fix comments from code review.

src/quantization.cpp

@@ -18,8 +18,8 @@ inline namespace MIGRAPHX_INLINE_NS {
 
 instruction_ref convert_fp32_fp16(program& prog, instruction_ref& ins)
 {
-    assert(ins->get_shape().type() == shape::float_type);
-    assert(ins->name().front() == '@');
+    assert(ins->get_shape().type() == shape::float_type || ins->get_shape().type() == shape::double_type);
+    assert(contains({"@literal", "@param"}, ins->name()));
     instruction_ref ins_fp16{};
     if(ins->name() == "@literal")
     {
@@ -47,11 +47,15 @@ instruction_ref convert_fp32_fp16(program& prog, instruction_ref& ins)
 void quantize(program& prog)
 {
     bool reduced_precision = false;
+    shape::type_t orig_type = shape::float_type;
     for(auto ins : iterator_for(prog))
     {
         // convert float_type to half_type
-        if(ins->name().front() == '@' && ins->get_shape().type() == shape::float_type)
+        if(contains({"@literal", "@param"}, ins->name()) && 
+            (ins->get_shape().type() == shape::float_type ||
+            ins->get_shape().type() == shape::double_type))
         {
+            orig_type = ins->get_shape().type();
             auto ins_fp16 = convert_fp32_fp16(prog, ins);
             auto outputs  = ins->outputs();
             for(auto output : outputs)
@@ -71,7 +75,7 @@ void quantize(program& prog)
     {
         for(auto ins : iterator_for(prog))
         {
-            if(ins->name().front() != '@')
+            if(!contains({"@literal", "@param"}, ins->name()))
             {
                 ins->recompute_ins_shape();
             }
@@ -80,7 +84,7 @@ void quantize(program& prog)
         auto ins = std::prev(prog.end());
         if(ins->get_shape().type() == shape::half_type)
         {
-            prog.add_instruction(op::fp_conversion{shape::float_type}, ins);
+            prog.add_instruction(op::fp_conversion{orig_type}, ins);
         }
     }
 }