[REFACTOR][PY][API-Change] Polish tvm.runtime, tvm.runtime.module API update

apps/bundle_deploy/bundle.cc

@@ -34,7 +34,7 @@ TVM_BUNDLE_FUNCTION void *tvm_runtime_create() {
   const std::string json_data(&build_graph_json[0],
                               &build_graph_json[0] + build_graph_json_len);
   tvm::runtime::Module mod_syslib =
-      (*tvm::runtime::Registry::Get("module._GetSystemLib"))();
+      (*tvm::runtime::Registry::Get("runtime.SystemLib"))();
   int device_type = kDLCPU;
   int device_id = 0;
   tvm::runtime::Module mod =

apps/dso_plugin_module/README.md

@@ -19,7 +19,7 @@
 Example Plugin Module
 =====================
 This folder contains an example that implements a C++ module
-that can be directly loaded as TVM's DSOModule (via tvm.module.load)
+that can be directly loaded as TVM's DSOModule (via tvm.runtime.load_module)
 
 ## Guideline
 

apps/dso_plugin_module/test_plugin_module.py

@@ -19,7 +19,7 @@
 
 def test_plugin_module():
     curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
-    mod = tvm.module.load(os.path.join(curr_path, "lib", "plugin_module.so"))
+    mod = tvm.runtime.load_module(os.path.join(curr_path, "lib", "plugin_module.so"))
     # NOTE: we need to make sure all managed resources returned
     # from mod get destructed before mod get unloaded.
     #

apps/extension/tests/test_ext.py

@@ -30,7 +30,7 @@ def test_ext_dev():
     B = tvm.compute((n,), lambda *i: A(*i) + 1.0, name='B')
     s = tvm.create_schedule(B.op)
     def check_llvm():
-        if not tvm.module.enabled("llvm"):
+        if not tvm.runtime.enabled("llvm"):
             return
         f = tvm.build(s, [A, B], "ext_dev", "llvm")
         ctx = tvm.ext_dev(0)
@@ -74,7 +74,7 @@ def test_extern_call():
     s = tvm.create_schedule(B.op)
 
     def check_llvm():
-        if not tvm.module.enabled("llvm"):
+        if not tvm.runtime.enabled("llvm"):
             return
         f = tvm.build(s, [A, B], "llvm")
         ctx = tvm.cpu(0)

apps/howto_deploy/cpp_deploy.cc

@@ -6,9 +6,9 @@
  * to you under the Apache License, Version 2.0 (the
  * "License"); you may not use this file except in compliance
  * with the License.  You may obtain a copy of the License at
- * 
+ *
  *   http://www.apache.org/licenses/LICENSE-2.0
- * 
+ *
  * Unless required by applicable law or agreed to in writing,
  * software distributed under the License is distributed on an
  * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
@@ -79,7 +79,7 @@ int main(void) {
   // For libraries that are directly packed as system lib and linked together with the app
   // We can directly use GetSystemLib to get the system wide library.
   LOG(INFO) << "Verify load function from system lib";
-  tvm::runtime::Module mod_syslib = (*tvm::runtime::Registry::Get("module._GetSystemLib"))();
+  tvm::runtime::Module mod_syslib = (*tvm::runtime::Registry::Get("runtime.SystemLib"))();
   Verify(mod_syslib, "addonesys");
   return 0;
 }

apps/howto_deploy/python_deploy.py

@@ -5,9 +5,9 @@
 # to you under the Apache License, Version 2.0 (the
 # "License"); you may not use this file except in compliance
 # with the License.  You may obtain a copy of the License at
-# 
+#
 #   http://www.apache.org/licenses/LICENSE-2.0
-# 
+#
 # Unless required by applicable law or agreed to in writing,
 # software distributed under the License is distributed on an
 # "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
@@ -26,21 +26,21 @@ def verify(mod, fname):
   f = mod.get_function(fname)
   # Use tvm.nd.array to convert numpy ndarray to tvm
   # NDArray type, so that function can be invoked normally
-  N = 10 
+  N = 10
   x = tvm.nd.array(np.arange(N, dtype=np.float32))
   y = tvm.nd.array(np.zeros(N, dtype=np.float32))
   # Invoke the function
   f(x, y)
-  np_x = x.asnumpy() 
-  np_y = y.asnumpy() 
+  np_x = x.asnumpy()
+  np_y = y.asnumpy()
   # Verify correctness of function
   assert(np.all([xi+1 == yi for xi, yi in zip(np_x, np_y)]))
   print("Finish verification...")
-  
+
 
 if __name__ == "__main__":
   # The normal dynamic loading method for deployment
-  mod_dylib = tvm.module.load("lib/test_addone_dll.so")
+  mod_dylib = tvm.runtime.load_module("lib/test_addone_dll.so")
   print("Verify dynamic loading from test_addone_dll.so")
   verify(mod_dylib, "addone")
   # There might be methods to use the system lib way in

apps/sgx/run_model.py

@@ -23,7 +23,7 @@
 
 def main():
     ctx = tvm.context('cpu', 0)
-    model = tvm.module.load(osp.join(CWD, 'build', 'enclave.signed.so'))
+    model = tvm.runtime.load_module(osp.join(CWD, 'build', 'enclave.signed.so'))
     inp = tvm.nd.array(np.ones((1, 3, 224, 224), dtype='float32'), ctx)
     out = tvm.nd.array(np.empty((1, 1000), dtype='float32'), ctx)
     model(inp, out)

docs/api/python/contrib.rst

@@ -15,8 +15,8 @@
     specific language governing permissions and limitations
     under the License.
 
-Additional Contrib APIs
------------------------
+tvm.contrib
+-----------
 .. automodule:: tvm.contrib
 
 tvm.contrib.cblas
@@ -43,6 +43,11 @@ tvm.contrib.cublas
     :members:
 
 
+tvm.contrib.dlpack
+~~~~~~~~~~~~~~~~~~
+.. automodule:: tvm.contrib.dlpack
+    :members:
+
 tvm.contrib.emscripten
 ~~~~~~~~~~~~~~~~~~~~~~
 .. automodule:: tvm.contrib.emscripten
@@ -53,6 +58,11 @@ tvm.contrib.miopen
 .. automodule:: tvm.contrib.miopen
     :members:
 
+tvm.contrib.mxnet
+~~~~~~~~~~~~~~~~~
+.. automodule:: tvm.contrib.mxnet
+    :members:
+
 tvm.contrib.ndk
 ~~~~~~~~~~~~~~~
 .. automodule:: tvm.contrib.ndk
@@ -118,7 +128,6 @@ tvm.contrib.util
     :members:
 
 
-
 tvm.contrib.xcode
 ~~~~~~~~~~~~~~~~~
 .. automodule:: tvm.contrib.xcode

docs/api/python/dev.rst

@@ -20,14 +20,7 @@ Developer API
 This page contains modules that are used by developers of TVM.
 Many of these APIs are PackedFunc registered in C++ backend.
 
-tvm.object
-~~~~~~~~~~
-.. automodule:: tvm.object
 
-.. autoclass:: tvm.object.Object
-    :members:
-
-.. autofunction:: tvm.register_object
 
 tvm.expr
 ~~~~~~~~

docs/api/python/index.rst

@@ -22,21 +22,23 @@ Python API
    :maxdepth: 2
 
    tvm
+   runtime
+   ndarray
    intrin
    tensor
    schedule
    target
    build
    module
    error
-   ndarray
    container
    function
    autotvm
    graph_runtime
    rpc
    bridge
    contrib
+   ffi
    dev
    topi
    vta/index

docs/api/python/ndarray.rst

@@ -15,22 +15,22 @@
     specific language governing permissions and limitations
     under the License.
 
-tvm.ndarray
------------
-.. automodule:: tvm.ndarray
+tvm.runtime.ndarray
+-------------------
+.. automodule:: tvm.runtime.ndarray
 
-.. autoclass:: tvm.ndarray.TVMContext
+.. autoclass:: tvm.nd.NDArray
     :members:
+    :inherited-members:
+
 
-.. autoclass:: tvm.ndarray.NDArray
+.. autoclass:: tvm.runtime.TVMContext
     :members:
-    :inherited-members:
 
+.. autofunction:: tvm.context
 .. autofunction:: tvm.cpu
 .. autofunction:: tvm.gpu
 .. autofunction:: tvm.opencl
 .. autofunction:: tvm.metal
-.. autofunction:: tvm.ndarray.array
-.. autofunction:: tvm.ndarray.empty
-
-.. autofunction:: tvm.register_extension
+.. autofunction:: tvm.nd.array
+.. autofunction:: tvm.nd.empty

docs/api/python/function.rst → docs/api/python/runtime.rst

@@ -15,10 +15,31 @@
     specific language governing permissions and limitations
     under the License.
 
-tvm.Function
-------------
-.. autoclass:: tvm.Function
+tvm.runtime
+-----------
+
+.. automodule:: tvm.runtime
+
+
+.. autoclass:: tvm.runtime.PackedFunc
+   :members:
 
 .. autofunction:: tvm.register_func
 
 .. autofunction:: tvm.get_global_func
+
+
+.. autoclass:: tvm.runtime.Module
+   :members:
+
+.. autofunction:: tvm.runtime.load_module
+
+.. autofunction:: tvm.runtime.system_lib
+
+.. autofunction:: tvm.runtime.enabled
+
+
+.. autoclass:: tvm.runtime.Object
+   :members:
+
+.. autofunction:: tvm.register_object

docs/deploy/aocl_fpga.md

@@ -57,8 +57,8 @@ import os
 
 tgt="aocl_sw_emu"
 
-fadd = tvm.module.load("myadd.so")
-fadd_dev = tvm.module.load("myadd.aocx")
+fadd = tvm.runtime.load("myadd.so")
+fadd_dev = tvm.runtime.load("myadd.aocx")
 fadd.import_module(fadd_dev)
 
 ctx = tvm.context(tgt, 0)

docs/deploy/aws_fpga.md

@@ -57,11 +57,11 @@ import os
 
 tgt="sdaccel"
 
-fadd = tvm.module.load("myadd.so")
+fadd = tvm.runtime.load("myadd.so")
 if os.environ.get("XCL_EMULATION_MODE"):
-    fadd_dev = tvm.module.load("myadd.xclbin")
+    fadd_dev = tvm.runtime.load("myadd.xclbin")
 else:
-    fadd_dev = tvm.module.load("myadd.awsxclbin")
+    fadd_dev = tvm.runtime.load("myadd.awsxclbin")
 fadd.import_module(fadd_dev)
 
 ctx = tvm.context(tgt, 0)

docs/dev/introduction_to_module_serialization.rst

@@ -53,7 +53,7 @@ Let us build one ResNet-18 workload for GPU as an example first.
    resnet18_lib.export_library(path_lib)
 
    # load it back
-   loaded_lib = tvm.module.load(path_lib)
+   loaded_lib = tvm.runtime.load(path_lib)
    assert loaded_lib.type_key == "library"
    assert loaded_lib.imported_modules[0].type_key == "cuda"
 
@@ -177,7 +177,7 @@ support arbitrary modules to import ideally.
 Deserialization
 ****************
 
-The entrance API is ``tvm.module.load``. This function
+The entrance API is ``tvm.runtime.load``. This function
 is to call ``_LoadFromFile`` in fact. If we dig it a little deeper, this is
 ``Module::LoadFromFile``. In our example, the file is ``deploy.so``,
 according to the function logic, we will call ``module.loadfile_so`` in