Merge branch 'master' into 1.8_update

Author: narendasan

.gitignore

@@ -5,6 +5,7 @@ bazel-genfiles
 bazel-out
 bazel-testlogs
 bazel-TRTorch
+bazel-trtorch-testing
 third_party/pytorch
 *.jit
 *.jit.pt
@@ -37,4 +38,6 @@ bdist
 py/trtorch/_version.py
 py/wheelhouse
 py/.eggs
-notebooks/.ipynb_checkpoints/
+notebooks/.ipynb_checkpoints/
+*.cache
+tests/py/data

core/conversion/converters/impl/squeeze.cpp

@@ -25,6 +25,14 @@ auto squeeze_registrations TRTORCH_UNUSED = RegisterNodeConversionPatterns().pat
          dim = selfDim.size() + dim;
        }
 
+       if (selfDim[dim] != 1) {
+         auto out = ctx->AssociateValueAndTensor(n->outputs()[0], self);
+
+         LOG_DEBUG("Output tensor shape: " << out->getDimensions());
+
+         return true;
+       }
+
        auto shuffle_layer = ctx->net->addShuffle(*self);
        TRTORCH_CHECK(shuffle_layer, "Unable to create shuffle layer from node: " << *n);
        shuffle_layer->setReshapeDimensions(util::squeezeDims(self->getDimensions(), dim));

core/conversion/evaluators/aten.cpp

@@ -481,7 +481,61 @@ auto aten_registrations TRTORCH_UNUSED =
                       LOG_WARNING("Warning from TorchScript: " << *warning);
                       return {};
                     },
-                    EvalOptions()});
+                    EvalOptions()})
+        .evaluator({c10::Symbol::fromQualString("aten::arange"),
+                    [](const torch::jit::Node* n, kwargs& args) -> c10::optional<torch::jit::IValue> {
+                      int input_size = n->inputs().size();
+                      int scalar_count = 0;
+                      for (int i = 0; i < input_size; i++) {
+                        if (args.at(n->input(i)).IValue()->isScalar()) {
+                          scalar_count += 1;
+                        }
+                      }
+                      if (scalar_count == 1) {
+                        if (args.at(n->input(0)).IValue()->isInt()) {
+                          int end_scalar = args.at(n->input(0)).unwrapToInt();
+                          return torch::arange(end_scalar);
+                        } else if (args.at(n->input(0)).IValue()->isDouble()) {
+                          float end_scalar = args.at(n->input(0)).unwrapToScalar().to<float>();
+                          return torch::arange(end_scalar);
+                        }
+                      } else if (scalar_count == 2) {
+                        if (args.at(n->input(0)).IValue()->isDouble() || args.at(n->input(1)).IValue()->isDouble()) {
+                          float start_scalar = args.at(n->input(0)).unwrapToScalar().to<float>();
+                          float end_scalar = args.at(n->input(1)).unwrapToScalar().to<float>();
+                          return torch::arange(start_scalar, end_scalar);
+                        } else {
+                          int start_scalar = args.at(n->input(0)).unwrapToInt();
+                          int end_scalar = args.at(n->input(1)).unwrapToInt();
+                          return torch::arange(start_scalar, end_scalar);
+                        }
+                      } else if (scalar_count == 3) {
+                        if (args.at(n->input(0)).IValue()->isDouble() || args.at(n->input(1)).IValue()->isDouble() ||
+                            args.at(n->input(2)).IValue()->isDouble()) {
+                          float start_scalar = args.at(n->input(0)).unwrapToScalar().to<float>();
+                          float end_scalar = args.at(n->input(1)).unwrapToScalar().to<float>();
+                          float step_scalar = args.at(n->input(2)).unwrapToScalar().to<float>();
+                          return torch::arange(start_scalar, end_scalar, step_scalar);
+                        } else {
+                          int start_scalar = args.at(n->input(0)).unwrapToInt();
+                          int end_scalar = args.at(n->input(1)).unwrapToInt();
+                          int step_scalar = args.at(n->input(2)).unwrapToInt();
+                          return torch::arange(start_scalar, end_scalar, step_scalar);
+                        }
+                      } else {
+                        TRTORCH_THROW_ERROR(
+                            "Invalid input argument size for aten::arange, input argument size: " << input_size);
+                      }
+                      return {};
+                    },
+                    EvalOptions().validSchemas({
+                        R"SIG(aten::arange(Scalar end, *, int? dtype=None, int? layout=None,
+                            Device? device=None, bool? pin_memory=None) -> (Tensor))SIG",
+                        R"SIG(aten::arange.start(Scalar start, Scalar end, *, ScalarType? dtype=None,
+                            Layout? layout=None, Device? device=None, bool? pin_memory=None) -> (Tensor))SIG",
+                        R"SIG(aten::arange.start_step(Scalar start, Scalar end, Scalar step, *, ScalarType? dtype=None,
+                        Layout? layout=None, Device? device=None, bool? pin_memory=None) -> (Tensor))SIG",
+                    })});
 } // namespace
 } // namespace evaluators
 } // namespace conversion

docs/_notebooks/Resnet50-example.html

@@ -690,7 +690,7 @@
         </div>
        </div>
        <p>
-        <img alt="e694bc42c4f14beb8d4270227984ae7a" src="http://developer.download.nvidia.com/compute/machine-learning/frameworks/nvidia_logo.png"/>
+        <img alt="0904ed52551d45acab1d028795d47df1" src="http://developer.download.nvidia.com/compute/machine-learning/frameworks/nvidia_logo.png"/>
        </p>
        <h1 id="notebooks-resnet50-example--page-root">
         TRTorch Getting Started - ResNet 50

docs/_notebooks/lenet-getting-started.html

@@ -784,7 +784,7 @@
         </div>
        </div>
        <p>
-        <img alt="2b4b96299f2847c6a1beafa86b7dd5dc" src="http://developer.download.nvidia.com/compute/machine-learning/frameworks/nvidia_logo.png"/>
+        <img alt="eb83af0673044a89b9357bb998acd927" src="http://developer.download.nvidia.com/compute/machine-learning/frameworks/nvidia_logo.png"/>
        </p>
        <h1 id="notebooks-lenet-getting-started--page-root">
         TRTorch Getting Started - LeNet

docs/_notebooks/ssd-object-detection-demo.html

@@ -804,7 +804,7 @@
         </div>
        </div>
        <p>
-        <img alt="71b57a93fad34034a8a37d637b575dce" src="http://developer.download.nvidia.com/compute/machine-learning/frameworks/nvidia_logo.png"/>
+        <img alt="092b8936ef17484f92965a48c4cdea98" src="http://developer.download.nvidia.com/compute/machine-learning/frameworks/nvidia_logo.png"/>
        </p>
        <h1 id="notebooks-ssd-object-detection-demo--page-root">
         Object Detection with TRTorch (SSD)

docs/_sources/tutorials/ptq.rst.txt

@@ -14,14 +14,17 @@ the TensorRT calibrator. With TRTorch we look to leverage existing infrastructur
 calibrators easier.
 
 LibTorch provides a ``DataLoader`` and ``Dataset`` API which steamlines preprocessing and batching input data.
-This section of the PyTorch documentation has more information https://pytorch.org/tutorials/advanced/cpp_frontend.html#loading-data.
+These APIs are exposed via both C++ and Python interface which makes it easier for the end user.
+For C++ interface, we use ``torch::Dataset`` and ``torch::data::make_data_loader`` objects to construct and perform pre-processing on datasets.
+The equivalent functionality in python interface uses ``torch.utils.data.Dataset`` and ``torch.utils.data.DataLoader``.
+This section of the PyTorch documentation has more information https://pytorch.org/tutorials/advanced/cpp_frontend.html#loading-data and https://pytorch.org/tutorials/recipes/recipes/loading_data_recipe.html.
 TRTorch uses Dataloaders as the base of a generic calibrator implementation. So you will be able to reuse or quickly
 implement a ``torch::Dataset`` for your target domain, place it in a DataLoader and create a INT8 Calibrator
 which you can provide to TRTorch to run INT8 Calibration during compliation of your module.
 
-.. _writing_ptq:
+.. _writing_ptq_cpp:
 
-How to create your own PTQ application
+How to create your own PTQ application in C++
 ----------------------------------------
 
 Here is an example interface of a ``torch::Dataset`` class for CIFAR10:
@@ -132,14 +135,76 @@ Then all thats required to setup the module for INT8 calibration is to set the f
     auto trt_mod = trtorch::CompileGraph(mod, compile_spec);
 
 If you have an existing Calibrator implementation for TensorRT you may directly set the ``ptq_calibrator`` field with a pointer to your calibrator and it will work as well.
-
 From here not much changes in terms of how to execution works. You are still able to fully use LibTorch as the sole interface for inference. Data should remain
 in FP32 precision when it's passed into `trt_mod.forward`. There exists an example application in the TRTorch demo that takes you from training a VGG16 network on
 CIFAR10 to deploying in INT8 with TRTorch here: https://github.com/NVIDIA/TRTorch/tree/master/cpp/ptq
 
+.. _writing_ptq_python:
+
+How to create your own PTQ application in Python
+----------------------------------------
+
+TRTorch Python API provides an easy and convenient way to use pytorch dataloaders with TensorRT calibrators. ``DataLoaderCalibrator`` class can be used to create
+a TensorRT calibrator by providing desired configuration. The following code demonstrates an example on how to use it
+
+.. code-block:: python
+
+    testing_dataset = torchvision.datasets.CIFAR10(root='./data',
+                                                            train=False,
+                                                            download=True,
+                                                            transform=transforms.Compose([
+                                                                transforms.ToTensor(),
+                                                                transforms.Normalize((0.4914, 0.4822, 0.4465),
+                                                                                     (0.2023, 0.1994, 0.2010))
+                                                            ]))
+
+    testing_dataloader = torch.utils.data.DataLoader(testing_dataset,
+                                                          batch_size=1,
+                                                          shuffle=False,
+                                                          num_workers=1)
+    calibrator = trtorch.ptq.DataLoaderCalibrator(testing_dataloader,
+                                                  cache_file='./calibration.cache',
+                                                  use_cache=False,
+                                                  algo_type=trtorch.ptq.CalibrationAlgo.ENTROPY_CALIBRATION_2,
+                                                  device=torch.device('cuda:0'))
+
+    compile_spec = {
+             "input_shapes": [[1, 3, 32, 32]],
+             "op_precision": torch.int8,
+             "calibrator": calibrator,
+             "device": {
+                 "device_type": trtorch.DeviceType.GPU,
+                 "gpu_id": 0,
+                 "dla_core": 0,
+                 "allow_gpu_fallback": False,
+                 "disable_tf32": False
+             }
+         }
+    trt_mod = trtorch.compile(model, compile_spec)
+
+In the cases where there is a pre-existing calibration cache file that users want to use, ``CacheCalibrator`` can be used without any dataloaders. The following example demonstrates how
+to use ``CacheCalibrator`` to use in INT8 mode.
+
+.. code-block:: python
+
+  calibrator = trtorch.ptq.CacheCalibrator("./calibration.cache")
+
+  compile_settings = {
+        "input_shapes": [[1, 3, 32, 32]],
+        "op_precision": torch.int8,
+        "calibrator": calibrator,
+        "max_batch_size": 32,
+    }
+
+  trt_mod = trtorch.compile(model, compile_settings)
+
+If you already have an existing calibrator class (implemented directly using TensorRT API), you can directly set the calibrator field to your class which can be very convenient.
+For a demo on how PTQ can be performed on a VGG network using TRTorch API, you can refer to https://github.com/NVIDIA/TRTorch/blob/master/tests/py/test_ptq_dataloader_calibrator.py
+and https://github.com/NVIDIA/TRTorch/blob/master/tests/py/test_ptq_trt_calibrator.py
+
 Citations
 ^^^^^^^^^^^
 
 Krizhevsky, A., & Hinton, G. (2009). Learning multiple layers of features from tiny images.
 
-Simonyan, K., & Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556.
+Simonyan, K., & Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556.

docs/_static/pygments.css

@@ -432,7 +432,7 @@ <h1 id="py-api-logging--page-root">
             Debug
            </code>
            <em class="property">
-            = LogLevel.DEBUG
+            = &lt;LogLevel.DEBUG: 4&gt;
            </em>
            <a class="headerlink" href="#trtorch.logging.Level.Debug" title="Permalink to this definition">
             ¶
@@ -447,7 +447,7 @@ <h1 id="py-api-logging--page-root">
             Error
            </code>
            <em class="property">
-            = LogLevel.ERROR
+            = &lt;LogLevel.ERROR: 1&gt;
            </em>
            <a class="headerlink" href="#trtorch.logging.Level.Error" title="Permalink to this definition">
             ¶
@@ -462,7 +462,7 @@ <h1 id="py-api-logging--page-root">
             Info
            </code>
            <em class="property">
-            = LogLevel.INFO
+            = &lt;LogLevel.INFO: 3&gt;
            </em>
            <a class="headerlink" href="#trtorch.logging.Level.Info" title="Permalink to this definition">
             ¶
@@ -477,7 +477,7 @@ <h1 id="py-api-logging--page-root">
             InternalError
            </code>
            <em class="property">
-            = LogLevel.INTERNAL_ERROR
+            = &lt;LogLevel.INTERNAL_ERROR: 0&gt;
            </em>
            <a class="headerlink" href="#trtorch.logging.Level.InternalError" title="Permalink to this definition">
             ¶
@@ -492,7 +492,7 @@ <h1 id="py-api-logging--page-root">
             Warning
            </code>
            <em class="property">
-            = LogLevel.WARNING
+            = &lt;LogLevel.WARNING: 2&gt;
            </em>
            <a class="headerlink" href="#trtorch.logging.Level.Warning" title="Permalink to this definition">
             ¶