[DOC] Improve the order of tutorials within a subsection (apache#6880)

trevor-m · Dec 4, 2020 · 6dae2b0 · 6dae2b0
1 parent 3288531
commit 6dae2b0
Show file tree

Hide file tree

Showing 10 changed files with 86 additions and 29 deletions.
diff --git a/docs/README.txt b/docs/README.txt
@@ -51,3 +51,8 @@ You will need a gpu CI environment.
 ```bash
 ./tests/scripts/task_python_docs.sh
 ```
+
+Define the Order of Tutorials
+-----------------------------
+You can define the order of tutorials with `conf.py::subsection_order` and `conf.py::within_subsection_order`.
+By default, the tutorials within one subsection is sorted by filename.
diff --git a/docs/conf.py b/docs/conf.py
@@ -204,6 +204,74 @@
     ]
 )
 
+# Explicitly define the order within a subsection.
+# The listed files are sorted according to the list.
+# The unlisted files are sorted by filenames.
+# The unlisted files always appear after listed files.
+within_subsection_order = {
+    "get_started": [
+        "relay_quick_start.py",
+        "tensor_expr_get_started.py",
+        "tvmc_command_line_driver.py",
+        "cross_compilation_and_rpc.py",
+    ],
+    "frontend": [
+        "from_pytorch.py",
+        "from_tensorflow.py",
+        "from_mxnet.py",
+        "from_onnx.py",
+        "from_keras.py",
+        "from_tflite.py",
+        "from_coreml.py",
+        "from_darknet.py",
+        "from_caffe2.py",
+    ],
+    "language": [
+        "schedule_primitives.py",
+        "reduciton.py",
+        "intrin_math.py",
+        "scan.py",
+        "extern_op.py",
+        "tensorize.py",
+        "tuple_inputs.py",
+        "tedd.py",
+    ],
+    "optimize": [
+        "opt_gemm.py",
+        "opt_conv_cuda.py",
+        "opt_conv_tensorcore.py",
+        "opt_matmul_auto_tensorcore.py",
+    ],
+    "autotvm": [
+        "tune_simple_template.py",
+        "tune_conv2d_cuda.py",
+        "tune_relay_cuda.py",
+        "tune_relay_x86.py",
+        "tune_relay_arm.py",
+        "tune_relay_mobile_gpu.py",
+    ],
+    "auto_scheduler": ["tune_matmul_x86.py", "tune_conv2d_layer_cuda.py"],
+}
+
+
+class WithinSubsectionOrder:
+    def __init__(self, src_dir):
+        self.src_dir = src_dir.split("/")[-1]
+
+    def __call__(self, filename):
+        # If the order is provided, use the provided order
+        if (
+            self.src_dir in within_subsection_order
+            and filename in within_subsection_order[self.src_dir]
+        ):
+            index = within_subsection_order[self.src_dir].index(filename)
+            assert index < 1e10
+            return "\0%010d" % index
+
+        # Otherwise, sort by filename
+        return filename
+
+
 sphinx_gallery_conf = {
     "backreferences_dir": "gen_modules/backreferences",
     "doc_module": ("tvm", "numpy"),
@@ -213,6 +281,7 @@
         "numpy": "https://numpy.org/doc/stable",
     },
     "examples_dirs": examples_dirs,
+    "within_subsection_order": WithinSubsectionOrder,
     "gallery_dirs": gallery_dirs,
     "subsection_order": subsection_order,
     "filename_pattern": os.environ.get("TVM_TUTORIAL_EXEC_PATTERN", ".py"),

diff --git a/tutorials/auto_scheduler/tune_conv2d_layer_cuda.py b/tutorials/auto_scheduler/tune_conv2d_layer_cuda.py
@@ -22,8 +22,7 @@
 **Author**: `Lianmin Zheng <https://github.com/merrymercy>`_, \
             `Chengfan Jia <https://github.com/jcf94/>`_
 
-
-Different from the existing :ref:`autotvm <tutorials-autotvm-sec>` which relies on 
+Different from the template-based :ref:`autotvm <tutorials-autotvm-sec>` which relies on
 manual templates to define the search space, the auto-scheduler does not require any templates.
 Users only need to write the computation declaration without any schedule commands or templates.
 The auto-scheduler can automatically generate a large search space and
@@ -182,7 +181,6 @@ def conv2d_layer(N, H, W, CO, CI, KH, KW, stride, padding):
 # and resume the status of search policy and cost model with the log file.
 # In the example below we resume the status and do more 5 trials.
 
-
 cost_model = auto_scheduler.XGBModel()
 cost_model.update_from_file(log_file)
 search_policy = auto_scheduler.SketchPolicy(

diff --git a/tutorials/auto_scheduler/tune_matmul_x86.py b/tutorials/auto_scheduler/tune_matmul_x86.py
@@ -20,7 +20,7 @@
 **Author**: `Lianmin Zheng <https://github.com/merrymercy>`_, \
             `Chengfan Jia <https://github.com/jcf94/>`_
 
-Different from the existing :ref:`autotvm <tutorials-autotvm-sec>` which relies on 
+Different from the template-based :ref:`autotvm <tutorials-autotvm-sec>` which relies on
 manual templates to define the search space, the auto-scheduler does not require any templates.
 Users only need to write the computation declaration without any schedule commands or templates.
 The auto-scheduler can automatically generate a large search space and

diff --git a/tutorials/autotvm/tune_conv2d_cuda.py b/tutorials/autotvm/tune_conv2d_cuda.py
@@ -53,8 +53,7 @@
 import numpy as np
 
 import tvm
-from tvm import te
-from tvm import topi
+from tvm import te, topi, testing
 from tvm.topi.testing import conv2d_nchw_python
 
 from tvm import autotvm

diff --git a/tutorials/autotvm/tune_relay_arm.py b/tutorials/autotvm/tune_relay_arm.py
@@ -66,9 +66,7 @@
 
 import numpy as np
 import tvm
-from tvm import te
-from tvm import autotvm
-from tvm import relay
+from tvm import relay, autotvm
 import tvm.relay.testing
 from tvm.autotvm.tuner import XGBTuner, GATuner, RandomTuner, GridSearchTuner
 from tvm.contrib.utils import tempdir
@@ -104,7 +102,7 @@ def get_network(name, batch_size):
             batch_size=batch_size, version="1.1", dtype=dtype
         )
     elif name == "inception_v3":
-        input_shape = (1, 3, 299, 299)
+        input_shape = (batch_size, 3, 299, 299)
         mod, params = relay.testing.inception_v3.get_workload(batch_size=batch_size, dtype=dtype)
     elif name == "mxnet":
         # an example for mxnet model

diff --git a/tutorials/autotvm/tune_relay_cuda.py b/tutorials/autotvm/tune_relay_cuda.py
@@ -64,12 +64,9 @@
 import numpy as np
 
 import tvm
-from tvm import te
-from tvm import autotvm
-from tvm import relay
+from tvm import relay, autotvm
 import tvm.relay.testing
 from tvm.autotvm.tuner import XGBTuner, GATuner, RandomTuner, GridSearchTuner
-from tvm.contrib.utils import tempdir
 import tvm.contrib.graph_runtime as runtime
 
 #################################################################
@@ -102,7 +99,7 @@ def get_network(name, batch_size):
             batch_size=batch_size, version="1.1", dtype=dtype
         )
     elif name == "inception_v3":
-        input_shape = (1, 3, 299, 299)
+        input_shape = (batch_size, 3, 299, 299)
         mod, params = relay.testing.inception_v3.get_workload(batch_size=batch_size, dtype=dtype)
     elif name == "mxnet":
         # an example for mxnet model
@@ -239,11 +236,6 @@ def tune_and_evaluate(tuning_opt):
         with tvm.transform.PassContext(opt_level=3):
             lib = relay.build_module.build(mod, target=target, params=params)
 
-        # export library
-        tmp = tempdir()
-        filename = "net.tar"
-        lib.export_library(tmp.relpath(filename))
-
         # load parameters
         ctx = tvm.context(str(target), 0)
         module = runtime.GraphModule(lib["default"](ctx))

diff --git a/tutorials/autotvm/tune_relay_mobile_gpu.py b/tutorials/autotvm/tune_relay_mobile_gpu.py
@@ -65,9 +65,7 @@
 import numpy as np
 
 import tvm
-from tvm import te
-from tvm import autotvm
-from tvm import relay
+from tvm import relay, autotvm
 import tvm.relay.testing
 from tvm.autotvm.tuner import XGBTuner, GATuner, RandomTuner, GridSearchTuner
 from tvm.contrib.utils import tempdir
@@ -103,7 +101,7 @@ def get_network(name, batch_size):
             batch_size=batch_size, version="1.1", dtype=dtype
         )
     elif name == "inception_v3":
-        input_shape = (1, 3, 299, 299)
+        input_shape = (batch_size, 3, 299, 299)
         mod, params = relay.testing.inception_v3.get_workload(batch_size=batch_size, dtype=dtype)
     elif name == "mxnet":
         # an example for mxnet model

diff --git a/tutorials/autotvm/tune_relay_x86.py b/tutorials/autotvm/tune_relay_x86.py
@@ -32,9 +32,7 @@
 import numpy as np
 
 import tvm
-from tvm import te
-from tvm import autotvm
-from tvm import relay
+from tvm import relay, autotvm
 from tvm.relay import testing
 from tvm.autotvm.tuner import XGBTuner, GATuner, RandomTuner, GridSearchTuner
 from tvm.autotvm.graph_tuner import DPTuner, PBQPTuner
@@ -73,7 +71,7 @@ def get_network(name, batch_size):
             batch_size=batch_size, version="1.1", dtype=dtype
         )
     elif name == "inception_v3":
-        input_shape = (1, 3, 299, 299)
+        input_shape = (batch_size, 3, 299, 299)
         mod, params = relay.testing.inception_v3.get_workload(batch_size=batch_size, dtype=dtype)
     elif name == "mxnet":
         # an example for mxnet model

diff --git a/tutorials/autotvm/tune_simple_template.py b/tutorials/autotvm/tune_simple_template.py
@@ -59,7 +59,7 @@
 
 import numpy as np
 import tvm
-from tvm import te
+from tvm import te, testing
 
 # the module is called `autotvm`
 from tvm import autotvm