[CMPL] Add Support for Other Data Types (#252)

* [CMPL] Add Support for Other Data Types

* [CMPL] Add test

* [CMPL] Fix

nnvm/include/nnvm/top/tensor.h

@@ -56,6 +56,7 @@ enum TypeFlag {
   kInt32 = 4,
   kInt8  = 5,
   kInt64 = 6,
+  kInt16 = 7,
 };
 
 struct CastParam : public dmlc::Parameter<CastParam> {
@@ -67,6 +68,9 @@ struct CastParam : public dmlc::Parameter<CastParam> {
     .add_enum("float16", kFloat16)
     .add_enum("uint8", kUint8)
     .add_enum("int32", kInt32)
+    .add_enum("int8", kInt8)
+    .add_enum("int64", kInt64)
+    .add_enum("int16", kInt16)
     .describe("Output data type.");
   }
 };

nnvm/python/nnvm/compiler/graph_attr.py

@@ -28,12 +28,26 @@ def set_shape_inputs(g, shape):
 
 DTYPE_TO_TCODE = {
     "default": -1,
-    "float32": 0
+    "float32": 0,
+    "float64": 1,
+    "float16": 2,
+    "uint8": 3,
+    "int32": 4,
+    "int8": 5,
+    "int64": 6,
+    "int16": 7,
 }
 
 TCODE_TO_DTYPE = {
     -1: None,
-    0: "float32"
+    0: "float32",
+    1: "float64",
+    2: "float16",
+    3: "uint8",
+    4: "int32",
+    5: "int8",
+    6: "int64",
+    7: "int16",
 }
 
 def set_dtype_inputs(g, dtype):

nnvm/src/compiler/compile_engine.cc

@@ -27,14 +27,39 @@ using namespace tvm;
  */
 int GetTypeFlag(tvm::Type type) {
   if (type == tvm::Float(32)) return 0;
+  if (type == tvm::Float(64)) return 1;
+  if (type == tvm::Float(16)) return 2;
+  if (type == tvm::UInt(8)) return 3;
+  if (type == tvm::Int(32)) return 4;
+  if (type == tvm::Int(8)) return 5;
+  if (type == tvm::Int(64)) return 6;
+  if (type == tvm::Int(16)) return 7;
   LOG(FATAL) << "cannot convert " << type;
   return 0;
 }
 // convert from type flag to tvm type.
 Type GetTVMType(int type_flag) {
-  if (type_flag == 0) return tvm::Float(32);
-  LOG(FATAL) << "unknown type_flag=" << type_flag;
-  return Float(32);
+  switch (type_flag) {
+    case 0:
+      return tvm::Float(32);
+    case 1:
+      return tvm::Float(64);
+    case 2:
+      return tvm::Float(16);
+    case 3:
+      return tvm::UInt(8);
+    case 4:
+      return tvm::Int(32);
+    case 5:
+      return tvm::Int(8);
+    case 6:
+      return tvm::Int(64);
+    case 7:
+      return tvm::Int(16);
+    default:
+      LOG(FATAL) << "unknown type_flag=" << type_flag;
+      return Float(32);
+  }
 }
 
 // internal compile engine

nnvm/src/compiler/compile_engine.h

@@ -90,6 +90,22 @@ GraphFunc GraphLower(Graph graph,
                      const Op* schedule_op_key,
                      const NodeAttrs& schedule_op_attr);
 
+/*!
+ * \brief Get type flag from TVM Type
+ *
+ * \param type the tvm type
+ * \return corresponding DLDataType
+ */
+int GetTypeFlag(tvm::Type type);
+
+/*!
+ * \brief Get TVM Type from type flag
+ *
+ * \param type_flag the type flag
+ * \return corresponding TVM type
+ */
+tvm::Type GetTVMType(int type_flag);
+
 }  // namespace compiler
 }  // namespace nnvm
 

nnvm/src/compiler/graph_fuse.cc

@@ -36,9 +36,7 @@ enum class FuseRule {
  * \return corresponding DLDataType
  */
 DLDataType GetDLType(int type_flag) {
-  if (type_flag == 0) return tvm::Type2TVMType(tvm::Float(32));
-  LOG(FATAL) << "unknown type_flag=" << type_flag;
-  return Type2TVMType(Float(32));
+  return Type2TVMType(GetTVMType(type_flag));
 }
 
 // Partition the graph into segments

nnvm/tests/python/compiler/test_build.py

@@ -77,7 +77,26 @@ def test_precompute_prune():
         res.asnumpy(), nx.asnumpy() + 1 + ny.asnumpy() + na.asnumpy())
 
 
+def test_dtypes():
+    x = sym.Variable("x")
+    y = sym.relu(x)
+    dshape = (1, 3, 32, 32)
+    oshape = dshape
+    for dtype in ['float32', 'float64', 'int32', 'int16', 'int8', 'int64']:
+        graph, lib, _ = nnvm.compiler.build(y, 'llvm', {"x": dshape}, dtype=dtype)
+        m = graph_runtime.create(graph, lib, tvm.cpu())
+        if 'float' in dtype:
+          data = np.random.uniform(size=dshape).astype(dtype)
+        elif 'int' in dtype:
+          data = np.random.randint(-127, 127, dshape).astype(dtype)
+        m.run(x=data)
+        data = (data > 0) * data
+        out = m.get_output(0, tvm.nd.empty(oshape, dtype))
+        np.testing.assert_allclose(out.asnumpy(), data, atol=1e-5, rtol=1e-5)
+
+
 if __name__ == "__main__":
     test_precompute_prune()
     test_compile()
     test_run()
+    test_dtypes()