.. DO NOT EDIT. .. THIS FILE WAS AUTOMATICALLY GENERATED BY SPHINX-GALLERY. .. TO MAKE CHANGES, EDIT THE SOURCE PYTHON FILE: .. "tutorial/autotvm_relay_x86.py" .. LINE NUMBERS ARE GIVEN BELOW. .. only:: html .. note:: :class: sphx-glr-download-link-note Click :ref:`here ` to download the full example code .. rst-class:: sphx-glr-example-title .. _sphx_glr_tutorial_autotvm_relay_x86.py: Compiling and Optimizing a Model with the Python Interface (AutoTVM) ==================================================================== **Author**: `Chris Hoge `_ In the `TVMC Tutorial `_, we covered how to compile, run, and tune a pre-trained vision model, ResNet-50 v2 using the command line interface for TVM, TVMC. TVM is more that just a command-line tool though, it is an optimizing framework with APIs available for a number of different languages that gives you tremendous flexibility in working with machine learning models. In this tutorial we will cover the same ground we did with TVMC, but show how it is done with the Python API. Upon completion of this section, we will have used the Python API for TVM to accomplish the following tasks: * Compile a pre-trained ResNet-50 v2 model for the TVM runtime. * Run a real image through the compiled model, and interpret the output and model performance. * Tune the model that model on a CPU using TVM. * Re-compile an optimized model using the tuning data collected by TVM. * Run the image through the optimized model, and compare the output and model performance. The goal of this section is to give you an overview of TVM's capabilites and how to use them through the Python API. .. GENERATED FROM PYTHON SOURCE LINES 44-46 .. code-block:: default .. GENERATED FROM PYTHON SOURCE LINES 52-62 TVM is a deep learning compiler framework, with a number of different modules available for working with deep learning models and operators. In this tutorial we will work through how to load, compile, and optimize a model using the Python API. We begin by importing a number of dependencies, including ``onnx`` for loading and converting the model, helper utilities for downloading test data, the Python Image Library for working with the image data, ``numpy`` for pre and post-processing of the image data, the TVM Relay framework, and the TVM Graph Executor. .. GENERATED FROM PYTHON SOURCE LINES 62-71 .. code-block:: default import onnx from tvm.contrib.download import download_testdata from PIL import Image import numpy as np import tvm.relay as relay import tvm from tvm.contrib import graph_executor .. GENERATED FROM PYTHON SOURCE LINES 72-93 Downloading and Loading the ONNX Model -------------------------------------- For this tutorial, we will be working with ResNet-50 v2. ResNet-50 is a convolutional neural network that is 50 layers deep and designed to classify images. The model we will be using has been pre-trained on more than a million images with 1000 different classifications. The network has an input image size of 224x224. If you are interested exploring more of how the ResNet-50 model is structured, we recommend downloading `Netron `_, a freely available ML model viewer. TVM provides a helper library to download pre-trained models. By providing a model URL, file name, and model type through the module, TVM will download the model and save it to disk. For the instance of an ONNX model, you can then load it into memory using the ONNX runtime. .. admonition:: Working with Other Model Formats TVM supports many popular model formats. A list can be found in the :ref:`Compile Deep Learning Models ` section of the TVM Documentation. .. GENERATED FROM PYTHON SOURCE LINES 93-106 .. code-block:: default model_url = ( "https://github.com/onnx/models/raw/main/" "vision/classification/resnet/model/" "resnet50-v2-7.onnx" ) model_path = download_testdata(model_url, "resnet50-v2-7.onnx", module="onnx") onnx_model = onnx.load(model_path) # Seed numpy's RNG to get consistent results np.random.seed(0) .. GENERATED FROM PYTHON SOURCE LINES 107-124 Downloading, Preprocessing, and Loading the Test Image ------------------------------------------------------ Each model is particular when it comes to expected tensor shapes, formats and data types. For this reason, most models require some pre and post-processing, to ensure the input is valid and to interpret the output. TVMC has adopted NumPy's ``.npz`` format for both input and output data. As input for this tutorial, we will use the image of a cat, but you can feel free to substitute this image for any of your choosing. .. image:: https://s3.amazonaws.com/model-server/inputs/kitten.jpg :height: 224px :width: 224px :align: center Download the image data, then convert it to a numpy array to use as an input to the model. .. GENERATED FROM PYTHON SOURCE LINES 124-143 .. code-block:: default img_url = "https://s3.amazonaws.com/model-server/inputs/kitten.jpg" img_path = download_testdata(img_url, "imagenet_cat.png", module="data") # Resize it to 224x224 resized_image = Image.open(img_path).resize((224, 224)) img_data = np.asarray(resized_image).astype("float32") # Our input image is in HWC layout while ONNX expects CHW input, so convert the array img_data = np.transpose(img_data, (2, 0, 1)) # Normalize according to the ImageNet input specification imagenet_mean = np.array([0.485, 0.456, 0.406]).reshape((3, 1, 1)) imagenet_stddev = np.array([0.229, 0.224, 0.225]).reshape((3, 1, 1)) norm_img_data = (img_data / 255 - imagenet_mean) / imagenet_stddev # Add the batch dimension, as we are expecting 4-dimensional input: NCHW. img_data = np.expand_dims(norm_img_data, axis=0) .. GENERATED FROM PYTHON SOURCE LINES 144-151 Compile the Model With Relay ---------------------------- The next step is to compile the ResNet model. We begin by importing the model to relay using the `from_onnx` importer. We then build the model, with standard optimizations, into a TVM library. Finally, we create a TVM graph runtime module from the library. .. GENERATED FROM PYTHON SOURCE LINES 151-154 .. code-block:: default target = "llvm" .. GENERATED FROM PYTHON SOURCE LINES 155-167 .. admonition:: Defining the Correct Target Specifying the correct target can have a huge impact on the performance of the compiled module, as it can take advantage of hardware features available on the target. For more information, please refer to :ref:`Auto-tuning a convolutional network for x86 CPU `. We recommend identifying which CPU you are running, along with optional features, and set the target appropriately. For example, for some processors ``target = "llvm -mcpu=skylake"``, or ``target = "llvm -mcpu=skylake-avx512"`` for processors with the AVX-512 vector instruction set. .. GENERATED FROM PYTHON SOURCE LINES 167-181 .. code-block:: default # The input name may vary across model types. You can use a tool # like Netron to check input names input_name = "data" shape_dict = {input_name: img_data.shape} mod, params = relay.frontend.from_onnx(onnx_model, shape_dict) with tvm.transform.PassContext(opt_level=3): lib = relay.build(mod, target=target, params=params) dev = tvm.device(str(target), 0) module = graph_executor.GraphModule(lib["default"](dev)) .. GENERATED FROM PYTHON SOURCE LINES 182-190 Execute on the TVM Runtime -------------------------- Now that we've compiled the model, we can use the TVM runtime to make predictions with it. To use TVM to run the model and make predictions, we need two things: - The compiled model, which we just produced. - Valid input to the model to make predictions on. .. GENERATED FROM PYTHON SOURCE LINES 190-197 .. code-block:: default dtype = "float32" module.set_input(input_name, img_data) module.run() output_shape = (1, 1000) tvm_output = module.get_output(0, tvm.nd.empty(output_shape)).numpy() .. GENERATED FROM PYTHON SOURCE LINES 198-205 Collect Basic Performance Data ------------------------------ We want to collect some basic performance data associated with this unoptimized model and compare it to a tuned model later. To help account for CPU noise, we run the computation in multiple batches in multiple repetitions, then gather some basis statistics on the mean, median, and standard deviation. .. GENERATED FROM PYTHON SOURCE LINES 205-222 .. code-block:: default import timeit timing_number = 10 timing_repeat = 10 unoptimized = ( np.array(timeit.Timer(lambda: module.run()).repeat(repeat=timing_repeat, number=timing_number)) * 1000 / timing_number ) unoptimized = { "mean": np.mean(unoptimized), "median": np.median(unoptimized), "std": np.std(unoptimized), } print(unoptimized) .. rst-class:: sphx-glr-script-out .. code-block:: none {'mean': 512.4081050999996, 'median': 511.8093637999948, 'std': 1.8920776301187912} .. GENERATED FROM PYTHON SOURCE LINES 223-232 Postprocess the output ---------------------- As previously mentioned, each model will have its own particular way of providing output tensors. In our case, we need to run some post-processing to render the outputs from ResNet-50 v2 into a more human-readable form, using the lookup-table provided for the model. .. GENERATED FROM PYTHON SOURCE LINES 232-249 .. code-block:: default from scipy.special import softmax # Download a list of labels labels_url = "https://s3.amazonaws.com/onnx-model-zoo/synset.txt" labels_path = download_testdata(labels_url, "synset.txt", module="data") with open(labels_path, "r") as f: labels = [l.rstrip() for l in f] # Open the output and read the output tensor scores = softmax(tvm_output) scores = np.squeeze(scores) ranks = np.argsort(scores)[::-1] for rank in ranks[0:5]: print("class='%s' with probability=%f" % (labels[rank], scores[rank])) .. rst-class:: sphx-glr-script-out .. code-block:: none class='n02123045 tabby, tabby cat' with probability=0.621103 class='n02123159 tiger cat' with probability=0.356379 class='n02124075 Egyptian cat' with probability=0.019712 class='n02129604 tiger, Panthera tigris' with probability=0.001215 class='n04040759 radiator' with probability=0.000262 .. GENERATED FROM PYTHON SOURCE LINES 250-259 This should produce the following output: .. code-block:: bash # class='n02123045 tabby, tabby cat' with probability=0.610553 # class='n02123159 tiger cat' with probability=0.367179 # class='n02124075 Egyptian cat' with probability=0.019365 # class='n02129604 tiger, Panthera tigris' with probability=0.001273 # class='n04040759 radiator' with probability=0.000261 .. GENERATED FROM PYTHON SOURCE LINES 261-283 Tune the model -------------- The previous model was compiled to work on the TVM runtime, but did not include any platform specific optimization. In this section, we will show you how to build an optimized model using TVM to target your working platform. In some cases, we might not get the expected performance when running inferences using our compiled module. In cases like this, we can make use of the auto-tuner, to find a better configuration for our model and get a boost in performance. Tuning in TVM refers to the process by which a model is optimized to run faster on a given target. This differs from training or fine-tuning in that it does not affect the accuracy of the model, but only the runtime performance. As part of the tuning process, TVM will try running many different operator implementation variants to see which perform best. The results of these runs are stored in a tuning records file. In the simplest form, tuning requires you to provide three things: - the target specification of the device you intend to run this model on - the path to an output file in which the tuning records will be stored - a path to the model to be tuned. .. GENERATED FROM PYTHON SOURCE LINES 283-288 .. code-block:: default import tvm.auto_scheduler as auto_scheduler from tvm.autotvm.tuner import XGBTuner from tvm import autotvm .. GENERATED FROM PYTHON SOURCE LINES 289-299 Set up some basic parameters for the runner. The runner takes compiled code that is generated with a specific set of parameters and measures the performance of it. ``number`` specifies the number of different configurations that we will test, while ``repeat`` specifies how many measurements we will take of each configuration. ``min_repeat_ms`` is a value that specifies how long need to run configuration test. If the number of repeats falls under this time, it will be increased. This option is necessary for accurate tuning on GPUs, and is not required for CPU tuning. Setting this value to 0 disables it. The ``timeout`` places an upper limit on how long to run training code for each tested configuration. .. GENERATED FROM PYTHON SOURCE LINES 299-314 .. code-block:: default number = 10 repeat = 1 min_repeat_ms = 0 # since we're tuning on a CPU, can be set to 0 timeout = 10 # in seconds # create a TVM runner runner = autotvm.LocalRunner( number=number, repeat=repeat, timeout=timeout, min_repeat_ms=min_repeat_ms, enable_cpu_cache_flush=True, ) .. GENERATED FROM PYTHON SOURCE LINES 315-329 Create a simple structure for holding tuning options. We use an XGBoost algorithim for guiding the search. For a production job, you will want to set the number of trials to be larger than the value of 20 used here. For CPU we recommend 1500, for GPU 3000-4000. The number of trials required can depend on the particular model and processor, so it's worth spending some time evaluating performance across a range of values to find the best balance between tuning time and model optimization. Because running tuning is time intensive we set number of trials to 10, but do not recommend a value this small. The ``early_stopping`` parameter is the minimum number of trails to run before a condition that stops the search early can be applied. The measure option indicates where trial code will be built, and where it will be run. In this case, we're using the ``LocalRunner`` we just created and a ``LocalBuilder``. The ``tuning_records`` option specifies a file to write the tuning data to. .. GENERATED FROM PYTHON SOURCE LINES 329-340 .. code-block:: default tuning_option = { "tuner": "xgb", "trials": 20, "early_stopping": 100, "measure_option": autotvm.measure_option( builder=autotvm.LocalBuilder(build_func="default"), runner=runner ), "tuning_records": "resnet-50-v2-autotuning.json", } .. GENERATED FROM PYTHON SOURCE LINES 341-346 .. admonition:: Defining the Tuning Search Algorithm By default this search is guided using an `XGBoost Grid` algorithm. Depending on your model complexity and amount of time available, you might want to choose a different algorithm. .. GENERATED FROM PYTHON SOURCE LINES 349-356 .. admonition:: Setting Tuning Parameters In this example, in the interest of time, we set the number of trials and early stopping to 10. You will likely see more performance improvements if you set these values to be higher but this comes at the expense of time spent tuning. The number of trials required for convergence will vary depending on the specifics of the model and the target platform. .. GENERATED FROM PYTHON SOURCE LINES 356-374 .. code-block:: default # begin by extracting the tasks from the onnx model tasks = autotvm.task.extract_from_program(mod["main"], target=target, params=params) # Tune the extracted tasks sequentially. for i, task in enumerate(tasks): prefix = "[Task %2d/%2d] " % (i + 1, len(tasks)) tuner_obj = XGBTuner(task, loss_type="rank") tuner_obj.tune( n_trial=min(tuning_option["trials"], len(task.config_space)), early_stopping=tuning_option["early_stopping"], measure_option=tuning_option["measure_option"], callbacks=[ autotvm.callback.progress_bar(tuning_option["trials"], prefix=prefix), autotvm.callback.log_to_file(tuning_option["tuning_records"]), ], ) .. rst-class:: sphx-glr-script-out .. code-block:: none [Task 1/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 1/25] Current/Best: 17.58/ 17.58 GFLOPS | Progress: (4/20) | 6.34 s [Task 1/25] Current/Best: 6.09/ 17.58 GFLOPS | Progress: (8/20) | 9.43 s [Task 1/25] Current/Best: 11.01/ 22.33 GFLOPS | Progress: (12/20) | 11.98 s [Task 1/25] Current/Best: 16.48/ 22.33 GFLOPS | Progress: (16/20) | 13.68 s [Task 1/25] Current/Best: 11.24/ 23.66 GFLOPS | Progress: (20/20) | 15.47 s Done. [Task 2/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 2/25] Current/Best: 12.08/ 12.53 GFLOPS | Progress: (4/20) | 3.89 s [Task 2/25] Current/Best: 12.51/ 17.86 GFLOPS | Progress: (8/20) | 5.22 s [Task 2/25] Current/Best: 20.84/ 20.84 GFLOPS | Progress: (12/20) | 6.54 s [Task 2/25] Current/Best: 11.29/ 20.84 GFLOPS | Progress: (16/20) | 7.80 s [Task 2/25] Current/Best: 17.55/ 20.84 GFLOPS | Progress: (20/20) | 9.41 s Done. [Task 3/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 3/25] Current/Best: 1.63/ 9.86 GFLOPS | Progress: (4/20) | 5.86 s [Task 3/25] Current/Best: 15.38/ 16.88 GFLOPS | Progress: (8/20) | 7.83 s [Task 3/25] Current/Best: 15.06/ 16.88 GFLOPS | Progress: (12/20) | 9.55 s [Task 3/25] Current/Best: 6.75/ 23.47 GFLOPS | Progress: (16/20) | 11.50 s [Task 3/25] Current/Best: 11.10/ 23.47 GFLOPS | Progress: (20/20) | 16.15 s Done. [Task 4/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 4/25] Current/Best: 8.74/ 18.47 GFLOPS | Progress: (4/20) | 2.42 s [Task 4/25] Current/Best: 6.27/ 18.47 GFLOPS | Progress: (8/20) | 7.07 s [Task 4/25] Current/Best: 20.57/ 20.57 GFLOPS | Progress: (12/20) | 11.99 s [Task 4/25] Current/Best: 15.45/ 20.57 GFLOPS | Progress: (16/20) | 14.34 s [Task 4/25] Current/Best: 12.43/ 20.57 GFLOPS | Progress: (20/20) | 16.41 s Done. [Task 5/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 5/25] Current/Best: 9.18/ 9.64 GFLOPS | Progress: (4/20) | 2.60 s [Task 5/25] Current/Best: 11.46/ 11.46 GFLOPS | Progress: (8/20) | 4.69 s [Task 5/25] Current/Best: 11.86/ 17.99 GFLOPS | Progress: (12/20) | 7.73 s [Task 5/25] Current/Best: 11.42/ 22.57 GFLOPS | Progress: (16/20) | 9.19 s [Task 5/25] Current/Best: 12.08/ 22.57 GFLOPS | Progress: (20/20) | 11.14 s Done. [Task 6/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 6/25] Current/Best: 11.97/ 19.51 GFLOPS | Progress: (4/20) | 4.13 s [Task 6/25] Current/Best: 18.90/ 19.51 GFLOPS | Progress: (8/20) | 5.90 s [Task 6/25] Current/Best: 13.21/ 19.51 GFLOPS | Progress: (12/20) | 7.91 s [Task 6/25] Current/Best: 19.09/ 19.51 GFLOPS | Progress: (16/20) | 10.15 s [Task 6/25] Current/Best: 3.75/ 19.51 GFLOPS | Progress: (20/20) | 12.75 s Done. [Task 7/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 7/25] Current/Best: 9.77/ 12.11 GFLOPS | Progress: (4/20) | 3.62 s [Task 7/25] Current/Best: 19.29/ 19.89 GFLOPS | Progress: (8/20) | 5.15 s [Task 7/25] Current/Best: 16.17/ 19.89 GFLOPS | Progress: (12/20) | 7.05 s [Task 7/25] Current/Best: 12.15/ 20.21 GFLOPS | Progress: (16/20) | 9.11 s [Task 7/25] Current/Best: 6.12/ 20.58 GFLOPS | Progress: (20/20) | 11.62 s Done. [Task 8/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 8/25] Current/Best: 9.86/ 13.29 GFLOPS | Progress: (4/20) | 2.97 s [Task 8/25] Current/Best: 9.63/ 13.29 GFLOPS | Progress: (8/20) | 8.06 s [Task 8/25] Current/Best: 12.86/ 13.29 GFLOPS | Progress: (12/20) | 14.49 s [Task 8/25] Current/Best: 18.76/ 18.76 GFLOPS | Progress: (16/20) | 16.61 s [Task 8/25] Current/Best: 18.17/ 18.76 GFLOPS | Progress: (20/20) | 23.59 s Done. [Task 9/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 9/25] Current/Best: 14.32/ 14.32 GFLOPS | Progress: (4/20) | 11.96 s [Task 9/25] Current/Best: 22.48/ 22.48 GFLOPS | Progress: (8/20) | 13.75 s [Task 9/25] Current/Best: 8.02/ 22.48 GFLOPS | Progress: (12/20) | 16.27 s [Task 9/25] Current/Best: 17.85/ 22.48 GFLOPS | Progress: (16/20) | 19.11 s [Task 9/25] Current/Best: 9.10/ 22.48 GFLOPS | Progress: (20/20) | 27.56 s [Task 10/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 10/25] Current/Best: 18.29/ 18.29 GFLOPS | Progress: (4/20) | 2.59 s [Task 10/25] Current/Best: 15.66/ 18.29 GFLOPS | Progress: (8/20) | 4.23 s [Task 10/25] Current/Best: 11.58/ 18.69 GFLOPS | Progress: (12/20) | 5.78 s [Task 10/25] Current/Best: 19.21/ 20.22 GFLOPS | Progress: (16/20) | 6.90 s [Task 10/25] Current/Best: 8.10/ 20.22 GFLOPS | Progress: (20/20) | 8.45 s Done. [Task 11/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 11/25] Current/Best: 10.97/ 18.10 GFLOPS | Progress: (4/20) | 3.36 s [Task 11/25] Current/Best: 14.82/ 18.10 GFLOPS | Progress: (8/20) | 6.16 s [Task 11/25] Current/Best: 15.90/ 18.10 GFLOPS | Progress: (12/20) | 8.30 s [Task 11/25] Current/Best: 11.89/ 20.56 GFLOPS | Progress: (16/20) | 11.17 s [Task 11/25] Current/Best: 16.47/ 20.56 GFLOPS | Progress: (20/20) | 13.31 s Done. [Task 12/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 12/25] Current/Best: 7.81/ 17.96 GFLOPS | Progress: (4/20) | 5.68 s [Task 12/25] Current/Best: 5.13/ 17.96 GFLOPS | Progress: (8/20) | 9.57 s [Task 12/25] Current/Best: 18.86/ 18.86 GFLOPS | Progress: (12/20) | 11.60 s [Task 12/25] Current/Best: 15.24/ 18.86 GFLOPS | Progress: (16/20) | 14.50 s [Task 12/25] Current/Best: 15.06/ 18.86 GFLOPS | Progress: (20/20) | 16.49 s Done. [Task 13/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 13/25] Current/Best: 8.41/ 17.19 GFLOPS | Progress: (4/20) | 3.79 s [Task 13/25] Current/Best: 14.46/ 20.69 GFLOPS | Progress: (8/20) | 6.39 s [Task 13/25] Current/Best: 18.61/ 21.28 GFLOPS | Progress: (12/20) | 9.54 s [Task 13/25] Current/Best: 12.24/ 21.28 GFLOPS | Progress: (16/20) | 12.93 s [Task 13/25] Current/Best: 17.11/ 21.28 GFLOPS | Progress: (20/20) | 15.31 s Done. [Task 14/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 14/25] Current/Best: 12.49/ 13.41 GFLOPS | Progress: (4/20) | 3.44 s [Task 14/25] Current/Best: 6.02/ 13.41 GFLOPS | Progress: (8/20) | 5.65 s [Task 14/25] Current/Best: 19.62/ 19.62 GFLOPS | Progress: (12/20) | 8.33 s [Task 14/25] Current/Best: 16.08/ 19.62 GFLOPS | Progress: (16/20) | 10.02 s Done. [Task 14/25] Current/Best: 16.97/ 19.62 GFLOPS | Progress: (20/20) | 11.74 s [Task 15/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 15/25] Current/Best: 14.69/ 16.91 GFLOPS | Progress: (4/20) | 2.75 s [Task 15/25] Current/Best: 12.70/ 17.97 GFLOPS | Progress: (8/20) | 4.11 s [Task 15/25] Current/Best: 9.41/ 20.25 GFLOPS | Progress: (12/20) | 6.33 s [Task 15/25] Current/Best: 20.21/ 20.25 GFLOPS | Progress: (16/20) | 9.57 s [Task 15/25] Current/Best: 9.13/ 20.25 GFLOPS | Progress: (20/20) | 10.60 s [Task 16/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 16/25] Current/Best: 17.48/ 17.48 GFLOPS | Progress: (4/20) | 3.04 s [Task 16/25] Current/Best: 3.01/ 17.48 GFLOPS | Progress: (8/20) | 4.68 s [Task 16/25] Current/Best: 18.39/ 19.45 GFLOPS | Progress: (12/20) | 5.92 s [Task 16/25] Current/Best: 17.87/ 19.45 GFLOPS | Progress: (16/20) | 7.31 s [Task 16/25] Current/Best: 10.12/ 21.37 GFLOPS | Progress: (20/20) | 9.46 s Done. [Task 17/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 17/25] Current/Best: 12.65/ 16.08 GFLOPS | Progress: (4/20) | 4.82 s [Task 17/25] Current/Best: 13.05/ 22.36 GFLOPS | Progress: (8/20) | 7.72 s [Task 17/25] Current/Best: 16.32/ 22.36 GFLOPS | Progress: (12/20) | 9.85 s [Task 17/25] Current/Best: 16.38/ 22.36 GFLOPS | Progress: (16/20) | 12.04 s [Task 17/25] Current/Best: 10.01/ 22.36 GFLOPS | Progress: (20/20) | 14.17 s Done. [Task 18/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 18/25] Current/Best: 10.44/ 16.69 GFLOPS | Progress: (4/20) | 3.82 s [Task 18/25] Current/Best: 10.55/ 16.98 GFLOPS | Progress: (8/20) | 7.50 s [Task 18/25] Current/Best: 18.89/ 18.89 GFLOPS | Progress: (12/20) | 9.50 s [Task 18/25] Current/Best: 10.33/ 18.89 GFLOPS | Progress: (16/20) | 13.30 s [Task 18/25] Current/Best: 19.99/ 19.99 GFLOPS | Progress: (20/20) | 14.87 s Done. [Task 19/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 19/25] Current/Best: 7.38/ 19.92 GFLOPS | Progress: (4/20) | 6.07 s [Task 19/25] Current/Best: 2.69/ 19.92 GFLOPS | Progress: (8/20) | 9.44 s [Task 19/25] Current/Best: 18.70/ 20.50 GFLOPS | Progress: (12/20) | 12.41 s [Task 19/25] Current/Best: 12.60/ 20.84 GFLOPS | Progress: (16/20) | 15.42 s [Task 19/25] Current/Best: 2.70/ 22.11 GFLOPS | Progress: (20/20) | 18.30 s Done. [Task 20/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 20/25] Current/Best: 8.61/ 15.36 GFLOPS | Progress: (4/20) | 3.34 s Done. Done. [Task 20/25] Current/Best: 10.01/ 15.36 GFLOPS | Progress: (8/20) | 6.91 s [Task 20/25] Current/Best: 2.33/ 15.36 GFLOPS | Progress: (12/20) | 10.88 s [Task 20/25] Current/Best: 11.25/ 15.36 GFLOPS | Progress: (16/20) | 14.81 s [Task 20/25] Current/Best: 11.35/ 21.29 GFLOPS | Progress: (20/20) | 16.96 s [Task 21/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 21/25] Current/Best: 6.36/ 17.79 GFLOPS | Progress: (4/20) | 3.27 s [Task 21/25] Current/Best: 14.60/ 17.79 GFLOPS | Progress: (8/20) | 4.81 s [Task 21/25] Current/Best: 1.61/ 17.79 GFLOPS | Progress: (12/20) | 6.91 s [Task 21/25] Current/Best: 15.93/ 17.79 GFLOPS | Progress: (16/20) | 10.42 s [Task 21/25] Current/Best: 4.41/ 17.79 GFLOPS | Progress: (20/20) | 17.68 s [Task 22/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 22/25] Current/Best: 2.70/ 16.76 GFLOPS | Progress: (4/20) | 2.72 s [Task 22/25] Current/Best: 8.75/ 20.83 GFLOPS | Progress: (8/20) | 4.78 s [Task 22/25] Current/Best: 19.79/ 20.83 GFLOPS | Progress: (12/20) | 7.14 s [Task 22/25] Current/Best: 15.42/ 20.83 GFLOPS | Progress: (16/20) | 9.30 s [Task 22/25] Current/Best: 12.14/ 20.83 GFLOPS | Progress: (20/20) | 11.06 s Done. [Task 23/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 23/25] Current/Best: 16.62/ 19.80 GFLOPS | Progress: (4/20) | 3.28 s [Task 23/25] Current/Best: 13.02/ 19.81 GFLOPS | Progress: (8/20) | 6.76 s [Task 23/25] Current/Best: 20.66/ 21.75 GFLOPS | Progress: (12/20) | 8.58 s [Task 23/25] Current/Best: 6.61/ 21.75 GFLOPS | Progress: (16/20) | 15.61 s [Task 23/25] Current/Best: 7.83/ 21.75 GFLOPS | Progress: (20/20) | 19.80 s Done. [Task 24/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 24/25] Current/Best: 8.10/ 8.10 GFLOPS | Progress: (4/20) | 11.80 s [Task 24/25] Current/Best: 3.20/ 8.10 GFLOPS | Progress: (8/20) | 23.05 s [Task 24/25] Current/Best: 3.78/ 8.10 GFLOPS | Progress: (12/20) | 33.78 s Done. [Task 24/25] Current/Best: 5.23/ 8.87 GFLOPS | Progress: (16/20) | 39.29 s [Task 24/25] Current/Best: 3.01/ 8.87 GFLOPS | Progress: (20/20) | 45.16 s Done. [Task 25/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s [Task 25/25] Current/Best: 1.55/ 2.73 GFLOPS | Progress: (4/20) | 11.61 s [Task 25/25] Current/Best: 6.18/ 8.42 GFLOPS | Progress: (8/20) | 22.91 s [Task 25/25] Current/Best: 6.10/ 8.42 GFLOPS | Progress: (12/20) | 34.36 s [Task 25/25] Current/Best: 5.93/ 8.74 GFLOPS | Progress: (16/20) | 36.18 s [Task 25/25] Current/Best: 2.80/ 9.20 GFLOPS | Progress: (20/20) | 46.90 s .. GENERATED FROM PYTHON SOURCE LINES 375-404 The output from this tuning process will look something like this: .. code-block:: bash # [Task 1/24] Current/Best: 10.71/ 21.08 GFLOPS | Progress: (60/1000) | 111.77 s Done. # [Task 1/24] Current/Best: 9.32/ 24.18 GFLOPS | Progress: (192/1000) | 365.02 s Done. # [Task 2/24] Current/Best: 22.39/ 177.59 GFLOPS | Progress: (960/1000) | 976.17 s Done. # [Task 3/24] Current/Best: 32.03/ 153.34 GFLOPS | Progress: (800/1000) | 776.84 s Done. # [Task 4/24] Current/Best: 11.96/ 156.49 GFLOPS | Progress: (960/1000) | 632.26 s Done. # [Task 5/24] Current/Best: 23.75/ 130.78 GFLOPS | Progress: (800/1000) | 739.29 s Done. # [Task 6/24] Current/Best: 38.29/ 198.31 GFLOPS | Progress: (1000/1000) | 624.51 s Done. # [Task 7/24] Current/Best: 4.31/ 210.78 GFLOPS | Progress: (1000/1000) | 701.03 s Done. # [Task 8/24] Current/Best: 50.25/ 185.35 GFLOPS | Progress: (972/1000) | 538.55 s Done. # [Task 9/24] Current/Best: 50.19/ 194.42 GFLOPS | Progress: (1000/1000) | 487.30 s Done. # [Task 10/24] Current/Best: 12.90/ 172.60 GFLOPS | Progress: (972/1000) | 607.32 s Done. # [Task 11/24] Current/Best: 62.71/ 203.46 GFLOPS | Progress: (1000/1000) | 581.92 s Done. # [Task 12/24] Current/Best: 36.79/ 224.71 GFLOPS | Progress: (1000/1000) | 675.13 s Done. # [Task 13/24] Current/Best: 7.76/ 219.72 GFLOPS | Progress: (1000/1000) | 519.06 s Done. # [Task 14/24] Current/Best: 12.26/ 202.42 GFLOPS | Progress: (1000/1000) | 514.30 s Done. # [Task 15/24] Current/Best: 31.59/ 197.61 GFLOPS | Progress: (1000/1000) | 558.54 s Done. # [Task 16/24] Current/Best: 31.63/ 206.08 GFLOPS | Progress: (1000/1000) | 708.36 s Done. # [Task 17/24] Current/Best: 41.18/ 204.45 GFLOPS | Progress: (1000/1000) | 736.08 s Done. # [Task 18/24] Current/Best: 15.85/ 222.38 GFLOPS | Progress: (980/1000) | 516.73 s Done. # [Task 19/24] Current/Best: 15.78/ 203.41 GFLOPS | Progress: (1000/1000) | 587.13 s Done. # [Task 20/24] Current/Best: 30.47/ 205.92 GFLOPS | Progress: (980/1000) | 471.00 s Done. # [Task 21/24] Current/Best: 46.91/ 227.99 GFLOPS | Progress: (308/1000) | 219.18 s Done. # [Task 22/24] Current/Best: 13.33/ 207.66 GFLOPS | Progress: (1000/1000) | 761.74 s Done. # [Task 23/24] Current/Best: 53.29/ 192.98 GFLOPS | Progress: (1000/1000) | 799.90 s Done. # [Task 24/24] Current/Best: 25.03/ 146.14 GFLOPS | Progress: (1000/1000) | 1112.55 s Done. .. GENERATED FROM PYTHON SOURCE LINES 406-415 Compiling an Optimized Model with Tuning Data ---------------------------------------------- As an output of the tuning process above, we obtained the tuning records stored in ``resnet-50-v2-autotuning.json``. The compiler will use the results to generate high performance code for the model on your specified target. Now that tuning data for the model has been collected, we can re-compile the model using optimized operators to speed up our computations. .. GENERATED FROM PYTHON SOURCE LINES 415-423 .. code-block:: default with autotvm.apply_history_best(tuning_option["tuning_records"]): with tvm.transform.PassContext(opt_level=3, config={}): lib = relay.build(mod, target=target, params=params) dev = tvm.device(str(target), 0) module = graph_executor.GraphModule(lib["default"](dev)) .. rst-class:: sphx-glr-script-out .. code-block:: none Done. .. GENERATED FROM PYTHON SOURCE LINES 424-425 Verify that the optimized model runs and produces the same results: .. GENERATED FROM PYTHON SOURCE LINES 425-438 .. code-block:: default dtype = "float32" module.set_input(input_name, img_data) module.run() output_shape = (1, 1000) tvm_output = module.get_output(0, tvm.nd.empty(output_shape)).numpy() scores = softmax(tvm_output) scores = np.squeeze(scores) ranks = np.argsort(scores)[::-1] for rank in ranks[0:5]: print("class='%s' with probability=%f" % (labels[rank], scores[rank])) .. rst-class:: sphx-glr-script-out .. code-block:: none class='n02123045 tabby, tabby cat' with probability=0.621104 class='n02123159 tiger cat' with probability=0.356378 class='n02124075 Egyptian cat' with probability=0.019712 class='n02129604 tiger, Panthera tigris' with probability=0.001215 class='n04040759 radiator' with probability=0.000262 .. GENERATED FROM PYTHON SOURCE LINES 439-448 Verifying that the predictions are the same: .. code-block:: bash # class='n02123045 tabby, tabby cat' with probability=0.610550 # class='n02123159 tiger cat' with probability=0.367181 # class='n02124075 Egyptian cat' with probability=0.019365 # class='n02129604 tiger, Panthera tigris' with probability=0.001273 # class='n04040759 radiator' with probability=0.000261 .. GENERATED FROM PYTHON SOURCE LINES 450-456 Comparing the Tuned and Untuned Models -------------------------------------- We want to collect some basic performance data associated with this optimized model to compare it to the unoptimized model. Depending on your underlying hardware, number of iterations, and other factors, you should see a performance improvement in comparing the optimized model to the unoptimized model. .. GENERATED FROM PYTHON SOURCE LINES 456-472 .. code-block:: default import timeit timing_number = 10 timing_repeat = 10 optimized = ( np.array(timeit.Timer(lambda: module.run()).repeat(repeat=timing_repeat, number=timing_number)) * 1000 / timing_number ) optimized = {"mean": np.mean(optimized), "median": np.median(optimized), "std": np.std(optimized)} print("optimized: %s" % (optimized)) print("unoptimized: %s" % (unoptimized)) .. rst-class:: sphx-glr-script-out .. code-block:: none optimized: {'mean': 412.1540658899994, 'median': 413.704081700007, 'std': 4.31069322949314} unoptimized: {'mean': 512.4081050999996, 'median': 511.8093637999948, 'std': 1.8920776301187912} .. GENERATED FROM PYTHON SOURCE LINES 473-485 Final Remarks ------------- In this tutorial, we gave a short example of how to use the TVM Python API to compile, run, and tune a model. We also discussed the need for pre and post-processing of inputs and outputs. After the tuning process, we demonstrated how to compare the performance of the unoptimized and optimize models. Here we presented a simple example using ResNet-50 v2 locally. However, TVM supports many more features including cross-compilation, remote execution and profiling/benchmarking. .. rst-class:: sphx-glr-timing **Total running time of the script:** ( 10 minutes 23.196 seconds) .. _sphx_glr_download_tutorial_autotvm_relay_x86.py: .. only:: html .. container:: sphx-glr-footer sphx-glr-footer-example .. container:: sphx-glr-download sphx-glr-download-python :download:`Download Python source code: autotvm_relay_x86.py ` .. container:: sphx-glr-download sphx-glr-download-jupyter :download:`Download Jupyter notebook: autotvm_relay_x86.ipynb ` .. only:: html .. rst-class:: sphx-glr-signature `Gallery generated by Sphinx-Gallery `_