You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 9 Next »

(authored by Kellen Sunderland)


This document described how to use the MXNet-TensorRT runtime integration to accelerate model inference. 

Why is TensorRT integration useful?

TensorRT can greatly speed up inference of deep learning models. One experiment on a Titan V (V100) GPU shows that with MXNet 1.2, we can get an approximately 3x speed-up when running inference of the ResNet-50 model on the CIFAR-10 dataset in single precision (fp32). As batch sizes and image sizes go up (for CNN inference), the benefit may be less, but in general, TensorRT helps especially in cases which have:

  • Many bandwidth-bound layers (e.g. pointwise operations) that benefit from GPU kernel fusion.
  • Inference use cases which have tight latency requirements and where the client application can't wait for large batches to be queued up.
  • Embedded systems, where memory constraints are tighter than on servers.
  • When performing inference in reduced precision, especially for integer (e.g. int8) inference.

In the past, the main hindrance for the user wishing to benefit from TensorRT was the fact that the model needed to be exported from the framework first. Once the model got exported through some means (NNVM to TensorRT graph rewrite, via ONNX, etc.), one had to then write a TensorRT client application, which would feed the data into the TensorRT engine. Since at that point the model was independent of the original framework, and since TensorRT could only compute the neural network layers but the user had to bring their own data pipeline, this increased the burden on the user and reduced the likelihood of reproducibility (e.g. different frameworks may have slightly different data pipelines, or flexibility of data pipeline operation ordering). Moreover, since frameworks typically support more operators than TensorRT, one could have to resort to TensorRT plugins for operations that aren't already available via the TensorRT graph API.

The current experimental runtime integration of TensorRT with MXNet resolves the above concerns by ensuring that:

  • The graph is still executed by MXNet.
  • The MXNet data pipeline is preserved.
  • The TensorRT runtime integration logic partitions the graph into subgraphs that are either TensorRT compatible or incompatible.
  • The graph partitioner collects the TensorRT-compatible subgraphs, hands them over to TensorRT, and substitutes the TensorRT compatible subgraph with a TensorRT library call, represented as a TensorRT node in NNVM.
  • If a node is not TensorRT compatible, it won't be extracted and substituted with a TensorRT call, and will still execute within MXNet.

The above points ensure that we find a compromise between the flexibility of MXNet, and fast inference in TensorRT. We do this with no additional burden to the user. Users do not need to learn how TensorRT APIs work, and do not need to write their own client application or data pipeline.

How do I build MXNet with TensorRT integration?

Building MXNet together with TensorRT is somewhat complex. The recipe will hopefully be simplified in the near future, but for now, it's easiest to build a Docker container with a Ubuntu 16.04 base. This Dockerfile can be found under the ci subdirectory of the MXNet repository. You can build the container as follows:

docker build -t ci/docker/Dockerfile.build.ubuntu_gpu_tensorrt mxnet_with_tensorrt


Next, we can run this container as follows (don't forget to install nvidia-docker):
nvidia-docker run -ti --rm mxnet_with_tensorrt

After starting the container, you will find yourself in the /opt/mxnet directory by default.

Running a "hello, world" model / unit test (LeNet-5 on MNIST)

You can then run the LeNet-5 unit test, which will train LeNet-5 on MNIST using the symbolic API. The test will then run inference in MXNet both with, and without MXNet-TensorRT runtime integration. Finally, the test will display a comparison of both runtime's accuracy scores. The test can be run as follows:

python ${MXNET_HOME}/tests/python/tensorrt/test_tensorrt_lenet5.py

You should get a result similar to the following:

Running inference in MXNet
[03:31:18] src/operator/nn/./cudnn/./cudnn_algoreg-inl.h:107: Running performance tests to find the best convolution algorithm, this can take a while... (setting env variable MXNET_CUDNN_AUTOTUNE_DEFAULT to 0 to disable)
Running inference in MXNet-TensorRT
MXNet accuracy: 98.680000
MXNet-TensorRT accuracy: 98.680000

Running more complex models

The unit test directory also provides a way to run models from the Gluon model zoo after slight modifications. The models that are tested are CNN classification models from the Gluon zoo. They are mostly based on ResNet, but include ResNeXtas well:

  • cifar_resnet20_v1
  • cifar_resnet56_v1
  • cifar_resnet110_v1
  • cifar_resnet20_v2
  • cifar_resnet56_v2
  • cifar_resnet110_v2
  • cifar_wideresnet16_10
  • cifar_wideresnet28_10
  • cifar_wideresnet40_8
  • cifar_resnext29_16x64d

Please note that even those examples are based on CIFAR-10 due to the ease of accessing the dataset without formal registration and preprocessing, everything should work fine with models trained on ImageNet, using MXNet's ImageNet iterators, based on the RecordIO representation of the ImageNet dataset.

The script can be run simply as

python ${MXNET_HOME}/tests/python/tensorrt/test_tensorrt_resnet_resnext.py

Here's some sample output, for inference with batch size 16 (TensorRT is especially useful for small batches for low-latency production inference):

===========================================
Model: cifar_resnet56_v1
===========================================

*** Running inference using pure MXNet ***
MXNet: time elapsed: 2.463s, accuracy: 94.19%

*** Running inference using MXNet + TensorRT ***
TensorRT: time elapsed: 1.652s, accuracy: 94.19%

TensorRT speed-up (not counting compilation): 1.49x
Absolute accuracy difference: 0.000000


===========================================
Model: cifar_resnet110_v1
===========================================

*** Running inference using pure MXNet ***
MXNet: time elapsed: 4.000s, accuracy: 95.20%

*** Running inference using MXNet + TensorRT ***

TensorRT: time elapsed: 2.085s, accuracy: 95.20%

TensorRT speed-up (not counting compilation): 1.92x
Absolute accuracy difference: 0.000000

As you can see, the speed-up varies by model. ResNet-110 has more layers that can be fused than ResNet-56, hence the speed-up is greater.

Running TensorRT with your own models with the symbolic API

When building your own models, feel free to use the above ResNet-50 model as an example. Here, we highlight a small number of issues that need to be taken into account.

1. When loading a pre-trained model, the inference will be handled using the Symbol API, rather than the Module API.

2. In order to provide the weights from MXNet (NNVM) to the TensorRT graph converter before the symbol is fully bound (before the memory is allocated, etc.), the arg_params and aux_params need to be provided to the symbol's simple_bind method. The weights and other values (e.g. moments learned from data by batch normalization, provided via aux_params) will be provided via the shared_buffer argument to simple_bind as follows:

executor = sym.simple_bind(ctx=ctx, data = data_shape,
softmax_label=sm_shape, grad_req='null', shared_buffer=all_params, force_rebind=True)

3. To collect arg_params and aux_params from the dictionaries loaded by model.load(), we need to combine them into one dictionary:

def merge_dicts(*dict_args):
    result = {}
    for dictionary in dict_args:
        result.update(dictionary)
        return result

sym, arg_params, aux_params = mx.model.load_checkpoint(model_prefix, epoch)
all_params = merge_dicts(arg_params, aux_params)

This all_params dictionary can be seen in use in the simple_bind call in #2. 4. Once the symbol is bound, we need to feed the data and run the forward() method. Let's say we're using a test set data iterator called test_iter. We can run inference as follows:

for idx, dbatch in enumerate(test_iter):
    data = dbatch.data[0]
    executor.arg_dict["data"][:] = data
    executor.forward(is_train=False)
    preds = executor.outputs[0].asnumpy() 
    top1 = np.argmax(preds, axis=1)

5. Note: One can choose between running inference with and without TensorRT. This can be selected by changing the state of the MXNET_USE_TENSORRT environment variable. Let's first write a convenience function to change the state of this environment variable:

def set_use_tensorrt(status = False):
        os.environ["MXNET_USE_TENSORRT"] = str(int(status))

Now, assuming that the logic to bind a symbol and run inference in batches of batch_size on dataset dataset is wrapped in the run_inference function, we can do the following:

print("Running inference in MXNet")
set_use_tensorrt(False)
mx_pct = run_inference(sym, arg_params, aux_params, mnist, all_test_labels, batch_size=batch_size)

print("Running inference in MXNet-TensorRT")
set_use_tensorrt(True)
trt_pct = run_inference(sym, arg_params, aux_params, mnist, all_test_labels,  batch_size=batch_size)

Simply switching the flag allows us to go back and forth between MXNet and MXNet-TensorRT inference. See the details in the unit test at ${MXNET_HOME}/tests/python/tensorrt/test_tensorrt_lenet5.py.

Running TensorRT with your own models with the Gluon API

Note: Please first read the previous section titled "Running TensorRT with your own models with the symbolic API" - it contains information that will also be useful for Gluonusers.

Note: If the user wishes to use the Gluon vision models, it's necessary to install the gluoncv pip package:

pip install gluoncv

The above package is based on a separate repository.

For Gluon models specifically, we need to add a data symbol to the model to load the data, as well as apply the softmax layer, because the Gluon models only present the logits that are to be presented for softmax. This is shown in python ${MXNET_HOME}/tests/python/tensorrt/test_tensorrt_resnet_resnext.py. Here's the relevant code:

net = gluoncv.model_zoo.get_model(model_name, pretrained=True)
data = mx.sym.var('data')
out = net(data)
softmax = mx.sym.SoftmaxOutput(out, name='softmax')

Since as in the symbolic API case, we need to provide the weights during the simple_bind call, we need to extract them. The Gluon symbol allows very easy access to the weights - we can extract them directly from the network object, and then provide them during the simple_bind call:

net = gluoncv.model_zoo.get_model(model_name, pretrained=True)
all_params = dict([(k, v.data()) for k, v in net.collect_params().items()])
executor = softmax.simple_bind(ctx=ctx, data=(batch_size, 3, 32, 32), softmax_label=(batch_size,), grad_req='null',
                                   shared_buffer=all_params, force_rebind=True)

Note that for Gluon-trained models, we should use Gluon's data pipeline to replicate the behavior of the pipeline that was used for training (e.g. using the same data scaling). Here's how to get the Gluon data iterator for the CIFAR-10 examples:

gluon.data.DataLoader(
        gluon.data.vision.CIFAR10(train=False).transform_first(transform_test),
        batch_size=batch_size, shuffle=False, num_workers=num_workers)

For more details, see the unit test examples at ${MXNET_HOME}/tests/python/tensorrt/test_tensorrt_resnet_resnext.py.

Examples

Image classification examples are available here.



  • No labels