Fix several typos (#1169)

Co-authored-by: isaacw <isaacw@nvidia.com>

examples/44_multi_gemm_ir_and_codegen/ir_gen/gen_device.py

@@ -49,7 +49,7 @@ class gen_device:
         self.arg_member = []
         self.gen_class_name = gen_class_name
         self.gen_kernel_name = gen_class_name + "Kernel"
-        self.tempalte_args = []
+        self.template_args = []
         self.__tempalate_arg_list = {'Stages': int, 'SplitKSerial': bool, 'IsBetaZero': bool, 'AlignmentA': int, 'AlignmentB': int}
 
         self.file_name = output_dir + "/device/" +gen_class_name +".h"
@@ -63,7 +63,7 @@ class gen_device:
         self.first_use_1stage = False
 
         ## gen kernel
-        self.gen_kernel = gen_ker.gen_kernel(self.tempalte_args, self.gen_class_name, self.b2b_num, output_dir, cutlass_deps_root, project_root)
+        self.gen_kernel = gen_ker.gen_kernel(self.template_args, self.gen_class_name, self.b2b_num, output_dir, cutlass_deps_root, project_root)
 
 
     def __check_arg_type(self, temp_arg):
@@ -126,7 +126,7 @@ class gen_device:
         func_code = self.gen_all_func()
         member_var_code = "private:\n typename B2bGemmKernel::Params params_;\n"
 
-        gen_code = gen_ir.gen_template_class(self.gen_class_name, self.tempalte_args, func_code + member_var_code)
+        gen_code = gen_ir.gen_template_class(self.gen_class_name, self.template_args, func_code + member_var_code)
         code = self.gen_include_header() + gen_ir.gen_namespace("cutlass", gen_ir.gen_namespace("gemm", gen_ir.gen_namespace("device", gen_code)))
 
         if ifprint:
@@ -142,7 +142,7 @@ class gen_device:
 
     def update_b2b_class_template_args(self):
         for arg in self.args.keys():
-            self.tempalte_args.append([self.__check_arg_type(arg), arg, self.args[arg]])
+            self.template_args.append([self.__check_arg_type(arg), arg, self.args[arg]])
 
     def update_b2b_args(self):
 

examples/44_multi_gemm_ir_and_codegen/ir_gen/gen_kernel.py

@@ -444,7 +444,7 @@ class gen_kernel:
 
         self.gen_class_name = "B2bGemm"
         self.gen_kernel_name = gen_class_name + "Kernel"
-        self.tempalte_args = []
+        self.template_args = []
 
         self.cutlass_deps_root = cutlass_deps_root
         self.project_root = project_root

examples/44_multi_gemm_ir_and_codegen/ir_gen/gen_threadblock.py

@@ -957,13 +957,13 @@ public:\n\
         
     def gen_code(self):
 
-        tempalte_arg = []
+        template_arg = []
         for i in range(self.b2b_num):
-            tempalte_arg.append(("typename", helper.var_idx("Shape", i)))
+            template_arg.append(("typename", helper.var_idx("Shape", i)))
         for i in range(self.b2b_num):
-            tempalte_arg.append(("typename", helper.var_idx("Policy", i)))
+            template_arg.append(("typename", helper.var_idx("Policy", i)))
         for i in range(self.b2b_num):
-            tempalte_arg.append((int, helper.var_idx("Stage", i)))
+            template_arg.append((int, helper.var_idx("Stage", i)))
         
      
 
@@ -971,7 +971,7 @@ public:\n\
         code_body += self.gen_protected()
         code_body += self.gen_public_member()
 
-        class_code = gen_ir.gen_template_class("B2bMmaBase", tempalte_arg, code_body)
+        class_code = gen_ir.gen_template_class("B2bMmaBase", template_arg, code_body)
 
         code = self.gen_include_header() + gen_ir.gen_namespace("cutlass", gen_ir.gen_namespace("gemm", gen_ir.gen_namespace("threadblock", class_code)))
 

examples/python/04_epilogue_visitor.ipynb

@@ -68,7 +68,7 @@
    "source": [
     "## Define the epilogue visitor functor\n",
     "The epilogue functor can be defined as a simple Python function and a set of example tensors for inputs and outputs. The example below illustrates a complex epilogue under the directed acyclic graph structure (`F` is used twice). The epilogue takes source tensors in different ranks: `alpha`, `beta` are scalars, `bias` is a column vector to broadcast, and `C`, `aux` are matrices. It contains various math operations from basic arithmatic operations and built-in callable functions like `relu`. It also accomodates multiple outputs `D` and `F`. Note that there are some restrictions on syntax.\n",
-    "* Each named variable must be assigned exactly once and defined before it it used.\n",
+    "* Each named variable must be assigned exactly once and defined before it used.\n",
     "* Reserved names: `accum`, `C`, and `D` are reserved for accumulator, tensor_C, and tensor_D.\n",
     "* Return values must be a named variable.\n",
     "\n",