diff --git a/CMakeLists.txt b/CMakeLists.txt
index c52f3b363a8a..4eb2468e4e2f 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -334,7 +334,7 @@ tvm_file_glob(GLOB RUNTIME_SRCS
   src/runtime/memory/*.cc
   src/runtime/disco/*.cc
   src/runtime/minrpc/*.cc
-  src/runtime/relax_vm/*.cc
+  src/runtime/vm/*.cc
 )
 set(TVM_RUNTIME_EXT_OBJS "")
 
diff --git a/cmake/modules/CUDA.cmake b/cmake/modules/CUDA.cmake
index 84261c6ea0ae..64f41d65fae4 100644
--- a/cmake/modules/CUDA.cmake
+++ b/cmake/modules/CUDA.cmake
@@ -138,8 +138,8 @@ if(USE_CUDA)
   endif(USE_NVTX)
 
   # Add CUDA builtins to RelaxVM
-  tvm_file_glob(GLOB RELAX_VM_CUDA_BUILTIN_SRC_CC src/runtime/relax_vm/cuda/*.cc)
-  list(APPEND RUNTIME_SRCS ${RELAX_VM_CUDA_BUILTIN_SRC_CC})
+  tvm_file_glob(GLOB VM_CUDA_BUILTIN_SRC_CC src/runtime/vm/cuda/*.cc)
+  list(APPEND RUNTIME_SRCS ${VM_CUDA_BUILTIN_SRC_CC})
 else(USE_CUDA)
   list(APPEND COMPILER_SRCS src/target/opt/build_cuda_off.cc)
 endif(USE_CUDA)
diff --git a/cmake/modules/Hexagon.cmake b/cmake/modules/Hexagon.cmake
index 75b0094ed611..2c153bbac96c 100644
--- a/cmake/modules/Hexagon.cmake
+++ b/cmake/modules/Hexagon.cmake
@@ -126,8 +126,8 @@ if(BUILD_FOR_HEXAGON)
     "${TVMRT_SOURCE_DIR}/hexagon/*.cc"
   )
   # Add builtins to RelaxVM
-  tvm_file_glob(GLOB RELAX_VM_BUILTIN_SRC_CC src/runtime/relax_vm/hexagon/*.cc)
-  list(APPEND RUNTIME_SRCS ${RELAX_VM_BUILTIN_SRC_CC})
+  tvm_file_glob(GLOB VM_BUILTIN_SRC_CC src/runtime/vm/hexagon/*.cc)
+  list(APPEND RUNTIME_SRCS ${VM_BUILTIN_SRC_CC})
 else()
   file_glob_append(RUNTIME_HEXAGON_SRCS
     "${TVMRT_SOURCE_DIR}/hexagon/hexagon_module.cc"
diff --git a/docs/reference/api/python/index.rst b/docs/reference/api/python/index.rst
index 2e0a9c7f2fb9..a233c69a0173 100644
--- a/docs/reference/api/python/index.rst
+++ b/docs/reference/api/python/index.rst
@@ -35,7 +35,7 @@ Python API
 
     runtime/runtime
     runtime/ndarray
-    runtime/relax_vm
+    runtime/vm
     runtime/disco
     runtime/profiling
 
diff --git a/docs/reference/api/python/runtime/relax_vm.rst b/docs/reference/api/python/runtime/vm.rst
similarity index 93%
rename from docs/reference/api/python/runtime/relax_vm.rst
rename to docs/reference/api/python/runtime/vm.rst
index 75afcb7939ab..2d7b980f864f 100644
--- a/docs/reference/api/python/runtime/relax_vm.rst
+++ b/docs/reference/api/python/runtime/vm.rst
@@ -15,7 +15,7 @@
     specific language governing permissions and limitations
     under the License.
 
-tvm.runtime.relax_vm
+tvm.runtime.vm
 --------------------
-.. automodule:: tvm.runtime.relax_vm
+.. automodule:: tvm.runtime.vm
     :members:
diff --git a/include/tvm/relax/exec_builder.h b/include/tvm/relax/exec_builder.h
index 81d6d4eb379e..16223d6bfb80 100644
--- a/include/tvm/relax/exec_builder.h
+++ b/include/tvm/relax/exec_builder.h
@@ -28,8 +28,8 @@
 #include <tvm/node/reflection.h>
 #include <tvm/node/repr_printer.h>
 #include <tvm/runtime/object.h>
-#include <tvm/runtime/relax_vm/bytecode.h>
-#include <tvm/runtime/relax_vm/executable.h>
+#include <tvm/runtime/vm/bytecode.h>
+#include <tvm/runtime/vm/executable.h>
 
 #include <string>
 #include <unordered_map>
@@ -38,7 +38,7 @@
 namespace tvm {
 namespace relax {
 
-namespace vm = tvm::runtime::relax_vm;
+namespace vm = tvm::runtime::vm;
 
 class ExecBuilder;
 
diff --git a/include/tvm/runtime/relax_vm/builtin.h b/include/tvm/runtime/vm/builtin.h
similarity index 92%
rename from include/tvm/runtime/relax_vm/builtin.h
rename to include/tvm/runtime/vm/builtin.h
index b994e44ae88d..3c4429b12250 100644
--- a/include/tvm/runtime/relax_vm/builtin.h
+++ b/include/tvm/runtime/vm/builtin.h
@@ -18,15 +18,15 @@
  */
 
 /*!
- * \file tvm/runtime/relax_vm/builtin.h
+ * \file tvm/runtime/vm/builtin.h
  * \brief Builtin runtime APIs.
  */
-#ifndef TVM_RUNTIME_RELAX_VM_BUILTIN_H_
-#define TVM_RUNTIME_RELAX_VM_BUILTIN_H_
+#ifndef TVM_RUNTIME_VM_BUILTIN_H_
+#define TVM_RUNTIME_VM_BUILTIN_H_
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 /*!
  * \brief Op code used in built-in match-shape function.
@@ -83,7 +83,7 @@ enum class MakeShapeCode : int {
   kLoadShape = 1,
 };
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
-#endif  // TVM_RUNTIME_RELAX_VM_BUILTIN_H_
+#endif  // TVM_RUNTIME_VM_BUILTIN_H_
diff --git a/include/tvm/runtime/relax_vm/bytecode.h b/include/tvm/runtime/vm/bytecode.h
similarity index 97%
rename from include/tvm/runtime/relax_vm/bytecode.h
rename to include/tvm/runtime/vm/bytecode.h
index 4526c6fffa1d..bc84578fd5d5 100644
--- a/include/tvm/runtime/relax_vm/bytecode.h
+++ b/include/tvm/runtime/vm/bytecode.h
@@ -18,11 +18,11 @@
  */
 
 /*!
- * \file tvm/runtime/relax_vm/bytecode.h
+ * \file tvm/runtime/vm/bytecode.h
  * \brief The bytecode for the virtual machine.
  */
-#ifndef TVM_RUNTIME_RELAX_VM_BYTECODE_H_
-#define TVM_RUNTIME_RELAX_VM_BYTECODE_H_
+#ifndef TVM_RUNTIME_VM_BYTECODE_H_
+#define TVM_RUNTIME_VM_BYTECODE_H_
 
 #include <tvm/runtime/data_type.h>
 #include <tvm/runtime/logging.h>
@@ -32,7 +32,7 @@
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 /*!
  * \brief The storage type for the bytecode in the VM.
@@ -238,8 +238,8 @@ struct Instruction {
   static Instruction If(RegName cond, Index false_offset);
 };
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
 
-#endif  // TVM_RUNTIME_RELAX_VM_BYTECODE_H_
+#endif  // TVM_RUNTIME_VM_BYTECODE_H_
diff --git a/include/tvm/runtime/relax_vm/executable.h b/include/tvm/runtime/vm/executable.h
similarity index 95%
rename from include/tvm/runtime/relax_vm/executable.h
rename to include/tvm/runtime/vm/executable.h
index 8a9fa024cab2..a84c902b6711 100644
--- a/include/tvm/runtime/relax_vm/executable.h
+++ b/include/tvm/runtime/vm/executable.h
@@ -18,10 +18,10 @@
  */
 
 /*!
- * \file tvm/runtime/relax_vm/executable.h
+ * \file tvm/runtime/vm/executable.h
  */
-#ifndef TVM_RUNTIME_RELAX_VM_EXECUTABLE_H_
-#define TVM_RUNTIME_RELAX_VM_EXECUTABLE_H_
+#ifndef TVM_RUNTIME_VM_EXECUTABLE_H_
+#define TVM_RUNTIME_VM_EXECUTABLE_H_
 
 #include <tvm/ffi/function.h>
 #include <tvm/runtime/module.h>
@@ -37,11 +37,11 @@
 // NOTE: this file only changes if we change relax vm format
 // for example if relax vm format do not change in 0.15, this should remain as 0.14
 // if it changes in 0.16, we will change it to 0.16
-#define RELAX_VM_VERSION "0.14"
+#define VM_VERSION "0.14"
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 /*!
  * \brief Information entry in executable function table.
@@ -214,11 +214,11 @@ class VMExecutable : public runtime::ModuleNode {
   void LoadPackedFuncNames(dmlc::Stream* strm);
 };
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
 
 namespace dmlc {
-DMLC_DECLARE_TRAITS(has_saveload, ::tvm::runtime::relax_vm::VMFuncInfo, true);
+DMLC_DECLARE_TRAITS(has_saveload, ::tvm::runtime::vm::VMFuncInfo, true);
 }  // namespace dmlc
-#endif  // TVM_RUNTIME_RELAX_VM_EXECUTABLE_H_
+#endif  // TVM_RUNTIME_VM_EXECUTABLE_H_
diff --git a/include/tvm/runtime/relax_vm/ndarray_cache_support.h b/include/tvm/runtime/vm/ndarray_cache_support.h
similarity index 93%
rename from include/tvm/runtime/relax_vm/ndarray_cache_support.h
rename to include/tvm/runtime/vm/ndarray_cache_support.h
index 579fbf306f68..3ab08df04389 100644
--- a/include/tvm/runtime/relax_vm/ndarray_cache_support.h
+++ b/include/tvm/runtime/vm/ndarray_cache_support.h
@@ -16,8 +16,8 @@
  * specific language governing permissions and limitations
  * under the License.
  */
-#ifndef TVM_RUNTIME_RELAX_VM_NDARRAY_CACHE_SUPPORT_H_
-#define TVM_RUNTIME_RELAX_VM_NDARRAY_CACHE_SUPPORT_H_
+#ifndef TVM_RUNTIME_VM_NDARRAY_CACHE_SUPPORT_H_
+#define TVM_RUNTIME_VM_NDARRAY_CACHE_SUPPORT_H_
 
 #include <tvm/ffi/container/array.h>
 #include <tvm/ffi/function.h>
@@ -29,7 +29,7 @@
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 /*!
  * \brief Metadata for NDArray cache, which by default, is named as "ndarray-cache.json".
@@ -89,8 +89,8 @@ struct NDArrayCacheMetadata {
   static NDArrayCacheMetadata LoadFromStr(const std::string& json_str, const std::string& path);
 };
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
 
-#endif  // TVM_RUNTIME_RELAX_VM_NDARRAY_CACHE_SUPPORT_H_
+#endif  // TVM_RUNTIME_VM_NDARRAY_CACHE_SUPPORT_H_
diff --git a/include/tvm/runtime/relax_vm/vm.h b/include/tvm/runtime/vm/vm.h
similarity index 96%
rename from include/tvm/runtime/relax_vm/vm.h
rename to include/tvm/runtime/vm/vm.h
index 884a8d0f4375..9aa34c9b4468 100644
--- a/include/tvm/runtime/relax_vm/vm.h
+++ b/include/tvm/runtime/vm/vm.h
@@ -18,13 +18,13 @@
  */
 
 /*!
- * \file tvm/runtime/relax_vm/vm.h
+ * \file tvm/runtime/vm/vm.h
  */
-#ifndef TVM_RUNTIME_RELAX_VM_VM_H_
-#define TVM_RUNTIME_RELAX_VM_VM_H_
+#ifndef TVM_RUNTIME_VM_VM_H_
+#define TVM_RUNTIME_VM_VM_H_
 
-#ifndef TVM_RELAX_VM_ENABLE_PROFILER
-#define TVM_RELAX_VM_ENABLE_PROFILER 1
+#ifndef TVM_VM_ENABLE_PROFILER
+#define TVM_VM_ENABLE_PROFILER 1
 #endif
 
 #include <memory>
@@ -45,7 +45,7 @@ using memory::MemoryManager;
 using memory::Storage;
 using memory::StorageObj;
 
-namespace relax_vm {
+namespace vm {
 
 /*!
  * \brief Possible instrument actions.
@@ -227,8 +227,8 @@ class VirtualMachine : public runtime::ModuleNode {
   std::unordered_map<uint32_t, VMExtension> extensions;
 };
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
 
-#endif  // TVM_RUNTIME_RELAX_VM_VM_H_
+#endif  // TVM_RUNTIME_VM_VM_H_
diff --git a/python/tvm/contrib/hexagon/session.py b/python/tvm/contrib/hexagon/session.py
index 4f017abcf6ba..b92c78309482 100644
--- a/python/tvm/contrib/hexagon/session.py
+++ b/python/tvm/contrib/hexagon/session.py
@@ -218,7 +218,7 @@ def get_executor_from_factory(
             The hexagon arch to be used
         """
         if isinstance(module, (runtime.Executable, str)):
-            return self._relax_vm_executable_executor(module, hexagon_arch=hexagon_arch)
+            return self._vm_executable_executor(module, hexagon_arch=hexagon_arch)
 
         raise TypeError(f"Unsupported executor type: {type(module)}")
 
@@ -244,7 +244,7 @@ def _set_device_type(self, module: Union[str, pathlib.Path]):
             else:
                 self._requires_cpu_device = False
 
-    def _relax_vm_executable_executor(
+    def _vm_executable_executor(
         self, executable: Union[runtime.Executable, str], hexagon_arch: str
     ):
         """Create a local TVM module which consumes a remote vm executable.
diff --git a/python/tvm/relax/__init__.py b/python/tvm/relax/__init__.py
index da288942edf0..b88000119897 100644
--- a/python/tvm/relax/__init__.py
+++ b/python/tvm/relax/__init__.py
@@ -16,8 +16,8 @@
 # under the License.
 # pylint: disable=invalid-name, wrong-import-position
 """The Relax IR namespace containing the IR, type, operator, builder, vm, etc."""
-from tvm.runtime import relax_vm as vm
-from tvm.runtime.relax_vm import VirtualMachine, VMInstrumentReturnKind
+from tvm.runtime import vm
+from tvm.runtime.vm import VirtualMachine, VMInstrumentReturnKind
 
 from .type_converter import args_converter
 
diff --git a/python/tvm/relax/frontend/nn/core.py b/python/tvm/relax/frontend/nn/core.py
index 5b0eae4acc0e..068b2090db5b 100644
--- a/python/tvm/relax/frontend/nn/core.py
+++ b/python/tvm/relax/frontend/nn/core.py
@@ -45,7 +45,7 @@
 from tvm.runtime import Device, NDArray
 from tvm.runtime import device as as_device
 from tvm.runtime import ndarray
-from tvm.runtime.relax_vm import VirtualMachine
+from tvm.runtime.vm import VirtualMachine
 from tvm.target import Target
 
 from .... import relax as rx
diff --git a/python/tvm/relax/frontend/nn/torch.py b/python/tvm/relax/frontend/nn/torch.py
index 28ccf3b8ea2f..ae98868dae09 100644
--- a/python/tvm/relax/frontend/nn/torch.py
+++ b/python/tvm/relax/frontend/nn/torch.py
@@ -22,7 +22,7 @@
 
 from tvm.ir import Array
 from tvm.runtime import NDArray, ShapeTuple, ndarray
-from tvm.runtime.relax_vm import VirtualMachine
+from tvm.runtime.vm import VirtualMachine
 
 from . import core
 from . import spec as _spec
diff --git a/python/tvm/runtime/relax_vm.py b/python/tvm/runtime/vm.py
similarity index 100%
rename from python/tvm/runtime/relax_vm.py
rename to python/tvm/runtime/vm.py
diff --git a/src/relax/backend/vm/codegen_vm.cc b/src/relax/backend/vm/codegen_vm.cc
index f61579e25e96..4cf2811922c8 100644
--- a/src/relax/backend/vm/codegen_vm.cc
+++ b/src/relax/backend/vm/codegen_vm.cc
@@ -24,7 +24,7 @@
 #include <tvm/relax/exec_builder.h>
 #include <tvm/relax/expr_functor.h>
 #include <tvm/relax/op_attr_types.h>
-#include <tvm/runtime/relax_vm/bytecode.h>
+#include <tvm/runtime/vm/bytecode.h>
 #include <tvm/target/target.h>
 #include <tvm/tir/function.h>
 
@@ -37,12 +37,12 @@
 
 namespace tvm {
 namespace relax {
-namespace relax_vm {
+namespace codegen_vm {
 
 using tvm::Target;
 using namespace relax;
 using namespace tvm::runtime;
-using namespace tvm::runtime::relax_vm;
+using namespace tvm::runtime::vm;
 
 /*!
  * \brief A class to generate VM executable for Relax functions.
@@ -492,6 +492,6 @@ Module VMLink(ExecBuilder builder, Target target, Optional<Module> lib, Array<Mo
 
 TVM_FFI_REGISTER_GLOBAL("relax.VMLink").set_body_typed(VMLink);
 
-}  // namespace relax_vm
+}  // namespace codegen_vm
 }  // namespace relax
 }  // namespace tvm
diff --git a/src/relax/backend/vm/codegen_vm_tir.cc b/src/relax/backend/vm/codegen_vm_tir.cc
index aa228822ded7..72789c9c1f86 100644
--- a/src/relax/backend/vm/codegen_vm_tir.cc
+++ b/src/relax/backend/vm/codegen_vm_tir.cc
@@ -25,7 +25,7 @@
 #include <tvm/relax/exec_builder.h>
 #include <tvm/relax/expr_functor.h>
 #include <tvm/relax/op_attr_types.h>
-#include <tvm/runtime/relax_vm/executable.h>
+#include <tvm/runtime/vm/executable.h>
 #include <tvm/target/target.h>
 #include <tvm/tir/builtin.h>
 #include <tvm/tir/expr.h>
@@ -39,7 +39,7 @@
 
 namespace tvm {
 namespace relax {
-namespace relax_vm {
+namespace codegen_vm {
 
 using vm::VMFuncInfo;
 
@@ -532,6 +532,6 @@ IRModule VMTIRCodeGen(ExecBuilder exec_builder, IRModule mod) {
 
 TVM_FFI_REGISTER_GLOBAL("relax.VMTIRCodeGen").set_body_typed(VMTIRCodeGen);
 
-}  // namespace relax_vm
+}  // namespace codegen_vm
 }  // namespace relax
 }  // namespace tvm
diff --git a/src/relax/backend/vm/vm_shape_lower.cc b/src/relax/backend/vm/vm_shape_lower.cc
index 0b60553034fe..50e71ccc14ed 100644
--- a/src/relax/backend/vm/vm_shape_lower.cc
+++ b/src/relax/backend/vm/vm_shape_lower.cc
@@ -25,7 +25,7 @@
 #include <tvm/relax/expr_functor.h>
 #include <tvm/relax/struct_info.h>
 #include <tvm/relax/struct_info_functor.h>
-#include <tvm/runtime/relax_vm/builtin.h>
+#include <tvm/runtime/vm/builtin.h>
 #include <tvm/tir/analysis.h>
 #include <tvm/tir/function.h>
 #include <tvm/tir/op.h>
@@ -176,7 +176,7 @@ class PrimExprSlotCollector : public ExprVisitor, public StructInfoVisitor {
  * - VisitExpr_(ShapeExprNode*): makes symbolic shape tuple.
  *
  * The checks and symbolic shape all maps to runtime builtin functions. Please checkout
- * runtime/relax_vm/builtin.cc for their definitions.
+ * runtime/vm/builtin.cc for their definitions.
  *
  * Shape computation are lowered to host-side TIR functions that load var from slot
  * and store computed results into the slot. For a given slot map: {0:m, 1: n+1: 2: n}
@@ -353,7 +353,7 @@ class VMShapeLowerMutator
   }
 
   std::pair<Expr, Expr> MakeSymbolicShapeArg(const PrimExpr& expr) {
-    using runtime::relax_vm::MakeShapeCode;
+    using runtime::vm::MakeShapeCode;
 
     if (auto* int_expr = expr.as<IntImmNode>()) {
       return {PrimValue::Int64(static_cast<int>(MakeShapeCode::kUseImm)),
@@ -370,7 +370,7 @@ class VMShapeLowerMutator
   }
 
   Expr VisitExpr_(const PrimValueNode* op) final {
-    using runtime::relax_vm::MakeShapeCode;
+    using runtime::vm::MakeShapeCode;
     // Constant shape can be preserved.
     bool is_const_value =
         op->value->IsInstance<IntImmNode>() || op->value->IsInstance<FloatImmNode>();
@@ -389,7 +389,7 @@ class VMShapeLowerMutator
   }
 
   Expr VisitExpr_(const ShapeExprNode* op) final {
-    using runtime::relax_vm::MakeShapeCode;
+    using runtime::vm::MakeShapeCode;
     // Constant shape can be preserved.
     bool is_const_shape = std::all_of(op->values.begin(), op->values.end(), [](const PrimExpr& e) {
       return e->IsInstance<IntImmNode>();
@@ -444,9 +444,9 @@ class VMShapeLowerMutator
    * \return The MatchShapeCode, and a relax expression specifying the
    *    argument used by that MatchShapeCode.
    */
-  std::pair<runtime::relax_vm::MatchShapeCode, Expr> MakeMatchArgs(const PrimExpr& expr,
-                                                                   bool require_value_computed) {
-    using runtime::relax_vm::MatchShapeCode;
+  std::pair<runtime::vm::MatchShapeCode, Expr> MakeMatchArgs(const PrimExpr& expr,
+                                                             bool require_value_computed) {
+    using runtime::vm::MatchShapeCode;
 
     if (auto* int_expr = expr.as<IntImmNode>()) {
       return {MatchShapeCode::kAssertEqualToImm, PrimValue::Int64(int_expr->value)};
@@ -496,7 +496,7 @@ class VMShapeLowerMutator
                                            bool require_value_computed) {
     std::vector<MatchShapeTodoItem> outstanding_todos;
 
-    using runtime::relax_vm::MatchShapeCode;
+    using runtime::vm::MatchShapeCode;
     for (const MatchShapeTodoItem& item : match_todos) {
       bool all_nop = true;
       bool any_nop = false;
diff --git a/src/runtime/disco/builtin.cc b/src/runtime/disco/builtin.cc
index 7c769b7dd081..9cd76e673a45 100644
--- a/src/runtime/disco/builtin.cc
+++ b/src/runtime/disco/builtin.cc
@@ -22,7 +22,7 @@
 #include <tvm/runtime/disco/builtin.h>
 #include <tvm/runtime/disco/disco_worker.h>
 #include <tvm/runtime/disco/session.h>
-#include <tvm/runtime/relax_vm/vm.h>
+#include <tvm/runtime/vm/vm.h>
 
 #include <sstream>
 
diff --git a/src/runtime/disco/loader.cc b/src/runtime/disco/loader.cc
index f93170d02f07..52e7833601f3 100644
--- a/src/runtime/disco/loader.cc
+++ b/src/runtime/disco/loader.cc
@@ -24,7 +24,7 @@
 #include <tvm/ffi/function.h>
 #include <tvm/runtime/data_type.h>
 #include <tvm/runtime/disco/builtin.h>
-#include <tvm/runtime/relax_vm/ndarray_cache_support.h>
+#include <tvm/runtime/vm/ndarray_cache_support.h>
 
 #include <functional>
 #include <numeric>
@@ -38,7 +38,7 @@
 namespace tvm {
 namespace runtime {
 
-using relax_vm::NDArrayCacheMetadata;
+using vm::NDArrayCacheMetadata;
 using FileRecord = NDArrayCacheMetadata::FileRecord;
 using ParamRecord = NDArrayCacheMetadata::FileRecord::ParamRecord;
 
diff --git a/src/runtime/relax_vm/attn_backend.cc b/src/runtime/vm/attn_backend.cc
similarity index 97%
rename from src/runtime/relax_vm/attn_backend.cc
rename to src/runtime/vm/attn_backend.cc
index 09f2d2c736fc..f3aaf3f68835 100644
--- a/src/runtime/relax_vm/attn_backend.cc
+++ b/src/runtime/vm/attn_backend.cc
@@ -17,13 +17,13 @@
  * under the License.
  */
 
-/*! \file src/runtime/relax_vm/attn_backend.cc */
+/*! \file src/runtime/vm/attn_backend.cc */
 
 #include "attn_backend.h"
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 std::unique_ptr<PagedPrefillFunc> ConvertPagedPrefillFunc(Array<ObjectRef> args,
                                                           AttnKind attn_kind) {
@@ -120,6 +120,6 @@ std::unique_ptr<RaggedPrefillTreeMaskFunc> ConvertRaggedPrefillTreeMaskFunc(Arra
   throw;
 }
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
diff --git a/src/runtime/relax_vm/attn_backend.h b/src/runtime/vm/attn_backend.h
similarity index 99%
rename from src/runtime/relax_vm/attn_backend.h
rename to src/runtime/vm/attn_backend.h
index 2eb9cf3d6677..21d8d81a8be8 100644
--- a/src/runtime/relax_vm/attn_backend.h
+++ b/src/runtime/vm/attn_backend.h
@@ -18,12 +18,12 @@
  */
 
 /*!
- * \file src/runtime/relax_vm/attn_backend.h
+ * \file src/runtime/vm/attn_backend.h
  * \brief The attention backend classes used by KV cache.
  */
 
-#ifndef TVM_RUNTIME_RELAX_VM_ATTN_BACKEND_H_
-#define TVM_RUNTIME_RELAX_VM_ATTN_BACKEND_H_
+#ifndef TVM_RUNTIME_VM_ATTN_BACKEND_H_
+#define TVM_RUNTIME_VM_ATTN_BACKEND_H_
 
 #include <tvm/ffi/container/array.h>
 #include <tvm/ffi/function.h>
@@ -39,7 +39,7 @@
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 /*! \brief The attention backend kinds. */
 enum class AttnBackendKind : int {
@@ -537,8 +537,8 @@ std::unique_ptr<PagedPrefillTreeMaskFunc> ConvertPagedPrefillTreeMaskFunc(Array<
 std::unique_ptr<RaggedPrefillTreeMaskFunc> ConvertRaggedPrefillTreeMaskFunc(Array<ObjectRef> args,
                                                                             AttnKind attn_kind);
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
 
-#endif  // TVM_RUNTIME_RELAX_VM_ATTN_BACKEND_H_
+#endif  // TVM_RUNTIME_VM_ATTN_BACKEND_H_
diff --git a/src/runtime/relax_vm/attn_utils.h b/src/runtime/vm/attn_utils.h
similarity index 99%
rename from src/runtime/relax_vm/attn_utils.h
rename to src/runtime/vm/attn_utils.h
index 547d32c1208e..dce577da0889 100644
--- a/src/runtime/relax_vm/attn_utils.h
+++ b/src/runtime/vm/attn_utils.h
@@ -17,12 +17,12 @@
  * under the License.
  */
 /*!
- * \file src/runtime/relax_vm/attn_utils.h
+ * \file src/runtime/vm/attn_utils.h
  * \brief Data structure and utilities for KV cache.
  */
 
-#ifndef TVM_RUNTIME_RELAX_VM_ATTN_UTILS_H_
-#define TVM_RUNTIME_RELAX_VM_ATTN_UTILS_H_
+#ifndef TVM_RUNTIME_VM_ATTN_UTILS_H_
+#define TVM_RUNTIME_VM_ATTN_UTILS_H_
 
 #include <tvm/runtime/ndarray.h>
 
@@ -36,7 +36,7 @@
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 /*!
  * \brief The maximum allowed block depth (a.k.a. number of common
@@ -1023,8 +1023,8 @@ class CachedPagedKVCacheAuxDataManager : public PagedKVCacheAuxDataManager {
   NDArray merged_compact_kv_aux_data_device_;
 };
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
 
-#endif  // TVM_RUNTIME_RELAX_VM_ATTN_UTILS_H_
+#endif  // TVM_RUNTIME_VM_ATTN_UTILS_H_
diff --git a/src/runtime/relax_vm/builtin.cc b/src/runtime/vm/builtin.cc
similarity index 99%
rename from src/runtime/relax_vm/builtin.cc
rename to src/runtime/vm/builtin.cc
index 3d7904bd8f48..a23e196e5c15 100644
--- a/src/runtime/relax_vm/builtin.cc
+++ b/src/runtime/vm/builtin.cc
@@ -17,7 +17,7 @@
  * under the License.
  */
 /*!
- * \file src/runtime/relax_vm/builtin.cc
+ * \file src/runtime/vm/builtin.cc
  */
 #include <tvm/ffi/any.h>
 #include <tvm/ffi/container/array.h>
@@ -29,13 +29,13 @@
 #include <tvm/runtime/logging.h>
 #include <tvm/runtime/memory/memory_manager.h>
 #include <tvm/runtime/ndarray.h>
-#include <tvm/runtime/relax_vm/builtin.h>
-#include <tvm/runtime/relax_vm/bytecode.h>
-#include <tvm/runtime/relax_vm/vm.h>
+#include <tvm/runtime/vm/builtin.h>
+#include <tvm/runtime/vm/bytecode.h>
+#include <tvm/runtime/vm/vm.h>
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 using tvm::runtime::NDArray;
 
@@ -562,7 +562,7 @@ TVM_FFI_REGISTER_GLOBAL("vm.builtin.ensure_zero_offset").set_body_typed([](NDArr
   }
 });
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
 
diff --git a/src/runtime/relax_vm/bytecode.cc b/src/runtime/vm/bytecode.cc
similarity index 93%
rename from src/runtime/relax_vm/bytecode.cc
rename to src/runtime/vm/bytecode.cc
index 9084207848b5..4356305521a3 100644
--- a/src/runtime/relax_vm/bytecode.cc
+++ b/src/runtime/vm/bytecode.cc
@@ -18,19 +18,19 @@
  */
 
 /*!
- * \file src/runtime/relax_vm/bytecode.cc
+ * \file src/runtime/vm/bytecode.cc
  * \brief The bytecode for Relax virtual machine.
  */
 
 #include <tvm/runtime/logging.h>
-#include <tvm/runtime/relax_vm/bytecode.h>
+#include <tvm/runtime/vm/bytecode.h>
 
 #include <functional>
 #include <sstream>
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 Instruction Instruction::Call(Index func_idx, Index num_args, Instruction::Arg* args, RegName dst) {
   Instruction instr;
@@ -63,6 +63,6 @@ Instruction Instruction::If(RegName cond, Index false_offset) {
   instr.false_offset = false_offset;
   return instr;
 }
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
diff --git a/src/runtime/relax_vm/cuda/cuda_graph_builtin.cc b/src/runtime/vm/cuda/cuda_graph_builtin.cc
similarity index 97%
rename from src/runtime/relax_vm/cuda/cuda_graph_builtin.cc
rename to src/runtime/vm/cuda/cuda_graph_builtin.cc
index d3484cbc7b3e..fb0534eec5af 100644
--- a/src/runtime/relax_vm/cuda/cuda_graph_builtin.cc
+++ b/src/runtime/vm/cuda/cuda_graph_builtin.cc
@@ -18,19 +18,19 @@
  */
 
 /*!
- * \file src/runtime/relax_vm/cuda_graph_builtin.cc
+ * \file src/runtime/vm/cuda_graph_builtin.cc
  * \brief The CUDA graph related builtin functions for Relax virtual machine.
  */
 
 #include <tvm/ffi/container/array.h>
 #include <tvm/ffi/function.h>
-#include <tvm/runtime/relax_vm/vm.h>
+#include <tvm/runtime/vm/vm.h>
 
 #include "../../../support/utils.h"
 #include "../../cuda/cuda_common.h"
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 namespace {
 
@@ -213,7 +213,7 @@ class CUDAGraphExtensionNode : public VMExtensionNode {
     return alloc_result;
   }
 
-  static constexpr const char* _type_key = "relax_vm.CUDAGraphExtension";
+  static constexpr const char* _type_key = "vm.CUDAGraphExtension";
 
  private:
   /*!
@@ -265,6 +265,6 @@ TVM_FFI_REGISTER_GLOBAL("vm.builtin.cuda_graph.get_cached_alloc")
       *rv = extension->GetCachedAllocation(vm, alloc_func, entry_index);
     });
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
diff --git a/src/runtime/relax_vm/executable.cc b/src/runtime/vm/executable.cc
similarity index 98%
rename from src/runtime/relax_vm/executable.cc
rename to src/runtime/vm/executable.cc
index 52a0588be35c..f33ce6bc0edb 100644
--- a/src/runtime/relax_vm/executable.cc
+++ b/src/runtime/vm/executable.cc
@@ -18,13 +18,13 @@
  */
 
 /*!
- * \file src/runtime/relax_vm/executable.cc
+ * \file src/runtime/vm/executable.cc
  */
 
 #include <dmlc/memory_io.h>
 #include <tvm/runtime/logging.h>
-#include <tvm/runtime/relax_vm/executable.h>
-#include <tvm/runtime/relax_vm/vm.h>
+#include <tvm/runtime/vm/executable.h>
+#include <tvm/runtime/vm/vm.h>
 
 #include <functional>
 #include <sstream>
@@ -33,7 +33,7 @@
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 /*! \brief The magic number for the serialized VM bytecode file  */
 constexpr uint64_t kTVMVMBytecodeMagic = 0xD225DE2F4214151D;
@@ -144,7 +144,7 @@ Instruction VMExecutable::GetInstruction(Index i) const {
 void SaveHeader(dmlc::Stream* strm) {
   uint64_t header = kTVMVMBytecodeMagic;
   strm->Write(header);
-  std::string version = RELAX_VM_VERSION;
+  std::string version = VM_VERSION;
   strm->Write(version);
 }
 
@@ -157,7 +157,7 @@ void LoadHeader(dmlc::Stream* strm) {
   // Check version.
   std::string version;
   STREAM_CHECK(strm->Read(&version), "version");
-  STREAM_CHECK(version == RELAX_VM_VERSION, "version");
+  STREAM_CHECK(version == VM_VERSION, "version");
 }
 
 void VMExecutable::SaveToBinary(dmlc::Stream* stream) {
@@ -559,6 +559,6 @@ String VMExecutable::AsPython() const {
 
 TVM_FFI_REGISTER_GLOBAL("relax.ExecutableLoadFromFile").set_body_typed(VMExecutable::LoadFromFile);
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
diff --git a/src/runtime/relax_vm/hexagon/builtin.cc b/src/runtime/vm/hexagon/builtin.cc
similarity index 95%
rename from src/runtime/relax_vm/hexagon/builtin.cc
rename to src/runtime/vm/hexagon/builtin.cc
index 89f1708b28ed..c48445130cf0 100644
--- a/src/runtime/relax_vm/hexagon/builtin.cc
+++ b/src/runtime/vm/hexagon/builtin.cc
@@ -18,18 +18,18 @@
  */
 
 /*!
- * \file src/runtime/relax_vm/hexagon/builtin.cc
+ * \file src/runtime/vm/hexagon/builtin.cc
  * \brief The hexagon graph related builtin functions for Relax virtual machine.
  */
 
 #include <tvm/ffi/function.h>
 #include <tvm/runtime/device_api.h>
-#include <tvm/runtime/relax_vm/vm.h>
+#include <tvm/runtime/vm/vm.h>
 
 #include "../../hexagon/hexagon_device_api.h"
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 TVM_FFI_REGISTER_GLOBAL("vm.builtin.hexagon.dma_copy")
     .set_body_typed([](ffi::AnyView vm_ptr, NDArray src_arr, NDArray dst_arr, int queue_id,
@@ -66,6 +66,6 @@ TVM_FFI_REGISTER_GLOBAL("vm.builtin.hexagon.dma_wait")
                              QURT_MEM_DCACHE);
       }
     });
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
diff --git a/src/runtime/relax_vm/kv_state.cc b/src/runtime/vm/kv_state.cc
similarity index 99%
rename from src/runtime/relax_vm/kv_state.cc
rename to src/runtime/vm/kv_state.cc
index 12f52c0794e9..466d41e3d31e 100644
--- a/src/runtime/relax_vm/kv_state.cc
+++ b/src/runtime/vm/kv_state.cc
@@ -23,7 +23,7 @@
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 // Register Object Type
 TVM_REGISTER_OBJECT_TYPE(KVStateObj);
@@ -114,6 +114,6 @@ TVM_FFI_REGISTER_GLOBAL("vm.builtin.rnn_state_set")
     });
 TVM_FFI_REGISTER_GLOBAL("vm.builtin.rnn_state_debug_get").set_body_method(&RNNStateObj::DebugGet);
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
diff --git a/src/runtime/relax_vm/kv_state.h b/src/runtime/vm/kv_state.h
similarity index 98%
rename from src/runtime/relax_vm/kv_state.h
rename to src/runtime/vm/kv_state.h
index 5800c4e2db93..46d8f4f59603 100644
--- a/src/runtime/relax_vm/kv_state.h
+++ b/src/runtime/vm/kv_state.h
@@ -16,8 +16,8 @@
  * specific language governing permissions and limitations
  * under the License.
  */
-#ifndef TVM_RUNTIME_RELAX_VM_KV_STATE_H_
-#define TVM_RUNTIME_RELAX_VM_KV_STATE_H_
+#ifndef TVM_RUNTIME_VM_KV_STATE_H_
+#define TVM_RUNTIME_VM_KV_STATE_H_
 #include <tvm/ffi/container/array.h>
 #include <tvm/ffi/function.h>
 #include <tvm/runtime/device_api.h>
@@ -28,7 +28,7 @@
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 /*! \brief The base class of attention KV cache and rnn state. */
 class KVStateObj : public Object {
@@ -345,8 +345,8 @@ class RNNState : public KVState {
   TVM_DEFINE_MUTABLE_OBJECT_REF_METHODS(RNNState, KVState, RNNStateObj);
 };
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
 
-#endif  // TVM_RUNTIME_RELAX_VM_KV_STATE_H_
+#endif  // TVM_RUNTIME_VM_KV_STATE_H_
diff --git a/src/runtime/relax_vm/lm_support.cc b/src/runtime/vm/lm_support.cc
similarity index 99%
rename from src/runtime/relax_vm/lm_support.cc
rename to src/runtime/vm/lm_support.cc
index 8abeddcf18dc..baf5bf480e37 100644
--- a/src/runtime/relax_vm/lm_support.cc
+++ b/src/runtime/vm/lm_support.cc
@@ -17,7 +17,7 @@
  * under the License.
  */
 /*!
- * \file src/runtime/relax_vm/lm_support.cc
+ * \file src/runtime/vm/lm_support.cc
  * \brief Runtime to support language model related task
  *
  * Including inplace attention kv cache for runtime and simple sampler.
@@ -42,13 +42,13 @@
 #include <tvm/runtime/logging.h>
 #include <tvm/runtime/memory/memory_manager.h>
 #include <tvm/runtime/ndarray.h>
-#include <tvm/runtime/relax_vm/vm.h>
+#include <tvm/runtime/vm/vm.h>
 
 #include <cmath>
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 //-------------------------------------------
 // We keep the implementation private as
@@ -621,6 +621,6 @@ void ApplySoftmaxWithTemperature(NDArray logits, double temperature) {
 TVM_FFI_REGISTER_GLOBAL("vm.builtin.apply_softmax_with_temperature")
     .set_body_typed(ApplySoftmaxWithTemperature);
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
diff --git a/src/runtime/relax_vm/ndarray_cache_support.cc b/src/runtime/vm/ndarray_cache_support.cc
similarity index 98%
rename from src/runtime/relax_vm/ndarray_cache_support.cc
rename to src/runtime/vm/ndarray_cache_support.cc
index 7341507e9a98..17f07df70c01 100644
--- a/src/runtime/relax_vm/ndarray_cache_support.cc
+++ b/src/runtime/vm/ndarray_cache_support.cc
@@ -17,7 +17,7 @@
  * under the License.
  */
 /*!
- * \file src/runtime/relax_vm/ndarray_cache_support.cc
+ * \file src/runtime/vm/ndarray_cache_support.cc
  * \brief Runtime to support ndarray cache file loading.
  *
  * This file provides a minimum support for ndarray cache file loading.
@@ -41,7 +41,7 @@
 #include <picojson.h>
 #include <tvm/ffi/function.h>
 #include <tvm/runtime/ndarray.h>
-#include <tvm/runtime/relax_vm/ndarray_cache_support.h>
+#include <tvm/runtime/vm/ndarray_cache_support.h>
 
 #include <string>
 #include <vector>
@@ -51,7 +51,7 @@
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 template <typename ExpectedType>
 inline ExpectedType AsType(const picojson::value& json) {
@@ -375,6 +375,6 @@ TVM_FFI_REGISTER_GLOBAL("vm.builtin.param_array_from_cache_by_name_unpacked")
       *rv = ParamModuleNode::GetParamByName(names);
     });
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
diff --git a/src/runtime/relax_vm/paged_kv_cache.cc b/src/runtime/vm/paged_kv_cache.cc
similarity index 99%
rename from src/runtime/relax_vm/paged_kv_cache.cc
rename to src/runtime/vm/paged_kv_cache.cc
index 2f21e6978a68..5b70c0317719 100644
--- a/src/runtime/relax_vm/paged_kv_cache.cc
+++ b/src/runtime/vm/paged_kv_cache.cc
@@ -17,7 +17,7 @@
  * under the License.
  */
 /*!
- * \file src/runtime/relax_vm/paged_kv_cache.cc
+ * \file src/runtime/vm/paged_kv_cache.cc
  * \brief Runtime paged KV cache object for language models.
  */
 #include <tvm/ffi/function.h>
@@ -39,7 +39,7 @@
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 //-------------------------------------------
 // We keep the implementation private as
@@ -2276,7 +2276,7 @@ class PagedAttentionKVCacheObj : public AttentionKVCacheObj {
     // - Reset the dirty flag to false.
     dirty_aux_data_device_ = false;
   }
-};  // namespace relax_vm
+};  // namespace vm
 
 TVM_REGISTER_OBJECT_TYPE(PagedAttentionKVCacheObj);
 
@@ -2387,6 +2387,6 @@ TVM_FFI_REGISTER_GLOBAL("vm.builtin.paged_attention_kv_cache_create")
       *rv = AttentionKVCache(std::move(n));
     });
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
diff --git a/src/runtime/relax_vm/rnn_state.cc b/src/runtime/vm/rnn_state.cc
similarity index 99%
rename from src/runtime/relax_vm/rnn_state.cc
rename to src/runtime/vm/rnn_state.cc
index d431fdb2ae2f..e7a655d392f8 100644
--- a/src/runtime/relax_vm/rnn_state.cc
+++ b/src/runtime/vm/rnn_state.cc
@@ -17,7 +17,7 @@
  * under the License.
  */
 /*!
- * \file src/runtime/relax_vm/rnn_state.cc
+ * \file src/runtime/vm/rnn_state.cc
  * \brief Runtime RNN state object for space state models.
  */
 
@@ -28,7 +28,7 @@
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 //-----------------------------------------------------------------------------
 // We keep the implementation private as they may subject to future changes.
@@ -495,6 +495,6 @@ TVM_FFI_REGISTER_GLOBAL("vm.builtin.rnn_state_create")
       return RNNState(std::move(n));
     });
 
-}  // namespace relax_vm
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
diff --git a/src/runtime/relax_vm/vm.cc b/src/runtime/vm/vm.cc
similarity index 99%
rename from src/runtime/relax_vm/vm.cc
rename to src/runtime/vm/vm.cc
index 8d0b928f8592..55a5a87d279a 100644
--- a/src/runtime/relax_vm/vm.cc
+++ b/src/runtime/vm/vm.cc
@@ -18,21 +18,21 @@
  */
 
 /*!
- * \file src/runtime/relax_vm/vm.cc
+ * \file src/runtime/vm/vm.cc
  */
 #include <dlpack/dlpack.h>
 #include <tvm/ffi/function.h>
 #include <tvm/runtime/memory/memory_manager.h>
 #include <tvm/runtime/nvtx.h>
 #include <tvm/runtime/profiling.h>
-#include <tvm/runtime/relax_vm/vm.h>
+#include <tvm/runtime/vm/vm.h>
 
 #include <optional>
 #include <thread>
 
 namespace tvm {
 namespace runtime {
-namespace relax_vm {
+namespace vm {
 
 //---------------------------------------------
 // VM Closure object
@@ -965,7 +965,7 @@ ffi::Function VirtualMachineImpl::_LookupFunction(const String& name) {
 //----------------------------------------------------------------
 // Profiler can be optionally disabled via a macro to reduce dep.
 //----------------------------------------------------------------
-#if TVM_RELAX_VM_ENABLE_PROFILER
+#if TVM_VM_ENABLE_PROFILER
 
 /*!
  * \brief An extension of VirtualMachineImpl to support per-op profiling
@@ -1081,7 +1081,7 @@ ObjectPtr<VirtualMachine> VirtualMachine::CreateProfiler() {
   LOG(FATAL) << "Profiler support is disabled";
   return nullptr;
 }
-#endif  // TVM_RELAX_VM_ENABLE_PROFILER
-}  // namespace relax_vm
+#endif  // TVM_VM_ENABLE_PROFILER
+}  // namespace vm
 }  // namespace runtime
 }  // namespace tvm
diff --git a/tests/python/contrib/test_tir_triton_integration.py b/tests/python/contrib/test_tir_triton_integration.py
index 69f727150544..b349d2fabce5 100644
--- a/tests/python/contrib/test_tir_triton_integration.py
+++ b/tests/python/contrib/test_tir_triton_integration.py
@@ -14,17 +14,18 @@
 # KIND, either express or implied.  See the License for the
 # specific language governing permissions and limitations
 # under the License.
-import numpy as np
 import sys
 
+import numpy as np
+import pytest
+
 import tvm
-from tvm.script import tir as T
-from tvm.script import relax as R
-from tvm.script import ir as I
+import tvm.testing
 from tvm import relax
 from tvm.relax.frontend import nn
-import tvm.testing
-import pytest
+from tvm.script import ir as I
+from tvm.script import relax as R
+from tvm.script import tir as T
 
 try:
     import triton
@@ -115,5 +116,5 @@ def add(x_handle: T.handle, y_handle: T.handle, output_handle: T.handle):
 
     with tvm.target.Target("cuda"):
         lib = tvm.compile(Module)
-        output_nd = tvm.runtime.relax_vm.VirtualMachine(lib, device)["main"](x_nd, y_nd)
+        output_nd = tvm.runtime.vm.VirtualMachine(lib, device)["main"](x_nd, y_nd)
         tvm.testing.assert_allclose(output_nd.numpy(), output_np, rtol=1e-5)
diff --git a/tests/python/disco/test_ccl.py b/tests/python/disco/test_ccl.py
index f93030f63923..649b865b6c3b 100644
--- a/tests/python/disco/test_ccl.py
+++ b/tests/python/disco/test_ccl.py
@@ -28,7 +28,7 @@
 from tvm import get_global_func
 from tvm import relax as rx
 from tvm.runtime import disco as di
-from tvm.runtime.relax_vm import VirtualMachine
+from tvm.runtime.vm import VirtualMachine
 from tvm.script import relax as R
 
 _all_session_kinds = [di.ThreadedSession, di.ProcessSession]
diff --git a/tests/python/relax/test_frontend_nn_extern_module.py b/tests/python/relax/test_frontend_nn_extern_module.py
index 5aa48dd18aba..cbc2e7f42922 100644
--- a/tests/python/relax/test_frontend_nn_extern_module.py
+++ b/tests/python/relax/test_frontend_nn_extern_module.py
@@ -189,7 +189,7 @@ def test_sym(self, a: nn.Tensor, b: nn.Tensor):  # pylint: disable=invalid-name
         )
         _check_ir_equality(mod)
         mod = AttachExternModules(ext_mods)(mod)  # pylint: disable=not-callable
-        compiled = tvm.runtime.relax_vm.VirtualMachine(
+        compiled = tvm.runtime.vm.VirtualMachine(
             tvm.compile(mod, target="llvm"),
             device=tvm.cpu(),
         )
@@ -238,7 +238,7 @@ def test_sym(self, a: nn.Tensor, b: nn.Tensor):  # pylint: disable=invalid-name
     )
     _check_ir_equality(mod)
     mod = AttachExternModules(ext_mods)(mod)  # pylint: disable=not-callable
-    compiled = tvm.runtime.relax_vm.VirtualMachine(
+    compiled = tvm.runtime.vm.VirtualMachine(
         tvm.compile(mod, target="llvm"),
         device=tvm.cpu(),
     )
diff --git a/tests/python/relax/test_tir_call_source_kernel.py b/tests/python/relax/test_tir_call_source_kernel.py
index 9aba489f480b..d7ca2a672b55 100644
--- a/tests/python/relax/test_tir_call_source_kernel.py
+++ b/tests/python/relax/test_tir_call_source_kernel.py
@@ -20,7 +20,9 @@
 import tvm
 import tvm.testing
 from tvm import relax
-from tvm.script import tir as T, ir as I, relax as R
+from tvm.script import ir as I
+from tvm.script import relax as R
+from tvm.script import tir as T
 
 add_cuda_source = """
 extern "C" __global__ void add_kernel(float* x, float* y, float* output, int n_elements) {
@@ -96,5 +98,5 @@ def add(x_handle: T.handle, y_handle: T.handle, output_handle: T.handle):
 
     with tvm.target.Target("cuda"):
         lib = tvm.compile(Module)
-        output_nd = tvm.runtime.relax_vm.VirtualMachine(lib, device)["main"](x_nd, y_nd)
+        output_nd = tvm.runtime.vm.VirtualMachine(lib, device)["main"](x_nd, y_nd)
         tvm.testing.assert_allclose(output_nd.numpy(), output_np, rtol=1e-5)
diff --git a/tests/python/relax/test_training_optimizer_numeric.py b/tests/python/relax/test_training_optimizer_numeric.py
index 937582adf715..6a9c34a5fb94 100644
--- a/tests/python/relax/test_training_optimizer_numeric.py
+++ b/tests/python/relax/test_training_optimizer_numeric.py
@@ -18,13 +18,13 @@
 from typing import Callable, List
 
 import numpy as np
+
 import tvm
 import tvm.testing
-from tvm import relax
-from tvm import IRModule
-from tvm.relax.training.optimizer import Adam, SGD, MomentumSGD
+from tvm import IRModule, relax
+from tvm.relax.training.optimizer import SGD, Adam, MomentumSGD
+from tvm.runtime.vm import VirtualMachine
 from tvm.script.parser import relax as R
-from tvm.runtime.relax_vm import VirtualMachine
 from tvm.testing import assert_allclose
 
 
diff --git a/web/emcc/wasm_runtime.cc b/web/emcc/wasm_runtime.cc
index 40dfb31ad19f..3c1c15a86123 100644
--- a/web/emcc/wasm_runtime.cc
+++ b/web/emcc/wasm_runtime.cc
@@ -60,16 +60,16 @@
 #include "ffi/src/ffi/traceback.cc"
 #include "src/runtime/memory/memory_manager.cc"
 #include "src/runtime/nvtx.cc"
-#include "src/runtime/relax_vm/attn_backend.cc"
-#include "src/runtime/relax_vm/builtin.cc"
-#include "src/runtime/relax_vm/bytecode.cc"
-#include "src/runtime/relax_vm/executable.cc"
-#include "src/runtime/relax_vm/kv_state.cc"
-#include "src/runtime/relax_vm/lm_support.cc"
-#include "src/runtime/relax_vm/ndarray_cache_support.cc"
-#include "src/runtime/relax_vm/paged_kv_cache.cc"
-#include "src/runtime/relax_vm/rnn_state.cc"
-#include "src/runtime/relax_vm/vm.cc"
+#include "src/runtime/vm/attn_backend.cc"
+#include "src/runtime/vm/builtin.cc"
+#include "src/runtime/vm/bytecode.cc"
+#include "src/runtime/vm/executable.cc"
+#include "src/runtime/vm/kv_state.cc"
+#include "src/runtime/vm/lm_support.cc"
+#include "src/runtime/vm/ndarray_cache_support.cc"
+#include "src/runtime/vm/paged_kv_cache.cc"
+#include "src/runtime/vm/rnn_state.cc"
+#include "src/runtime/vm/vm.cc"
 
 // --- Implementations of backend and wasm runtime API. ---
 
diff --git a/web/tests/node/test_relax_vm.js b/web/tests/node/test_vm.js
similarity index 100%
rename from web/tests/node/test_relax_vm.js
rename to web/tests/node/test_vm.js