Add support for int8 values to Categorify inference (#1818)

Author: oliverholworthy

cpp/nvtabular/inference/categorify.cc

@@ -93,6 +93,9 @@ namespace nvtabular
           case 'u':
             switch (dtype.itemsize())
             {
+            case 1:
+              insert_int_mapping<uint8_t>(values);
+              return;
             case 2:
               insert_int_mapping<uint16_t>(values);
               return;
@@ -107,6 +110,9 @@ namespace nvtabular
           case 'i':
             switch (dtype.itemsize())
             {
+            case 1:
+              insert_int_mapping<int8_t>(values);
+              return;
             case 2:
               insert_int_mapping<int16_t>(values);
               return;
@@ -204,6 +210,8 @@ namespace nvtabular
           case 'u':
             switch (itemsize)
             {
+            case 1:
+              return transform_int<uint8_t>(input);
             case 2:
               return transform_int<uint16_t>(input);
             case 4:
@@ -215,6 +223,8 @@ namespace nvtabular
           case 'i':
             switch (itemsize)
             {
+            case 1:
+              return transform_int<int8_t>(input);
             case 2:
               return transform_int<int16_t>(input);
             case 4:

tests/unit/ops/test_categorify.py

@@ -704,9 +704,11 @@ def test_categorify_inference():
         "unicode_string": np.random.randint(
             low=a_char, high=z_char, size=num_rows * 10, dtype="int32"
         ).view("U10"),
+        "int8_feature": np.random.randint(0, 10, dtype="int8", size=num_rows),
         "int16_feature": np.random.randint(0, 10, dtype="int16", size=num_rows),
         "int32_feature": np.random.randint(0, 10, dtype="int32", size=num_rows),
         "int64_feature": np.random.randint(0, 10, dtype="int64", size=num_rows),
+        "uint8_feature": np.random.randint(0, 10, dtype="uint8", size=num_rows),
         "uint16_feature": np.random.randint(0, 10, dtype="uint16", size=num_rows),
         "uint32_feature": np.random.randint(0, 10, dtype="uint32", size=num_rows),
         "uint64_feature": np.random.randint(0, 10, dtype="uint64", size=num_rows),