
 
-# this manual definition is a WAR and RMM team will fix it in the future
+# Add this manual definition
 add_compile_definitions(LIBCUDACXX_ENABLE_EXPERIMENTAL_MEMORY_RESOURCE)
 
 add_library(huge_ctr_hps SHARED ${huge_ctr_hps_src})

sbin/build-hadoop.sh

@@ -43,7 +43,6 @@ if [[ "${HDFS_BUILD_MODE}" == "MINIMAL" ]]; then
   cd hadoop-hdfs-project/hadoop-hdfs-native-client
 fi
 
-# WAR for https://issues.apache.org/jira/browse/YARN-11528
 # TODO: remove below lines when the fix released.
 sed -i '/scripts/i \
     \"resolutions\": { \

test/utest/data_reader/data_reader_benchmark.cu

@@ -34,7 +34,6 @@ using namespace HugeCTR;
 
 // const std::string fname = "/40m.limit_preshuffled/train_data.bin";
 // const std::string fname = "/criteo_kaggle/dlrm/test_data.bin";
-// const std::string fname = "/raid/datasets/criteo/mlperf/40m.limit_preshuffled/train_data.bin";
 const std::string fname = "/data_multihot/val_data.bin";
 
 __global__ void gpu_sleep(int64_t num_cycles) {

test/utest/embedding_collection/configuration.hpp

@@ -374,10 +374,9 @@ InputDataConfiguration synthetic_uniform_dataset = {.fixed_hotness = true,
 InputDataConfiguration criteo_dataset = {
     .fixed_hotness = true,
     .input_data_type = InputDataType::RawFormat,
-    .raw_format_param = {
-        .input_file = "/raid/datasets/criteo/mlperf/40m.limit_preshuffled/train_data.bin",
-        .label_dim = 1,
-        .dense_dim = 13}};
+    .raw_format_param = {.input_file = "/datasets/40m.limit_preshuffled/train_data.bin",
+                         .label_dim = 1,
+                         .dense_dim = 13}};
 
 InputDataConfiguration criteo_multi_hot_dataset = {
     .fixed_hotness = true,