Caution: Run the first three cells in the sequential order or you will end up getting errors in imports.
\n"},{"metadata":{},"cell_type":"markdown","source":"# Setup Pytorch for TPU."},{"metadata":{},"cell_type":"markdown","source":"\nBasically Pytorch uses torch_xla package to connect to do computations on TPU. Not only we can train on a single TPU core but we can also perform computations utilizing all the available TPU cores. The following cell loads the torch_xla package and setup dependencies for it.
\nThe next cell is inspired from the notebook of one of my favorite mentors- Abhishek Thakur.
\n "},{"metadata":{"trusted":true},"cell_type":"code","source":"!curl https://raw.githubusercontent.com/pytorch/xla/master/contrib/scripts/env-setup.py -o pytorch-xla-env-setup.py\n!python pytorch-xla-env-setup.py --apt-packages libomp5 libopenblas-dev","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"\nImports
\n"},{"metadata":{"trusted":true},"cell_type":"code","source":"# Declaring the path to load efficientNet models.\nimport sys\nsys.path.append('../input/efficientnet-pytorch/EfficientNet-PyTorch/EfficientNet-PyTorch-master')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"#IMPORTS\nimport os\nimport gc\nimport time\nimport numpy as np\nimport pandas as pd\nfrom scipy import stats\nfrom tqdm.auto import tqdm\nfrom typing import Dict\nimport matplotlib.pyplot as plt\nfrom tempfile import gettempdir\nfrom prettytable import PrettyTable\nfrom transformers import AdamW, get_linear_schedule_with_warmup, get_constant_schedule\n\n# TORCH\nimport torch\nfrom torch import nn, optim\nfrom torch.optim import lr_scheduler\nfrom torch.utils.data import DataLoader\n\n# TORCH XLA\nimport torch_xla\nimport torch_xla.debug.metrics as met\nimport torch_xla.distributed.data_parallel as dp\nimport torch_xla.distributed.parallel_loader as pl\nimport torch_xla.utils.utils as xu\nimport torch_xla.core.xla_model as xm\nimport torch_xla.distributed.xla_multiprocessing as xmp\nimport torch_xla.test.test_utils as test_utils\n\n# L5Kit\nfrom l5kit.configs import load_config_data\nfrom l5kit.geometry import transform_points\nfrom l5kit.rasterization import build_rasterizer\nfrom l5kit.dataset import AgentDataset, EgoDataset\nfrom l5kit.data import LocalDataManager, ChunkedDataset\nfrom l5kit.evaluation.metrics import neg_multi_log_likelihood, time_displace\nfrom l5kit.visualization import PREDICTED_POINTS_COLOR, TARGET_POINTS_COLOR, draw_trajectory\nfrom l5kit.evaluation import write_pred_csv, compute_metrics_csv, read_gt_csv, create_chopped_dataset\n\n# EfficientNet\nfrom efficientnet_pytorch import model as enet\n\n# Catalyst Module\nfrom catalyst import dl\nfrom catalyst.utils import metrics\nfrom catalyst.dl import utils\nfrom catalyst.data.sampler import DistributedSamplerWrapper, BalanceClassSampler\n\n# Filter warnings\nimport warnings\nwarnings.filterwarnings(\"ignore\")","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"# Configurations"},{"metadata":{"trusted":true},"cell_type":"code","source":"# CONFIGRATIONS\n\nclass Config:\n WEIGHT_FILE = None # Model state_dict path of previously trained model\n \n MODEL_NAME = \"efficientnet-b0\" # b0-b7 could be the different choices.\n \n IMG_SIZE = 222\n \n PIXEL_SIZE = 0.45\n \n EPOCHS = 3 # Epochs to train the model for.\n \n l_rate = 0.001 # Learning rate\n\n scheduler_params = dict(\n mode='max',\n factor=0.7,\n patience=0,\n verbose=False, \n threshold=0.0001,\n threshold_mode='abs',\n cooldown=0, \n min_lr=1e-8,\n eps=1e-08\n )\n \n SchedulerClass = torch.optim.lr_scheduler.ReduceLROnPlateau # Scheduler for learning rate.\n \n criterion = torch.nn.MSELoss(reduction=\"none\") # Loss function.\n \n verbose_steps = 50 # Steps to print model's training status after.\n \nconfig = Config()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# L5KIT'S CONFIGRATIONS\n\nos.environ[\"L5KIT_DATA_FOLDER\"] = \"../input/lyft-motion-prediction-autonomous-vehicles\"\ndm = LocalDataManager()\ncfg = {\n 'model_params': {'model_architecture': 'efficientnet-b6',\n 'history_num_frames': 0,\n 'history_step_size': 1,\n 'history_delta_time': 0.1,\n 'future_num_frames': 50,\n 'future_step_size': 1,\n 'future_delta_time': 0.1},\n\n 'raster_params': {'raster_size': [config.IMG_SIZE, config.IMG_SIZE],\n 'pixel_size': [config.PIXEL_SIZE, config.PIXEL_SIZE],\n 'ego_center': [0.25, 0.5],\n 'map_type': 'py_semantic',\n 'satellite_map_key': 'aerial_map/aerial_map.png',\n 'semantic_map_key': 'semantic_map/semantic_map.pb',\n 'dataset_meta_key': 'meta.json',\n 'filter_agents_threshold': 0.5},\n\n 'train_data_loader': {'key': 'scenes/train.zarr',\n 'batch_size': 12,\n 'shuffle': True,\n 'num_workers': 8},\n\n \"valid_data_loader\":{\"key\": \"scenes/validation.zarr\",\n \"batch_size\": 8,\n \"shuffle\": False,\n \"num_workers\": 5},\n \n }","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"# Utility Functions\n\nAt the end I would try to explain some of the parts of the code that I think should be discussed with concepts.
\n\nTo keep things simpler I did not add the code for logging. You could, if you want to or I would be anyways adding it in the next version of this notebook.
"},{"metadata":{},"cell_type":"markdown","source":"Future Scope:
\n\n1. LabelSmoothing\n2. Sampler (if necessary)\n3. Logging Functionality\n\nFollowing code cell is inspired from this [kernel](https://www.kaggle.com/shonenkov/tpu-training-super-fast-xlmroberta)."},{"metadata":{"trusted":true},"cell_type":"code","source":"class TPUFitter:\n def __init__(self, model, device):\n self.model = model\n self.device = device\n\n # Following some lines are for setting up the AdamW optimizer.\n # See below explanation for details.\n param_optimizer = list(self.model.named_parameters())\n no_decay = ['bias', 'LayerNorm.bias', 'LayerNorm.weight']\n optimizer_grouped_parameters = [\n {'params': [p for n, p in param_optimizer if not any(nd in n for nd in no_decay)],\n 'weight_decay': 0.001},\n \n {'params': [p for n, p in param_optimizer if any(nd in n for nd in no_decay)],\n 'weight_decay': 0.0}\n ]\n\n self.optimizer = AdamW(optimizer_grouped_parameters, lr=config.l_rate*xm.xrt_world_size())\n self.scheduler = config.SchedulerClass(self.optimizer, **config.scheduler_params)\n \n # Following function is used for printing to output efficiently. \n xm.master_print(f'Model Loaded on {self.device}')\n\n def fit(self, train_loader, validation_loader):\n \"\"\"Function to fit the model.\"\"\"\n \n # Loop over the Epochs\n for e in range(config.EPOCHS):\n \n t = time.time() # Get a measurement of time.\n para_loader = pl.ParallelLoader(train_loader, [self.device]) # Distributed loading of the model.\n loss = self.forward(para_loader.per_device_loader(self.device))\n xm.master_print(\n f'[RESULT]: Train. Epoch: {e+1}, ' + \\\n f'loss: {loss:.5f}, '+ \\\n f'time: {(time.time() - t):.5f}'\n )\n xm.master_print(\"\\n\")\n\n t = time.time()\n para_loader = pl.ParallelLoader(validation_loader, [self.device])\n loss = self.validation(para_loader.per_device_loader(self.device))\n xm.master_print(\n f'[RESULT]: Validation. Epoch: {e+1}, ' + \\\n f'loss: {loss:.5f}, ' + \\\n f'time: {(time.time() - t):.5f}'\n )\n xm.master_print(\"\\n\")\n\n def validation(self, val_loader):\n \"\"\"Function to validate the model's predictions.\"\"\"\n \n # Setting model to evaluation mode.\n self.model.eval()\n t = time.time()\n \n for step, data in enumerate(val_loader):\n with torch.no_grad():\n inputs = data[\"image\"].to(self.device)\n targets = data[\"target_positions\"].to(self.device)\n target_availabilities = data[\"target_availabilities\"].unsqueeze(-1).to(self.device)\n \n outputs = self.model(inputs)\n outputs = outputs.reshape(targets.shape)\n loss = config.criterion(outputs, targets)\n loss = loss * target_availabilities\n loss = loss.mean()\n\n if step % config.verbose_steps == 0:\n xm.master_print(\n f'Valid Step {step}, loss: ' + \\\n f'{loss:.4f}' + \\\n f'time: {(time.time() - t):.5f}'\n ) \n return loss\n \n def forward(self, train_loader):\n \"\"\"Function to perform custom forward propagation.\"\"\"\n \n # Setting model to training mode.\n self.model.train()\n \n t = time.time()\n for step, data in enumerate(train_loader):\n inputs = data[\"image\"].to(self.device)\n target_availabilities = data[\"target_availabilities\"].unsqueeze(-1).to(self.device)\n targets = data[\"target_positions\"].to(self.device)\n \n outputs = self.model(inputs)\n outputs = outputs.reshape(targets.shape)\n loss = config.criterion(outputs, targets)\n \n loss = loss * target_availabilities\n loss = loss.mean()\n\n if step % config.verbose_steps == 0:\n xm.master_print(\n f'Train Step {step}, loss: ' + \\\n f'{loss:.4f}' + \\\n f'time: {(time.time() - t):.5f}'\n )\n self.optimizer.zero_grad()\n \n loss.backward()\n xm.optimizer_step(self.optimizer)\n \n self.model.eval()\n self.save('last-checkpoint.bin')\n return loss\n\n def save(self, path):\n \"\"\"Function to save the model's current state.\"\"\"\n xm.save(self.model.state_dict(), path)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# Implementation of class to load the particular EfficientNet model.\nclass LyftModel(torch.nn.Module):\n def __init__(self, cfg):\n super().__init__()\n self.backbone = enet.EfficientNet.from_name(config.MODEL_NAME)\n \n num_history_channels = (cfg[\"model_params\"][\"history_num_frames\"] + 1) * 2\n num_in_channels = 3 + num_history_channels\n num_targets = 2*cfg[\"model_params\"][\"future_num_frames\"]\n \n self.backbone._conv_stem = nn.Conv2d(\n num_in_channels,\n self.backbone._conv_stem.out_channels,\n kernel_size=self.backbone._conv_stem.kernel_size,\n stride=self.backbone._conv_stem.stride,\n padding=self.backbone._conv_stem.padding,\n bias=False\n )\n \n self.backbone._fc = nn.Linear(in_features=self.backbone._fc.in_features, out_features=num_targets)\n \n def forward(self, x):\n \"\"\"Function to perform forward propagation.\"\"\"\n x = self.backbone(x)\n return x","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def get_dataloader(config, zarr_data, subset_len, map_type=\"py_semantic\"):\n \"\"\"Creates DataLoader instance for the given dataset.\"\"\"\n \n cfg[\"raster_params\"][\"map_type\"] = map_type\n rasterizer = build_rasterizer(cfg, dm)\n chunk_data = ChunkedDataset(zarr_data).open()\n agent_data = AgentDataset(cfg, chunk_data, rasterizer)\n \n # Getting Subset of the dataset.\n subset_data = torch.utils.data.Subset(agent_data, range(0, subset_len))\n \n dataloader = DataLoader(subset_data, \n batch_size=config[\"batch_size\"],\n num_workers=config[\"num_workers\"],\n shuffle=config[\"shuffle\"]\n )\n return dataloader\n\ndef train():\n device = xm.xla_device()\n model = LyftModel(cfg).to(device)\n fitter = TPUFitter(model, device)\n \n xm.master_print(\"Preparing the dataloader..\")\n train_dataloader = get_dataloader(cfg[\"train_data_loader\"], dm.require(\"scenes/train.zarr\"), 5000)\n val_dataloader = get_dataloader(cfg[\"valid_data_loader\"], dm.require(\"scenes/validate.zarr\"), 300)\n \n xm.master_print(\"Training the model..\")\n fitter.fit(train_dataloader, val_dataloader)\n \n return fitter","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"# Train the Model"},{"metadata":{"trusted":true},"cell_type":"code","source":"model = train()","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"\nHere I only used a subset of the original dataset to train faster as this is an illustration based notebook. However you must use a larger subset or the complete dataset for training of the model.
"},{"metadata":{},"cell_type":"markdown","source":"Just a note that you have saved your progress when you were iterating through the above code so now you could simply load the model to further train it. Please also note the TPU requirements for this competition.😏🙄 "},{"metadata":{},"cell_type":"markdown","source":"# Explanations\n\nIn the following notebook we would be looking at explanations of some important code areas which should be into your knowledge both for your experience as well as to completely understand this notebook.
"},{"metadata":{},"cell_type":"markdown","source":"## AdamW Optimizer\n\nAssuming you're having a good knowledge about [Adam](https://arxiv.org/abs/1412.6980) optimizer, we would head directly towards understanding of the [AdamW](https://www.fast.ai/2018/07/02/adam-weight-decay/) optimizer.\n\nSo, `AdamW` optimizer is equipped with an extra **weight decay** parameter (a hyperparameter) that it uses along with the general implementation of `Adam` algorithm. Note that apart from the case of vanilla SGD, L2 regularization (which Adam implements generally) and weight decay(which AdamW implements) are two different things.\n\nThe formula for L2 regularization would be-\n\n\n\nAnd formula for weight decay would be-\n\n\n\nSo a basic difference could be understood as that we add a penalizing term to the loss in the `L2-regularization` whereas we subtract an additional fraction of the weight from itself when updating it. Here `wd` is simply a hyperparameter which measures how larger decay in the weight you want at each update. \n\nThat's why we added a `weight_decay` parameter here (from above code)-\n \n \nHere we are simply specifying a `weight_decay` value of 0 for the parameter for which we don't want the weights to decay additionally. Finally we pass this configuration to `AdamW` optimizer. With this we just combined the power of `Adam` optimization with weight_decay so that weights are learned and updated much more efficiently.\n\n**Read more about AdamW from [here](https://www.fast.ai/2018/07/02/adam-weight-decay/).**","attachments":{"L2.PNG":{"image/png":"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"},"image.png":{"image/png":"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"},"weight_decay.PNG":{"image/png":"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"}}},{"metadata":{"trusted":true},"cell_type":"markdown","source":"Till now I find that only AdamW optimizer should be explained and the rest of the code is easily understandable. However you could ask for any explanations in the comments or reach me out through mail and if I would be knowing that, I would be delighted to help you. Thanks!
"},{"metadata":{},"cell_type":"markdown","source":"Hope You Guys Liked my notebook.
\nPlease upvote if you did to keep me motivated.😁😀
"},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat":4,"nbformat_minor":4}