diff --git a/docs/tutorials/index.md b/docs/tutorials/index.md index 7e0ffaa3f72a..ec3788c957f9 100644 --- a/docs/tutorials/index.md +++ b/docs/tutorials/index.md @@ -89,8 +89,9 @@ Select API:  * [Learning Rate Schedules](/tutorials/gluon/learning_rate_schedules.html) * [Advanced Learning Rate Schedules](/tutorials/gluon/learning_rate_schedules_advanced.html) * [Profiling MXNet Models](/tutorials/python/profiler.html) - * [Hybridize Gluon models with control flows](/tutorials/control_flow/ControlFlowTutorial.html) + * [Module to Gluon API](/tutorials/python/module_to_gluon.html) (new!) * [Gluon end to end from training to inference](/tutorials/gluon/gluon_from_experiment_to_deployment.html) + * API Guides * Core APIs * NDArray @@ -113,6 +114,7 @@ Select API:  * [HybridBlocks](/tutorials/gluon/hybrid.html) ([Alternative](http://gluon.mxnet.io/chapter07_distributed-learning/hybridize.html) External link) * [Block Naming](/tutorials/gluon/naming.html) * [Custom Operators](/tutorials/gluon/customop.html) + * [Control Flow operators](/tutorials/control_flow/ControlFlowTutorial.html) (new!) * Autograd * [AutoGrad API](/tutorials/gluon/autograd.html) * [AutoGrad API with chain rule](http://gluon.mxnet.io/chapter01_crashcourse/autograd.html) External link diff --git a/docs/tutorials/python/module_to_gluon.md b/docs/tutorials/python/module_to_gluon.md new file mode 100644 index 000000000000..5ab9d88cbd23 --- /dev/null +++ b/docs/tutorials/python/module_to_gluon.md @@ -0,0 +1,362 @@ + + + + + + + + + + + + + + + + + +# Converting Module API code to the Gluon API + +Sometimes you find yourself in the situation where the model you want to use has been written using the symbolic Module API rather than the simpler, easier-to-debug, more flexible, imperative Gluon API. In this tutorial, we will give you a comprehensive guide for transforming Module code to Gluon code. + +The different steps to take into consideration are: + +I) Data loading + +II) Model definition + +III) Loss + +IV) Training Loop + +V) Exporting Models + +VI) Loading Models for Inference + +In the following section we will look at 1:1 mappings between the Module and the Gluon ways of training a neural network. + +## I - Data Loading + +In this section we will be looking at the difference in loading data between Module and Gluon. +Let's first import a few Python modules. + +```python +from collections import namedtuple +import logging +logging.basicConfig(level=logging.INFO) +import random + +import numpy as np +import mxnet as mx +from mxnet.gluon.data import ArrayDataset, DataLoader +from mxnet.gluon import nn +from mxnet import gluon + +# parameters +batch_size = 5 +dataset_length = 50 + +# random seeds +random.seed(1) +np.random.seed(1) +mx.random.seed(1) + +``` + +#### Module + +When using the Module API we use a [`DataIter`](https://mxnet.incubator.apache.org/api/python/io/io.html?highlight=dataiter#mxnet.io.DataIter), in addition to the data itself, the [`DataIter`](https://mxnet.incubator.apache.org/api/python/io/io.html?highlight=dataiter#mxnet.io.DataIter) contains information about the name of the input symbols. + +In the Module API, `DataIter`s are responsible for both holding the data and iterating through it. Some `DataIter`s support multi-threading like the [`ImageRecordIter`](https://mxnet.incubator.apache.org/api/python/io/io.html#mxnet.io.ImageRecordIter), while other don't, such as the [`NDArrayIter`](https://mxnet.incubator.apache.org/api/python/io/io.html?highlight=ndarrayiter#mxnet.io.NDArrayIter). + +Let's create some random data, following the same format as grayscale 28x28 images. + + +```python +train_data = np.random.rand(dataset_length, 28,28).astype('float32') +train_label = np.random.randint(0, 10, (dataset_length,)).astype('float32') +``` + +We can now wraps this data into an ArrayIterator that will create batches of data using the first dimension of the provided array as the batch dimension. + +```python +data_iter = mx.io.NDArrayIter(data=train_data, label=train_label, batch_size=batch_size, shuffle=False, data_name='data', label_name='softmax_label') +for batch in data_iter: + print(batch.data[0].shape, batch.label[0]) + break +``` + + (5, 28, 28) + [5. 0. 3. 4. 9.] + + + +#### Gluon + +With Gluon, the preferred method is to use a [`DataLoader`](https://mxnet.incubator.apache.org/api/python/gluon/data.html?highlight=dataloader#mxnet.gluon.data.DataLoader) that makes use of a [`Dataset`](https://mxnet.incubator.apache.org/api/python/gluon/data.html?highlight=dataset#mxnet.gluon.data.Dataset) to asynchronously prefetch the data. + +The Gluon API offers you the ability to efficiently fetch data and separate the concerns of loading versus holding data. The DataLoader role is to request certain indices of the dataset. The Dataset has ownership of the data. +The `Dataset` data can be in or out of memory, and the `DataLoader` role is to request certain indices of the dataset, in the main thread or through multi-processing (or multi-threaded) workers and batch the data together. + +```python +dataset = ArrayDataset(train_data, train_label) +dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=False, num_workers=0) +for data, label in dataloader: + print(data.shape, label) + break +``` + + (5, 28, 28) + [5. 0. 3. 4. 9.] + + +You can check the [`Dataset` and `DataLoader` tutorials](https://mxnet.incubator.apache.org/tutorials/gluon/datasets.html) out. You can either rewrite your code in order to use one of the provided [`Dataset`](https://mxnet.incubator.apache.org/api/python/gluon/data.html?highlight=dataset#mxnet.gluon.data.Dataset) class, like the [`ArrayDataset`](https://mxnet.incubator.apache.org/api/python/gluon/data.html?highlight=arraydataset#mxnet.gluon.data.ArrayDataset) or the [`ImageFolderDataset`](https://mxnet.incubator.apache.org/api/python/gluon/data.html?highlight=imagefolderdataset#mxnet.gluon.data.vision.datasets.ImageFolderDataset) + + +## II - Model Definition + +Let's look at the model definition from the [MNIST Module Tutorial](https://mxnet.incubator.apache.org/tutorials/python/mnist.html): + +#### Module + +For the Module API, you define the data flow by setting `data` keyword argument of one layer to the next. +You then bind the symbolic model to a specific compute context and specify the symbol names for the data and the label. + +```python + +# context +ctx = mx.cpu() + +def get_module_network(): + data = mx.sym.var('data') + data = mx.sym.flatten(data=data) + fc1 = mx.sym.FullyConnected(data=data, num_hidden=128) + act1 = mx.sym.Activation(data=fc1, act_type="relu") + fc2 = mx.sym.FullyConnected(data=act1, num_hidden = 64) + act2 = mx.sym.Activation(data=fc2, act_type="relu") + fc3 = mx.sym.FullyConnected(data=act2, num_hidden=10) + mlp = mx.sym.SoftmaxOutput(data=fc3, name='softmax') + return mlp + +mlp = get_module_network() +# Bind model to Module +mlp_model = mx.mod.Module(symbol=mlp, context=ctx, data_names=['data'], label_names=['softmax_label']) +``` + +#### Gluon + +In Gluon, for the equivalent model, you would create a `Sequential` block, in that case a `HybridSequential` block to allow for future hybridization since we are only using [hybridizable blocks](https://mxnet.incubator.apache.org/tutorials/gluon/hybrid.html). The flow of the data will be automatically set from one layer to the next, since they are held in a `Sequential` block. +Note that we don't need named symbols for the input, and we show how the loss is handled in Gluon in the next section. + +```python +def get_gluon_network(): + net = nn.HybridSequential() + with net.name_scope(): + net.add( + nn.Flatten(), + nn.Dense(units=128, activation="relu"), + nn.Dense(units=64, activation="relu"), + nn.Dense(units=10) + ) + return net + +net = get_gluon_network() +``` + +## III - Loss + +The loss, that you are trying to minimize using an optimization algorithm like SGD, is defined differently in the Module API than in Gluon. + + +#### Module + + +In the module API, the loss is part of the network. It has usually a forward pass result, that is the inference value, and a backward pass that is the gradient of the output with respect to that particular loss. + +For example, the [sym.SoftmaxOutput](https://mxnet.incubator.apache.org/api/python/symbol/symbol.html?highlight=softmaxout#mxnet.symbol.SoftmaxOutput) is a softmax output in the forward pass and the gradient with respect to the cross-entropy loss in the backward pass. + +```python +# Softmax with cross entropy loss, directly part of the network +out = mx.sym.SoftmaxOutput(data=mlp, name='softmax') +``` + +#### Gluon + + +In Gluon, it is a lot more transparent. Losses, like the [SoftmaxCrossEntropyLoss](https://mxnet.incubator.apache.org/api/python/gluon/loss.html?highlight=softmaxcross#mxnet.gluon.loss.SoftmaxCrossEntropyLoss), are only computing the actual value of the loss. You then call `.backward()` on the loss value to compute the gradient of the parameters with respect to that loss. At inference time, you simply call `.softmax()` on your output to get the output of your network normalized according to the softmax function. + + +```python +# We simply create a loss function we will use in our training loop +loss_fn = gluon.loss.SoftmaxCrossEntropyLoss() +``` + +In the next section we will show how you use this loss function in Gluon to generate the loss value in the main training loop. + +## IV - Training Loop + + +#### Module + +The Module API provides a [`.fit()`](https://mxnet.incubator.apache.org/api/python/module/module.html?highlight=.fit#mxnet.module.BaseModule.fit) function that takes care of fitting training data to your symbolic model. With Gluon, your execution flow controls the data flow, so you need to write your own loop. It might seems like it is more verbose, but you have a lot more control as to what is happening during the training. +With the [`.fit()`](https://mxnet.incubator.apache.org/api/python/module/module.html?highlight=.fit#mxnet.module.BaseModule.fit) function, you control the metric reporting, checkpointing or weights initialization through a lot of different keyword arguments (check the [docs](https://mxnet.incubator.apache.org/api/python/module/module.html?highlight=.fit#mxnet.module.BaseModule.fit)). That is where you define the optimizer for example. + +```python +mlp_model.fit(data_iter, # train data + eval_data=data_iter, # validation data + optimizer='sgd', # use SGD to train + force_init=True, + force_rebind=True, + optimizer_params={'learning_rate':0.1}, # use fixed learning rate + eval_metric='acc', # report accuracy during training + num_epoch=5) # train for 5 full dataset passes +``` + +```INFO:root:Epoch[4] Train-accuracy=0.070000``` + +```INFO:root:Epoch[4] Time cost=0.038``` + +```INFO:root:Epoch[4] Validation-accuracy=0.125000``` + +#### Gluon + + +With Gluon, you do these operations directly in the training loop, and the optimizer is part of the [`Trainer`](https://mxnet.incubator.apache.org/api/python/gluon/gluon.html?highlight=trainer#mxnet.gluon.Trainer) object that handles the weight updates of your parameters. + +Notice the `loss.backward()` we call before updating the weight as mentionned in the previous section + +```python +net.initialize(mx.init.Xavier(magnitude=2.24), ctx=ctx) # Initialize network and trainer +trainer = gluon.Trainer(net.collect_params(), 'sgd', {'learning_rate': 0.1}) + +metric = mx.metric.Accuracy() # Pick a metric + +for e in range(5): # start of epoch + + for data, label in dataloader: # start of mini-batch + data = data.as_in_context(ctx) + label = label.as_in_context(ctx) + + with mx.autograd.record(): + output = net(data) # forward pass + loss = loss_fn(output, label) # get loss + + loss.backward() # compute gradients + trainer.step(data.shape[0]) # update weights with SGD + metric.update(label, output) # update the metrics # end of mini-batch + + name, acc = metric.get() + print('training metrics at epoch %d: %s=%f'%(e, name, acc)) + metric.reset() # end of epoch +``` + +```training metrics at epoch 3: accuracy=0.155000``` + +```training metrics at epoch 4: accuracy=0.145000``` + +The Gluon training code is more verbose than the simple `.fit` from Module. However that is also the main advantage, there is no black magic going on here, you have full control of your training loop. You can for example easily set breakpoints, modify a learning rate or print data during the training flow. This flexibility also makes easy to implement more complex use-case like gradient accumulation across batches. + +## V - Exporting Model + +The ultimate purpose of training a model is to be able to export it and share it, whether it is for deployment or simply reproducibility purposes. + +#### Module + + +With the Module API, you can save model using the [`.save_checkpoint()`](https://mxnet.incubator.apache.org/api/python/module/module.html?highlight=save_chec#mxnet.module.Module.save_checkpoint) and get a `-symbol.json` and a `.params` file that represent your network. + + +```python +mlp_model.save_checkpoint('module-model', epoch=5) +# module-model-0005.params module-model-symbol.json +``` + +```INFO:root:Saved checkpoint to "module-model-0005.params"``` + +#### Gluon + + + +With Gluon, network parameters are associated with a `Block`, but the execution flow is controlled in python through the code in `.forward()` function. Hence only [hybridized networks]() can be exported with a `-symbol.json` and `.params` file using [`.export()`](https://mxnet.incubator.apache.org/api/python/gluon/gluon.html?highlight=export#mxnet.gluon.HybridBlock.export), non-hybridized models can only have their parameters exported using [`.save_parameters()`](https://mxnet.incubator.apache.org/api/python/gluon/gluon.html?highlight=save_pa#mxnet.gluon.Block.save_parameters). Check this great tutorial to learn more: [Saving and Loading Gluon Models](https://mxnet.incubator.apache.org/tutorials/gluon/save_load_params.html). + + +Any models: + +```python +# save only the parameters +net.save_parameters('gluon-model.params') +# gluon-model.params +``` + +Hybridized models: + +```python +# save the parameters and the symbolic representation +net.hybridize() +net(mx.nd.ones((1,1,28,28), ctx)) + +net.export('gluon-model-hybrid', epoch=5) +# gluon-model-hybrid-symbol.json gluon-model-hybrid-0005.params +``` + +## VI - Loading Model for Inference + + +#### Module + + +For inference, in the Module API, you need to first load the parameters and symbol, bind the symbol to a module and load the corresponding parameters. You can then pass a batch of data through that module and request the output of the network. + + +```python +# Load the symbol and parameters +sym, arg_params, aux_params = mx.model.load_checkpoint('module-model', 5) + +# Bind them in a module +mod = mx.mod.Module(symbol=sym, context=ctx, label_names=None) +mod.bind(for_training=False, data_shapes=[('data', (1,1,28,28))], + label_shapes=mod._label_shapes) + +# Set the parameters +mod.set_params(arg_params, aux_params, allow_missing=True) + +# Run the inference +Batch = namedtuple('Batch', ['data']) +mod.forward(Batch([mx.nd.ones((1,28,28))])) +prob = mod.get_outputs()[0].asnumpy() +print("Output probabilities: {}".format(prob)) +``` + +`Output probabilities: [[0.05537598 0.03889056 0.06126577 0.08879893 0.12371024 0.05759033 0.1378248 0.26134694 0.07905186 0.09614458]]` + +#### Gluon (Symbolic Model) + +For the Gluon API, it is a lot simpler. You can just load a serialized model in a [`SymbolBlock`](https://mxnet.incubator.apache.org/api/python/gluon/gluon.html?highlight=symbolblo#mxnet.gluon.SymbolBlock) and run inference directly. + +```python +net = gluon.SymbolBlock.imports('module-model-symbol.json', ['data', 'softmax_label'], 'module-model-0005.params') +prob = net(mx.nd.ones((1,1,28,28)), mx.nd.ones(1)) # note the second argument here to account for the softmax_label symbol +print("Output probabilities: {}".format(prob.asnumpy())) +``` + +`Output probabilities: [[0.05537598 0.03889056 0.06126577 0.08879893 0.12371024 0.05759033 0.1378248 0.26134694 0.07905186 0.09614458]]` + +#### Gluon (Imperative Model) + +```python +net = get_gluon_network() +net.load_parameters('gluon-model.params') +prob = net(mx.nd.ones((1,1,28,28))).softmax() +print("Output probabilities: {}".format(prob.asnumpy())) +``` + +`Output probabilities: [[0.01298077 0.00173413 0.01661885 0.3362421 0.00536332 0.02099853 0.01413316 0.5528366 0.0133819 0.02571066]]` + +## Conclusion + +This tutorial lead you through the steps necessary to train a deep learning model and showed you the differences between the symbolic approach of the Module API and the imperative one of the Gluon API. If you need more help converting your Module API code to the Gluon API, reach out to the community on the [discuss forum](https://discuss.mxnet.io)! +You can also compare the scripts for training MNIST in [Gluon](https://mxnet.incubator.apache.org/tutorials/gluon/mnist.html) and [Module](https://mxnet.incubator.apache.org/tutorials/python/mnist.html). + + + + diff --git a/tests/tutorials/test_tutorials.py b/tests/tutorials/test_tutorials.py index 37ba9918fb70..3ccbf764e6ff 100644 --- a/tests/tutorials/test_tutorials.py +++ b/tests/tutorials/test_tutorials.py @@ -169,6 +169,9 @@ def test_python_data_augmentation_with_masks(): def test_python_kvstore(): assert _test_tutorial_nb('python/kvstore') +def test_module_to_gluon(): + assert _test_tutorial_nb('python/module_to_gluon') + def test_python_types_of_data_augmentation(): assert _test_tutorial_nb('python/types_of_data_augmentation') @@ -198,4 +201,3 @@ def test_vision_cnn_visualization(): def test_control_flow(): assert _test_tutorial_nb('control_flow/ControlFlowTutorial') -