_h -> _fp16

Summary: _h -> _fp16 for better verbosity

Differential Revision: D54401567

Author: Michael Gschwind

examples/models/llama2/export_llama_lib.py

@@ -424,7 +424,7 @@ def _export_llama(modelname, args) -> str:  # noqa: C901
         generate_memory_trace(builder.export_program, "memory_profile.json")
 
     if builder.dtype == DType.fp16:
-        modelname = f"{modelname}_h"
+        modelname = f"{modelname}_fp16"
 
     if args.use_kv_cache:
         modelname = f"{modelname}_kv"