graph-mode-vs-eager-mode-in-tensorflow.ipynb

{"cells":[{"source":"<a href=\"https://www.kaggle.com/code/peremartramanonellas/improve-tensorflow-performance-with-graph-mode?scriptVersionId=124612302\" target=\"_blank\"><img align=\"left\" alt=\"Kaggle\" title=\"Open in Kaggle\" src=\"https://kaggle.com/static/images/open-in-kaggle.svg\"></a>","metadata":{},"cell_type":"markdown"},{"cell_type":"code","execution_count":1,"id":"1abaa1b7","metadata":{"_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","_kg_hide-input":true,"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","execution":{"iopub.execute_input":"2023-04-04T21:11:26.55822Z","iopub.status.busy":"2023-04-04T21:11:26.557573Z","iopub.status.idle":"2023-04-04T21:11:26.577577Z","shell.execute_reply":"2023-04-04T21:11:26.576719Z"},"papermill":{"duration":0.040584,"end_time":"2023-04-04T21:11:26.581939","exception":false,"start_time":"2023-04-04T21:11:26.541355","status":"completed"},"tags":[]},"outputs":[{"data":{"text/html":["\n","<style>\n","font-family: monospace;\n",".output_png {\n","    display: table-cell;\n","    text-align: center;\n","    vertical-align: middle;\n","    horizontal-align: middle;\n","}\n","h1 {\n","    text-align: center;\n","    background-color: AliceBlue;\n","    padding: 10px;\n","    margin: 0;\n","    font-family: monospace;\n","    color:DimGray;\n","    border-radius: 10px\n","}\n","\n","h2 {\n","    text-align: center;\n","    background-color: HoneyDew;\n","    padding: 10px;\n","    margin: 0;\n","    font-family: monospace;\n","    color:DimGray;\n","    border-radius: 10px\n","}\n","\n","h3 {\n","    text-align: center;\n","    background-color: MintCream;\n","    padding: 10px;\n","    margin: 0;\n","    font-family: monospace;\n","    color:DimGray;\n","    border-radius: 10px\n","}\n","\n","\n","body, p {\n","    font-family: monospace;\n","    font-size: 15px;\n","    color: charcoal;\n","}\n","div {\n","    font-size: 14px;\n","    margin: 0;\n","\n","}\n","\n","h4 {\n","    padding: 0px;\n","    margin: 0;\n","    font-family: monospace;\n","    color: purple;\n","}\n","</style>\n"],"text/plain":["<IPython.core.display.HTML object>"]},"execution_count":1,"metadata":{},"output_type":"execute_result"}],"source":["from IPython.core.display import HTML\n","HTML(\"\"\"\n","<style>\n","font-family: monospace;\n",".output_png {\n","    display: table-cell;\n","    text-align: center;\n","    vertical-align: middle;\n","    horizontal-align: middle;\n","}\n","h1 {\n","    text-align: center;\n","    background-color: AliceBlue;\n","    padding: 10px;\n","    margin: 0;\n","    font-family: monospace;\n","    color:DimGray;\n","    border-radius: 10px\n","}\n","\n","h2 {\n","    text-align: center;\n","    background-color: HoneyDew;\n","    padding: 10px;\n","    margin: 0;\n","    font-family: monospace;\n","    color:DimGray;\n","    border-radius: 10px\n","}\n","\n","h3 {\n","    text-align: center;\n","    background-color: MintCream;\n","    padding: 10px;\n","    margin: 0;\n","    font-family: monospace;\n","    color:DimGray;\n","    border-radius: 10px\n","}\n","\n","\n","body, p {\n","    font-family: monospace;\n","    font-size: 15px;\n","    color: charcoal;\n","}\n","div {\n","    font-size: 14px;\n","    margin: 0;\n","\n","}\n","\n","h4 {\n","    padding: 0px;\n","    margin: 0;\n","    font-family: monospace;\n","    color: purple;\n","}\n","</style>\n","\"\"\")\n","\n"]},{"cell_type":"markdown","id":"705997b3","metadata":{"papermill":{"duration":0.008174,"end_time":"2023-04-04T21:11:26.600594","exception":false,"start_time":"2023-04-04T21:11:26.59242","status":"completed"},"tags":[]},"source":["# Graph Mode vs Eager Mode. Improve drastically and easily the performance of your TensorFlow Notebooks. \n","The code in graph mode is much more efficient and can give us a huge performance boost.\n","\n","At the beginning of this notebook, there is a brief introduction to graph mode. Initially, we will see their differences and a few examples of how the code looks in graph mode.  As well as how we can easily go from eager mode to graph mode.\n","\n","After the brief introduction, we will move on to solving a classification problem with the human_vs_horses dataset. Running the solution in both eager and graphical modes, so we can easily compare performance.\n","\n","*The notebook is ready to work also with the cats_vs_dogs Dataset. But the memory limitations we encountered in Kaggle prevent us from running it with such a large Dataset. In case you want to test with other Datasets or even with a more complex model, you can download the notebook. This is configured to run without issues in a local environment.*"]},{"cell_type":"markdown","id":"8ace6c4b","metadata":{"execution":{"iopub.execute_input":"2022-10-26T20:50:41.188836Z","iopub.status.busy":"2022-10-26T20:50:41.18834Z","iopub.status.idle":"2022-10-26T20:50:41.214108Z","shell.execute_reply":"2022-10-26T20:50:41.2132Z","shell.execute_reply.started":"2022-10-26T20:50:41.188735Z"},"papermill":{"duration":0.008113,"end_time":"2023-04-04T21:11:26.616988","exception":false,"start_time":"2023-04-04T21:11:26.608875","status":"completed"},"tags":[]},"source":["## brief introduction to graph mode & differences with eager mode\n","To simplify, the eager mode is more intuitive, easier to code and debug, and is the natural way to develop. The graph mode is better for performance. \n","\n","Let's look at some code, in eager and graph mode. \n","\n","#### Eager Mode\n","def func(x):\n","\n","    if x > 0:\n","    \n","        x = x + 1\n","        \n","    return x\n","\n","#### Graph Mode\n","def func(x):\n","\n","    def if_true():\n","    \n","        return x + 1\n","        \n","    def if_false():\n","    \n","        return x\n","        \n","    x = tf.cond(tf.greater(x, 0), if_true, if_false)\n","    \n","    return x\n","    \n","As you can see, the code in graph mode is much more difficult to write and understand than the code in eager mode. But there is great news. We do not need to write the code in graph mode. Autograph does it for us.\n","\n","Another great news is that the two types of code can coexist. So, we can use graph mode only for those functions that are expensive in time. Two points where they are usually used in the execution of custom models, and in the transformation of data.\n","\n","### Trying the autograph. \n","We are going to use the autograph to convert some functions from eager to graph. \n"]},{"cell_type":"code","execution_count":2,"id":"b35cdcfe","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:26.636527Z","iopub.status.busy":"2023-04-04T21:11:26.636175Z","iopub.status.idle":"2023-04-04T21:11:31.066849Z","shell.execute_reply":"2023-04-04T21:11:31.065871Z"},"papermill":{"duration":4.442597,"end_time":"2023-04-04T21:11:31.069252","exception":false,"start_time":"2023-04-04T21:11:26.626655","status":"completed"},"tags":[]},"outputs":[],"source":["import tensorflow as tf\n","import time"]},{"cell_type":"code","execution_count":3,"id":"b3ae7cf5","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:31.090001Z","iopub.status.busy":"2023-04-04T21:11:31.088796Z","iopub.status.idle":"2023-04-04T21:11:31.094427Z","shell.execute_reply":"2023-04-04T21:11:31.093561Z"},"papermill":{"duration":0.016987,"end_time":"2023-04-04T21:11:31.096374","exception":false,"start_time":"2023-04-04T21:11:31.079387","status":"completed"},"tags":[]},"outputs":[],"source":["@tf.function\n","def func(X):\n","    if x > 0: \n","        x = x + 1\n","    return x"]},{"cell_type":"markdown","id":"5688448e","metadata":{"papermill":{"duration":0.008235,"end_time":"2023-04-04T21:11:31.113106","exception":false,"start_time":"2023-04-04T21:11:31.104871","status":"completed"},"tags":[]},"source":["We have decorated our function with ***@tf.function***. We don't see the graph code, but when we call it, the function will run in graph mode. \n","\n","Now let's see the generated code."]},{"cell_type":"code","execution_count":4,"id":"2b4f2477","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:31.131777Z","iopub.status.busy":"2023-04-04T21:11:31.130828Z","iopub.status.idle":"2023-04-04T21:11:31.176495Z","shell.execute_reply":"2023-04-04T21:11:31.175006Z"},"papermill":{"duration":0.057343,"end_time":"2023-04-04T21:11:31.178777","exception":false,"start_time":"2023-04-04T21:11:31.121434","status":"completed"},"tags":[]},"outputs":[{"name":"stdout","output_type":"stream","text":["def tf__func(X):\n","    with ag__.FunctionScope('func', 'fscope', ag__.ConversionOptions(recursive=True, user_requested=True, optional_features=(), internal_convert_user_code=True)) as fscope:\n","        do_return = False\n","        retval_ = ag__.UndefinedReturnValue()\n","\n","        def get_state():\n","            return (x,)\n","\n","        def set_state(vars_):\n","            nonlocal x\n","            (x,) = vars_\n","\n","        def if_body():\n","            nonlocal x\n","            x = (ag__.ld(x) + 1)\n","\n","        def else_body():\n","            nonlocal x\n","            pass\n","        x = ag__.Undefined('x')\n","        ag__.if_stmt((ag__.ld(x) > 0), if_body, else_body, get_state, set_state, ('x',), 1)\n","        try:\n","            do_return = True\n","            retval_ = ag__.ld(x)\n","        except:\n","            do_return = False\n","            raise\n","        return fscope.ret(retval_, do_return)\n","\n"]}],"source":["print (tf.autograph.to_code(func.python_function))"]},{"cell_type":"markdown","id":"6579a6c9","metadata":{"papermill":{"duration":0.008507,"end_time":"2023-04-04T21:11:31.196064","exception":false,"start_time":"2023-04-04T21:11:31.187557","status":"completed"},"tags":[]},"source":["Ok! That's the code autogenerated in graph mode. As you can see is even more complex that the one that i writed as a sample. But the structure is similar.  The functions used are defined at the beginning, and we execute the code inside a function calling the functions witted previously. \n","\n","You may ask,why can't I decorate every function in the notebook?  Easy, not all works fine in graph mode, there are some tricks that we must know. \n","\n","\n","\n","**If we need to use Tensor variables, they can't be declared inside the function.**"]},{"cell_type":"code","execution_count":5,"id":"ed3bf23f","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:31.215112Z","iopub.status.busy":"2023-04-04T21:11:31.214168Z","iopub.status.idle":"2023-04-04T21:11:31.218847Z","shell.execute_reply":"2023-04-04T21:11:31.217982Z"},"papermill":{"duration":0.01637,"end_time":"2023-04-04T21:11:31.220896","exception":false,"start_time":"2023-04-04T21:11:31.204526","status":"completed"},"scrolled":true,"tags":[]},"outputs":[],"source":["#This function will fail as it does not support declaring tf variables in the body. \n","#if you want to execute and test it, remove the comments. \n","\n","#@tf.function\n","#def f(x):\n","#    v = tf.Variable(1.0)\n","#    return v.assign_add(x)\n","\n","#print(f(5))"]},{"cell_type":"code","execution_count":6,"id":"6f2305a1","metadata":{"_kg_hide-output":true,"execution":{"iopub.execute_input":"2023-04-04T21:11:31.240278Z","iopub.status.busy":"2023-04-04T21:11:31.239413Z","iopub.status.idle":"2023-04-04T21:11:34.119273Z","shell.execute_reply":"2023-04-04T21:11:34.118272Z"},"papermill":{"duration":2.891598,"end_time":"2023-04-04T21:11:34.121206","exception":false,"start_time":"2023-04-04T21:11:31.229608","status":"completed"},"scrolled":true,"tags":[]},"outputs":[{"name":"stdout","output_type":"stream","text":["tf.Tensor(6.0, shape=(), dtype=float32)\n"]}],"source":["#For it to work, we just have to remove the variable and declare it outside the function\n","v = tf.Variable(1.0)\n","@tf.function\n","def f(x):\n","    return v.assign_add(x)\n","\n","print(f(5))"]},{"cell_type":"markdown","id":"d52fb9de","metadata":{"papermill":{"duration":0.008498,"end_time":"2023-04-04T21:11:34.138825","exception":false,"start_time":"2023-04-04T21:11:34.130327","status":"completed"},"tags":[]},"source":["Another thing to consider is how the print() function works. **In graphical mode, the print function, will only execute once,** no matter how many times we call it in a loop. \n","\n","We must replace it with ***tf.print()***."]},{"cell_type":"code","execution_count":7,"id":"dc436870","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:34.157212Z","iopub.status.busy":"2023-04-04T21:11:34.156872Z","iopub.status.idle":"2023-04-04T21:11:34.200981Z","shell.execute_reply":"2023-04-04T21:11:34.200006Z"},"papermill":{"duration":0.055728,"end_time":"2023-04-04T21:11:34.20302","exception":false,"start_time":"2023-04-04T21:11:34.147292","status":"completed"},"tags":[]},"outputs":[{"name":"stdout","output_type":"stream","text":["with print\n","with tf.print\n","with tf.print\n","with tf.print\n","with tf.print\n","with tf.print\n"]}],"source":["@tf.function\n","def print_test(): \n","    tf.print(\"with tf.print\")\n","    print(\"with print\")\n","    \n","\n","for i in range(5):\n","    print_test()"]},{"cell_type":"markdown","id":"6c5192b2","metadata":{"papermill":{"duration":0.008685,"end_time":"2023-04-04T21:11:34.220589","exception":false,"start_time":"2023-04-04T21:11:34.211904","status":"completed"},"tags":[]},"source":["Something similar happens with ***ASSERT***, which we must replace with the corresponding ***tf.debugging.assert_***"]},{"cell_type":"markdown","id":"76ec2844","metadata":{"execution":{"iopub.execute_input":"2022-10-26T21:50:11.854769Z","iopub.status.busy":"2022-10-26T21:50:11.853531Z","iopub.status.idle":"2022-10-26T21:50:11.859706Z","shell.execute_reply":"2022-10-26T21:50:11.858203Z","shell.execute_reply.started":"2022-10-26T21:50:11.854717Z"},"papermill":{"duration":0.008526,"end_time":"2023-04-04T21:11:34.237818","exception":false,"start_time":"2023-04-04T21:11:34.229292","status":"completed"},"tags":[]},"source":["# Compare eager vs graph using the Cats vs Dogs dataset or Human vs Horses. \n","We are going to use a custom model to solve an image classification problem using the horse_or_humans dataset, which we can find in TensorFlow Datasets.\n","\n","I've decided to use a Dataset from the TensorFlow Datasets database to make it easier to download the notebook and run tests on our own computer, with other datasets like cats_vs_dogs, or even with more complex models like the ones we found in TensorFlow HUB."]},{"cell_type":"code","execution_count":8,"id":"61759d1c","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:34.257008Z","iopub.status.busy":"2023-04-04T21:11:34.256617Z","iopub.status.idle":"2023-04-04T21:11:36.402206Z","shell.execute_reply":"2023-04-04T21:11:36.401211Z"},"papermill":{"duration":2.157952,"end_time":"2023-04-04T21:11:36.404631","exception":false,"start_time":"2023-04-04T21:11:34.246679","status":"completed"},"tags":[]},"outputs":[],"source":["import tensorflow_datasets as tfds\n","import tensorflow_hub as hub\n","import matplotlib.pyplot as plt\n","import numpy as np\n","import time"]},{"cell_type":"markdown","id":"edfae44f","metadata":{"papermill":{"duration":0.008898,"end_time":"2023-04-04T21:11:36.423181","exception":false,"start_time":"2023-04-04T21:11:36.414283","status":"completed"},"tags":[]},"source":["## Import & data\n","We retrieved the data from a TensorFlow dataset instead of Kaggle to make it easier to run the notebook in other environments and make it easier to test."]},{"cell_type":"code","execution_count":9,"id":"0084ba7f","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:36.442743Z","iopub.status.busy":"2023-04-04T21:11:36.442079Z","iopub.status.idle":"2023-04-04T21:11:40.288265Z","shell.execute_reply":"2023-04-04T21:11:40.286861Z"},"papermill":{"duration":3.858515,"end_time":"2023-04-04T21:11:40.290582","exception":false,"start_time":"2023-04-04T21:11:36.432067","status":"completed"},"tags":[]},"outputs":[{"name":"stdout","output_type":"stream","text":["\u001b[1mDownloading and preparing dataset 153.59 MiB (download: 153.59 MiB, generated: Unknown size, total: 153.59 MiB) to ./data/horses_or_humans/3.0.0...\u001b[0m\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"2a87c9dc4fed481d86a1bb208c52b5f2","version_major":2,"version_minor":0},"text/plain":["Dl Completed...: 0 url [00:00, ? url/s]"]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"0696ada07e5541f08e39253659eb71e4","version_major":2,"version_minor":0},"text/plain":["Dl Size...: 0 MiB [00:00, ? MiB/s]"]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"d3e5a3b930e145f0bbbda44bae0972e0","version_major":2,"version_minor":0},"text/plain":["Generating splits...:   0%|          | 0/2 [00:00<?, ? splits/s]"]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"64d19d80c2ef4925a90ce4e72ecf604a","version_major":2,"version_minor":0},"text/plain":["Generating train examples...:   0%|          | 0/1027 [00:00<?, ? examples/s]"]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"0346cc08825c455b83c383747f27df02","version_major":2,"version_minor":0},"text/plain":["Shuffling horses_or_humans-train.tfrecord...:   0%|          | 0/1027 [00:00<?, ? examples/s]"]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"e630f4e736344ab488cd4d9685d35cff","version_major":2,"version_minor":0},"text/plain":["Generating test examples...:   0%|          | 0/256 [00:00<?, ? examples/s]"]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"b4db46091dc6438ab8dfb43ac2eae42f","version_major":2,"version_minor":0},"text/plain":["Shuffling horses_or_humans-test.tfrecord...:   0%|          | 0/256 [00:00<?, ? examples/s]"]},"metadata":{},"output_type":"display_data"},{"name":"stdout","output_type":"stream","text":["\u001b[1mDataset horses_or_humans downloaded and prepared to ./data/horses_or_humans/3.0.0. Subsequent calls will reuse this data.\u001b[0m\n"]}],"source":["#Choose the Dataset you want to use. \n","#In Kaggle I do not recommend cats_vs_dogs \n","#because we can reach the limit of memory consumption.\n","\n","#setattr(tfds.image_classification.cats_vs_dogs, '_URL',\"https://download.microsoft.com/download/3/E/1/3E1C3F21-ECDB-4869-8368-6DEBA77B919F/kagglecatsanddogs_5340.zip\")\n","#TS_2_DOWNLOAD = 'cats_vs_dogs'\n","\n","TS_2_DOWNLOAD = 'horses_or_humans'\n","\n","splits, info_cd = tfds.load(TS_2_DOWNLOAD, \n","                         as_supervised=True, \n","                         with_info=True, \n","                         split=['train[:70%]', 'train[70%:90%]', 'train[90%:]'], \n","                         data_dir='./data')"]},{"cell_type":"code","execution_count":10,"id":"385a51b5","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:40.312165Z","iopub.status.busy":"2023-04-04T21:11:40.311585Z","iopub.status.idle":"2023-04-04T21:11:40.317787Z","shell.execute_reply":"2023-04-04T21:11:40.316776Z"},"papermill":{"duration":0.019032,"end_time":"2023-04-04T21:11:40.31985","exception":false,"start_time":"2023-04-04T21:11:40.300818","status":"completed"},"tags":[]},"outputs":[],"source":["num_examples_cd = info_cd.splits['train'].num_examples\n","num_classes_cd = info_cd.features['label'].num_classes\n","\n","(train_examples, validation_examples, test_examples) = splits\n","num_examples = num_examples_cd\n","num_classes = num_classes_cd"]},{"cell_type":"markdown","id":"90715ccc","metadata":{"papermill":{"duration":0.009489,"end_time":"2023-04-04T21:11:40.339203","exception":false,"start_time":"2023-04-04T21:11:40.329714","status":"completed"},"tags":[]},"source":["## Eager vs Graph"]},{"cell_type":"code","execution_count":11,"id":"5426f99b","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:40.360573Z","iopub.status.busy":"2023-04-04T21:11:40.359693Z","iopub.status.idle":"2023-04-04T21:11:40.364618Z","shell.execute_reply":"2023-04-04T21:11:40.36375Z"},"papermill":{"duration":0.017664,"end_time":"2023-04-04T21:11:40.366553","exception":false,"start_time":"2023-04-04T21:11:40.348889","status":"completed"},"tags":[]},"outputs":[],"source":["BATCH_SIZE = 32\n","IMAGE_SIZE = 224"]},{"cell_type":"markdown","id":"c3c15440","metadata":{"papermill":{"duration":0.009694,"end_time":"2023-04-04T21:11:40.386254","exception":false,"start_time":"2023-04-04T21:11:40.37656","status":"completed"},"tags":[]},"source":["Now I'm going to create two functions to deal with the images. Nothing will be done other than adapt the size and normalize the image.\n","\n","It is an operation that can be heavy, depending on the size of the images and the Dataset. Besides that, it is a function that can be passed effortlessly to Graph mode.\n","\n","The code is the same in both functions. The only difference is the decoration in the map_fn_graph. "]},{"cell_type":"code","execution_count":12,"id":"320d9be9","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:40.407637Z","iopub.status.busy":"2023-04-04T21:11:40.406748Z","iopub.status.idle":"2023-04-04T21:11:40.412255Z","shell.execute_reply":"2023-04-04T21:11:40.411424Z"},"papermill":{"duration":0.018259,"end_time":"2023-04-04T21:11:40.414238","exception":false,"start_time":"2023-04-04T21:11:40.395979","status":"completed"},"tags":[]},"outputs":[],"source":["#Treat the image in eager mode. \n","def map_fn_eager(img, label):\n","    # resize the image\n","    img = tf.image.resize(img, size=[IMAGE_SIZE, IMAGE_SIZE])\n","    # normalize the image\n","    img /= 255.0\n","    return img, label"]},{"cell_type":"code","execution_count":13,"id":"9f87e511","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:40.43552Z","iopub.status.busy":"2023-04-04T21:11:40.434707Z","iopub.status.idle":"2023-04-04T21:11:40.440253Z","shell.execute_reply":"2023-04-04T21:11:40.439433Z"},"papermill":{"duration":0.018184,"end_time":"2023-04-04T21:11:40.442207","exception":false,"start_time":"2023-04-04T21:11:40.424023","status":"completed"},"tags":[]},"outputs":[],"source":["# #Treat the image in graph mode. \n","@tf.function\n","def map_fn_graph(img, label):\n","    # resize the image\n","    img = tf.image.resize(img, size=[IMAGE_SIZE, IMAGE_SIZE])\n","    # normalize the image\n","    img /= 255.0\n","    return img, label"]},{"cell_type":"code","execution_count":14,"id":"77b5b322","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:40.463503Z","iopub.status.busy":"2023-04-04T21:11:40.462708Z","iopub.status.idle":"2023-04-04T21:11:40.468392Z","shell.execute_reply":"2023-04-04T21:11:40.467542Z"},"papermill":{"duration":0.01822,"end_time":"2023-04-04T21:11:40.470362","exception":false,"start_time":"2023-04-04T21:11:40.452142","status":"completed"},"tags":[]},"outputs":[],"source":["# Prepare train dataset by using preprocessing with map_fn_eager or graph, shuffling and batching\n","def prepare_dataset(train_examples, validation_examples, test_examples, num_examples, map_fn, batch_size):\n","    train_ds = train_examples.map(map_fn).shuffle(buffer_size = num_examples).batch(batch_size)\n","    valid_ds = validation_examples.map(map_fn).batch(batch_size)\n","    test_ds = test_examples.map(map_fn).batch(batch_size)\n","    \n","    return train_ds, valid_ds, test_ds"]},{"cell_type":"code","execution_count":15,"id":"c791d5e6","metadata":{"_kg_hide-output":true,"execution":{"iopub.execute_input":"2023-04-04T21:11:40.491587Z","iopub.status.busy":"2023-04-04T21:11:40.490781Z","iopub.status.idle":"2023-04-04T21:11:40.559731Z","shell.execute_reply":"2023-04-04T21:11:40.558751Z"},"papermill":{"duration":0.081912,"end_time":"2023-04-04T21:11:40.561898","exception":false,"start_time":"2023-04-04T21:11:40.479986","status":"completed"},"tags":[]},"outputs":[{"name":"stdout","output_type":"stream","text":["Time spend: 0.062294721603393555\n"]}],"source":["start_time = time.time()\n","train_ds_eager, valid_ds_eager, test_ds_eager = prepare_dataset(train_examples, \n","                                                                validation_examples, \n","                                                                test_examples, \n","                                                                num_examples, \n","                                                                map_fn_eager, BATCH_SIZE)\n","end_time = time.time()\n","print (\"Time spend:\",  end_time - start_time)"]},{"cell_type":"code","execution_count":16,"id":"b51d614f","metadata":{"_kg_hide-output":true,"execution":{"iopub.execute_input":"2023-04-04T21:11:40.584441Z","iopub.status.busy":"2023-04-04T21:11:40.582729Z","iopub.status.idle":"2023-04-04T21:11:40.647424Z","shell.execute_reply":"2023-04-04T21:11:40.646289Z"},"papermill":{"duration":0.078374,"end_time":"2023-04-04T21:11:40.650157","exception":false,"start_time":"2023-04-04T21:11:40.571783","status":"completed"},"tags":[]},"outputs":[{"name":"stdout","output_type":"stream","text":["Time spend: 0.05785417556762695\n"]}],"source":["start_time = time.time()\n","train_ds_graph, valid_ds_graph, test_ds_graph = prepare_dataset(train_examples, \n","                                                                validation_examples, \n","                                                                test_examples, \n","                                                                num_examples, \n","                                                                map_fn_graph, BATCH_SIZE)\n","end_time = time.time()\n","print (\"Time spend:\",  end_time - start_time)"]},{"cell_type":"markdown","id":"5d026792","metadata":{"papermill":{"duration":0.009573,"end_time":"2023-04-04T21:11:40.669935","exception":false,"start_time":"2023-04-04T21:11:40.660362","status":"completed"},"tags":[]},"source":["With this data set and using the Kaggle GPUs, the elapsed time is minimal. Less than one second. Even so, an improvement of 30% is appreciated over time. \n","\n","With a larger dataset, or with quality images like the ones we could find in medical analysis problems, the time saved could be significant."]},{"cell_type":"markdown","id":"e2513dab","metadata":{"execution":{"iopub.execute_input":"2022-10-27T20:39:20.616778Z","iopub.status.busy":"2022-10-27T20:39:20.615708Z","iopub.status.idle":"2022-10-27T20:39:20.6231Z","shell.execute_reply":"2022-10-27T20:39:20.621545Z","shell.execute_reply.started":"2022-10-27T20:39:20.616726Z"},"papermill":{"duration":0.009522,"end_time":"2023-04-04T21:11:40.689256","exception":false,"start_time":"2023-04-04T21:11:40.679734","status":"completed"},"tags":[]},"source":["# Executing the model"]},{"cell_type":"markdown","id":"65f02d81","metadata":{"papermill":{"duration":0.00947,"end_time":"2023-04-04T21:11:40.708874","exception":false,"start_time":"2023-04-04T21:11:40.699404","status":"completed"},"tags":[]},"source":["If you want you can use some models from TensorFlow HUB, just uncomment the lines commented, and comment the line with the sequential model. \n","\n","With bigger models, it is easier to see the performance differences, but in Kaggle we have some limitations, and we can run out of memory. \n","\n","But if you decide to download the notebook, and execute in your computer, I think it would be a good idea to test the performance with several models, to get a better view of the improvement obtained."]},{"cell_type":"code","execution_count":17,"id":"4b35595b","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:40.730701Z","iopub.status.busy":"2023-04-04T21:11:40.730317Z","iopub.status.idle":"2023-04-04T21:11:40.831577Z","shell.execute_reply":"2023-04-04T21:11:40.829588Z"},"papermill":{"duration":0.1149,"end_time":"2023-04-04T21:11:40.83386","exception":false,"start_time":"2023-04-04T21:11:40.71896","status":"completed"},"tags":[]},"outputs":[{"name":"stdout","output_type":"stream","text":["Model: \"sequential\"\n","_________________________________________________________________\n","Layer (type)                 Output Shape              Param #   \n","=================================================================\n","conv2d (Conv2D)              (None, 221, 221, 16)      784       \n","_________________________________________________________________\n","max_pooling2d (MaxPooling2D) (None, 110, 110, 16)      0         \n","_________________________________________________________________\n","dropout (Dropout)            (None, 110, 110, 16)      0         \n","_________________________________________________________________\n","conv2d_1 (Conv2D)            (None, 107, 107, 32)      8224      \n","_________________________________________________________________\n","max_pooling2d_1 (MaxPooling2 (None, 53, 53, 32)        0         \n","_________________________________________________________________\n","dropout_1 (Dropout)          (None, 53, 53, 32)        0         \n","_________________________________________________________________\n","conv2d_2 (Conv2D)            (None, 50, 50, 64)        32832     \n","_________________________________________________________________\n","max_pooling2d_2 (MaxPooling2 (None, 25, 25, 64)        0         \n","_________________________________________________________________\n","dropout_2 (Dropout)          (None, 25, 25, 64)        0         \n","_________________________________________________________________\n","flatten (Flatten)            (None, 40000)             0         \n","_________________________________________________________________\n","dense (Dense)                (None, 512)               20480512  \n","_________________________________________________________________\n","dense_1 (Dense)              (None, 2)                 1026      \n","=================================================================\n","Total params: 20,523,378\n","Trainable params: 20,523,378\n","Non-trainable params: 0\n","_________________________________________________________________\n"]}],"source":["#MODULE_HANDLE = 'https://tfhub.dev/tensorflow/resnet_50/feature_vector/1'\n","#MODULE_HANDLE = 'https://tfhub.dev/google/imagenet/resnet_v2_152/classification/5'\n","#model = tf.keras.Sequential([\n","#    hub.KerasLayer(MODULE_HANDLE, input_shape=(IMAGE_SIZE, IMAGE_SIZE, 3)),\n","#    tf.keras.layers.Dense(num_classes, activation='softmax')\n","#])\n","\n","model = tf.keras.models.Sequential([\n","            tf.keras.layers.Conv2D(16, (4,4), activation=\"relu\", input_shape=(IMAGE_SIZE, IMAGE_SIZE, 3)),\n","            tf.keras.layers.MaxPooling2D(2,2),\n","            tf.keras.layers.Dropout(0.2),   \n","            tf.keras.layers.Conv2D(32, (4,4), activation=\"relu\"),\n","            tf.keras.layers.MaxPooling2D(2,2),  \n","            tf.keras.layers.Dropout(0.2),  \n","            tf.keras.layers.Conv2D(64, (4,4), activation=\"relu\"),\n","            tf.keras.layers.MaxPooling2D(2,2),  \n","            tf.keras.layers.Dropout(0.5), \n","            tf.keras.layers.Flatten(), \n","            tf.keras.layers.Dense(512, activation=\"relu\"), \n","            tf.keras.layers.Dense(2, activation=\"softmax\")])\n","model.summary()"]},{"cell_type":"code","execution_count":18,"id":"db1874b7","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:40.855634Z","iopub.status.busy":"2023-04-04T21:11:40.85502Z","iopub.status.idle":"2023-04-04T21:11:40.866738Z","shell.execute_reply":"2023-04-04T21:11:40.865866Z"},"papermill":{"duration":0.024649,"end_time":"2023-04-04T21:11:40.868704","exception":false,"start_time":"2023-04-04T21:11:40.844055","status":"completed"},"tags":[]},"outputs":[],"source":["optimizer = tf.keras.optimizers.Adam()\n","train_accuracy = tf.keras.metrics.SparseCategoricalAccuracy()\n","val_accuracy = tf.keras.metrics.SparseCategoricalAccuracy()\n","train_loss = tf.keras.losses.SparseCategoricalCrossentropy()\n","val_loss = tf.keras.losses.SparseCategoricalCrossentropy()"]},{"cell_type":"code","execution_count":19,"id":"02700f89","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:40.902467Z","iopub.status.busy":"2023-04-04T21:11:40.902028Z","iopub.status.idle":"2023-04-04T21:11:40.91493Z","shell.execute_reply":"2023-04-04T21:11:40.913283Z"},"papermill":{"duration":0.036664,"end_time":"2023-04-04T21:11:40.918812","exception":false,"start_time":"2023-04-04T21:11:40.882148","status":"completed"},"tags":[]},"outputs":[],"source":["# Custom training step. This function is executed in each step each epoch. \n","def train_one_step(model, optimizer, x, y, train_loss, train_accuracy):\n","    with tf.GradientTape() as tape:\n","        # Run the model on input x to get predictions\n","        predictions = model(x)\n","        # Compute the training loss using `train_loss`, passing in the true y and the predicted y\n","        loss = train_loss(y, predictions)\n","\n","    # Using the tape and loss, compute the gradients on model variables using tape.gradient\n","    grads = tape.gradient(loss, model.trainable_variables)\n","    \n","    # Zip the gradients and model variables, and then apply the result on the optimizer\n","    optimizer.apply_gradients(zip(grads, model.trainable_variables))\n","\n","    # Call the train accuracy object on ground truth and predictions\n","    train_accuracy(y, predictions)\n","    return loss"]},{"cell_type":"markdown","id":"f4df54be","metadata":{"papermill":{"duration":0.024137,"end_time":"2023-04-04T21:11:40.96925","exception":false,"start_time":"2023-04-04T21:11:40.945113","status":"completed"},"tags":[]},"source":["Now I am going to define two functions, one to execute in eager mode and the other in graph mode.\n","\n","Both functions will call ***train_on_step***. You can see that it isn't decorated, but it doesn't need to be, since TensorFlow not only transforms the decorated functions, but also all the ones called from it.\n","\n","That is, ***train_one_step*** will run in eager mode when called from ***train_eager***, and in graph mode when called from ***train_graph***."]},{"cell_type":"code","execution_count":20,"id":"9cd55afa","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:41.000036Z","iopub.status.busy":"2023-04-04T21:11:40.999598Z","iopub.status.idle":"2023-04-04T21:11:41.008832Z","shell.execute_reply":"2023-04-04T21:11:41.007886Z"},"papermill":{"duration":0.032344,"end_time":"2023-04-04T21:11:41.016355","exception":false,"start_time":"2023-04-04T21:11:40.984011","status":"completed"},"tags":[]},"outputs":[],"source":["\n","def train_eager(model, optimizer, epochs, train_ds, train_loss, train_accuracy, valid_ds, val_loss, val_accuracy):\n","    step = 0\n","    loss = 0.0\n","    for epoch in range(epochs):\n","        for x, y in train_ds:\n","            # training step number increments at each iteration\n","            step += 1\n","\n","            # Run one training step by passing appropriate model parameters\n","            # required by the function and finally get the loss to report the results\n","            loss = train_one_step(model, optimizer, x, y, train_loss, train_accuracy)\n","\n","            # Use tf.print to report your results.\n","            # Print the training step number, loss and accuracy\n","            print('Step', step, \n","                   ': train loss', loss, \n","                   '; train accuracy', train_accuracy.result())\n","\n","        for x, y in valid_ds:\n","            # Call the model on the batches of inputs x and get the predictions\n","            y_pred = model(x)\n","            loss = val_loss(y, y_pred)\n","            val_accuracy(y, y_pred)\n","        # Print the validation loss and accuracy\n","        \n","        tf.print('val loss', loss, '; val accuracy', val_accuracy.result())"]},{"cell_type":"markdown","id":"42546bf9","metadata":{"papermill":{"duration":0.014599,"end_time":"2023-04-04T21:11:41.04607","exception":false,"start_time":"2023-04-04T21:11:41.031471","status":"completed"},"tags":[]},"source":["The ***train_graph*** function has a small modification. \n","\n","I have replaced the call to the ***print*** function with a call to ***tf.print***.  \n","Since as we have seen at the beginning of the notebook, print is only executed once, and we need it to print the values at each step, and at each epoch."]},{"cell_type":"code","execution_count":21,"id":"131302cb","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:11:41.077194Z","iopub.status.busy":"2023-04-04T21:11:41.076757Z","iopub.status.idle":"2023-04-04T21:11:41.088812Z","shell.execute_reply":"2023-04-04T21:11:41.087601Z"},"papermill":{"duration":0.0316,"end_time":"2023-04-04T21:11:41.092154","exception":false,"start_time":"2023-04-04T21:11:41.060554","status":"completed"},"tags":[]},"outputs":[],"source":["@tf.function\n","def train_graph(model, optimizer, epochs, train_ds, train_loss, train_accuracy, valid_ds, val_loss, val_accuracy):\n","    step = 0\n","    loss = 0.0\n","    for epoch in range(epochs):\n","        for x, y in train_ds:\n","            # training step number increments at each iteration\n","            step += 1\n","\n","            # Run one training step by passing appropriate model parameters\n","            # required by the function and finally get the loss to report the results\n","            loss = train_one_step(model, optimizer, x, y, train_loss, train_accuracy)\n","\n","            # Use tf.print to report your results.\n","            # Print the training step number, loss and accuracy\n","            tf.print('Step', step, \n","                   ': train loss', loss, \n","                   '; train accuracy', train_accuracy.result())\n","\n","        for x, y in valid_ds:\n","            # Call the model on the batches of inputs x and get the predictions\n","            y_pred = model(x)\n","            loss = val_loss(y, y_pred)\n","            val_accuracy(y, y_pred)\n","\n","        # Print the validation loss and accuracy\n","        tf.print('val loss', loss, '; val accuracy', val_accuracy.result())"]},{"cell_type":"code","execution_count":22,"id":"ed165e03","metadata":{"_kg_hide-output":true,"execution":{"iopub.execute_input":"2023-04-04T21:11:41.13736Z","iopub.status.busy":"2023-04-04T21:11:41.136699Z","iopub.status.idle":"2023-04-04T21:12:01.756896Z","shell.execute_reply":"2023-04-04T21:12:01.755971Z"},"papermill":{"duration":20.637985,"end_time":"2023-04-04T21:12:01.758979","exception":false,"start_time":"2023-04-04T21:11:41.120994","status":"completed"},"scrolled":true,"tags":[]},"outputs":[{"name":"stderr","output_type":"stream","text":["Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n"]},{"name":"stdout","output_type":"stream","text":["Step 1 : train loss tf.Tensor(0.7318816, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.375, shape=(), dtype=float32)\n","Step 2 : train loss tf.Tensor(8.879311, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.515625, shape=(), dtype=float32)\n","Step 3 : train loss tf.Tensor(2.0693283, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.5208333, shape=(), dtype=float32)\n","Step 4 : train loss tf.Tensor(4.102291, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.4921875, shape=(), dtype=float32)\n","Step 5 : train loss tf.Tensor(1.8404226, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.49375, shape=(), dtype=float32)\n","Step 6 : train loss tf.Tensor(0.7700146, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.5104167, shape=(), dtype=float32)\n","Step 7 : train loss tf.Tensor(0.6617205, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.52678573, shape=(), dtype=float32)\n","Step 8 : train loss tf.Tensor(0.72194177, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.52734375, shape=(), dtype=float32)\n","Step 9 : train loss tf.Tensor(0.7641979, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.5138889, shape=(), dtype=float32)\n","Step 10 : train loss tf.Tensor(0.6492306, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.534375, shape=(), dtype=float32)\n","Step 11 : train loss tf.Tensor(0.7340541, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.5255682, shape=(), dtype=float32)\n","Step 12 : train loss tf.Tensor(0.63445956, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.5442708, shape=(), dtype=float32)\n","Step 13 : train loss tf.Tensor(0.61525023, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.5625, shape=(), dtype=float32)\n","Step 14 : train loss tf.Tensor(0.56193703, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.58928573, shape=(), dtype=float32)\n","Step 15 : train loss tf.Tensor(0.62151223, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.58958334, shape=(), dtype=float32)\n","Step 16 : train loss tf.Tensor(0.5694978, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.5996094, shape=(), dtype=float32)\n","Step 17 : train loss tf.Tensor(0.55128217, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.6139706, shape=(), dtype=float32)\n","Step 18 : train loss tf.Tensor(0.4306438, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.6319444, shape=(), dtype=float32)\n","Step 19 : train loss tf.Tensor(0.5202576, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.6381579, shape=(), dtype=float32)\n","Step 20 : train loss tf.Tensor(0.44210547, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.65, shape=(), dtype=float32)\n","Step 21 : train loss tf.Tensor(0.42333415, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.66220236, shape=(), dtype=float32)\n","Step 22 : train loss tf.Tensor(0.29722947, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.6775568, shape=(), dtype=float32)\n","Step 23 : train loss tf.Tensor(0.35897836, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.6828929, shape=(), dtype=float32)\n"]},{"name":"stderr","output_type":"stream","text":["Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n"]},{"name":"stdout","output_type":"stream","text":["val loss 0.229061604 ; val accuracy 0.89268291\n","Step 24 : train loss tf.Tensor(0.34080732, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.689747, shape=(), dtype=float32)\n","Step 25 : train loss tf.Tensor(0.17720252, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.6998723, shape=(), dtype=float32)\n","Step 26 : train loss tf.Tensor(0.2716396, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.70552146, shape=(), dtype=float32)\n","Step 27 : train loss tf.Tensor(0.2520623, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.7154664, shape=(), dtype=float32)\n","Step 28 : train loss tf.Tensor(0.2938758, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.7224118, shape=(), dtype=float32)\n","Step 29 : train loss tf.Tensor(0.16875121, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.72996706, shape=(), dtype=float32)\n","Step 30 : train loss tf.Tensor(0.35883933, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.73276776, shape=(), dtype=float32)\n","Step 31 : train loss tf.Tensor(0.11052806, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.74051285, shape=(), dtype=float32)\n","Step 32 : train loss tf.Tensor(0.2337941, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.7467726, shape=(), dtype=float32)\n","Step 33 : train loss tf.Tensor(0.30659693, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.7507219, shape=(), dtype=float32)\n","Step 34 : train loss tf.Tensor(0.19119927, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.75630254, shape=(), dtype=float32)\n","Step 35 : train loss tf.Tensor(0.20406711, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.7606528, shape=(), dtype=float32)\n","Step 36 : train loss tf.Tensor(0.12021639, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.76563877, shape=(), dtype=float32)\n","Step 37 : train loss tf.Tensor(0.1244132, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.7720651, shape=(), dtype=float32)\n","Step 38 : train loss tf.Tensor(0.20470458, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.7764804, shape=(), dtype=float32)\n","Step 39 : train loss tf.Tensor(0.101705216, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.78147846, shape=(), dtype=float32)\n","Step 40 : train loss tf.Tensor(0.10736162, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.787015, shape=(), dtype=float32)\n","Step 41 : train loss tf.Tensor(0.15120609, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.7915058, shape=(), dtype=float32)\n","Step 42 : train loss tf.Tensor(0.09921621, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.79577994, shape=(), dtype=float32)\n","Step 43 : train loss tf.Tensor(0.25550154, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.79764533, shape=(), dtype=float32)\n","Step 44 : train loss tf.Tensor(0.13730693, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8015816, shape=(), dtype=float32)\n","Step 45 : train loss tf.Tensor(0.0478521, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8053408, shape=(), dtype=float32)\n","Step 46 : train loss tf.Tensor(0.16322729, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8066759, shape=(), dtype=float32)\n","val loss 0.0223185085 ; val accuracy 0.929268301\n","Step 47 : train loss tf.Tensor(0.09554188, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8095238, shape=(), dtype=float32)\n","Step 48 : train loss tf.Tensor(0.09605236, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8129161, shape=(), dtype=float32)\n","Step 49 : train loss tf.Tensor(0.042726148, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8161669, shape=(), dtype=float32)\n","Step 50 : train loss tf.Tensor(0.020325564, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8199234, shape=(), dtype=float32)\n","Step 51 : train loss tf.Tensor(0.061691232, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.82290363, shape=(), dtype=float32)\n","Step 52 : train loss tf.Tensor(0.23864159, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.82576686, shape=(), dtype=float32)\n","Step 53 : train loss tf.Tensor(0.04734259, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8285199, shape=(), dtype=float32)\n","Step 54 : train loss tf.Tensor(0.25189495, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8305785, shape=(), dtype=float32)\n","Step 55 : train loss tf.Tensor(0.039212048, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8337196, shape=(), dtype=float32)\n","Step 56 : train loss tf.Tensor(0.32099864, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.834471, shape=(), dtype=float32)\n","Step 57 : train loss tf.Tensor(0.2260639, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.83631283, shape=(), dtype=float32)\n","Step 58 : train loss tf.Tensor(0.10847858, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.83863884, shape=(), dtype=float32)\n","Step 59 : train loss tf.Tensor(0.10128307, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.84088457, shape=(), dtype=float32)\n","Step 60 : train loss tf.Tensor(0.15187691, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.84305406, shape=(), dtype=float32)\n","Step 61 : train loss tf.Tensor(0.24169302, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8446298, shape=(), dtype=float32)\n","Step 62 : train loss tf.Tensor(0.28809315, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8451282, shape=(), dtype=float32)\n","Step 63 : train loss tf.Tensor(0.20067543, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.84611505, shape=(), dtype=float32)\n","Step 64 : train loss tf.Tensor(0.073611684, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8480635, shape=(), dtype=float32)\n","Step 65 : train loss tf.Tensor(0.108241774, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8504399, shape=(), dtype=float32)\n","Step 66 : train loss tf.Tensor(0.252618, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.85178053, shape=(), dtype=float32)\n","Step 67 : train loss tf.Tensor(0.3762957, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.85165876, shape=(), dtype=float32)\n","Step 68 : train loss tf.Tensor(0.3680907, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.85247433, shape=(), dtype=float32)\n","Step 69 : train loss tf.Tensor(0.32987252, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.85210943, shape=(), dtype=float32)\n","val loss 0.17172949 ; val accuracy 0.91707319\n","Step 70 : train loss tf.Tensor(0.12814012, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.85381454, shape=(), dtype=float32)\n","Step 71 : train loss tf.Tensor(0.1551708, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8550203, shape=(), dtype=float32)\n","Step 72 : train loss tf.Tensor(0.10541588, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8566356, shape=(), dtype=float32)\n","Step 73 : train loss tf.Tensor(0.04338018, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.85864335, shape=(), dtype=float32)\n","Step 74 : train loss tf.Tensor(0.2834276, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.85843766, shape=(), dtype=float32)\n","Step 75 : train loss tf.Tensor(0.10747389, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8603661, shape=(), dtype=float32)\n","Step 76 : train loss tf.Tensor(0.12292717, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.86140275, shape=(), dtype=float32)\n","Step 77 : train loss tf.Tensor(0.0668216, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.86282635, shape=(), dtype=float32)\n","Step 78 : train loss tf.Tensor(0.07998662, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8646217, shape=(), dtype=float32)\n","Step 79 : train loss tf.Tensor(0.08009334, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8659669, shape=(), dtype=float32)\n","Step 80 : train loss tf.Tensor(0.036061794, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8676764, shape=(), dtype=float32)\n","Step 81 : train loss tf.Tensor(0.03441244, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8693428, shape=(), dtype=float32)\n","Step 82 : train loss tf.Tensor(0.010616228, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.87096775, shape=(), dtype=float32)\n","Step 83 : train loss tf.Tensor(0.035255384, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8721689, shape=(), dtype=float32)\n","Step 84 : train loss tf.Tensor(0.09667572, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8733409, shape=(), dtype=float32)\n","Step 85 : train loss tf.Tensor(0.008153996, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8748595, shape=(), dtype=float32)\n","Step 86 : train loss tf.Tensor(0.11273857, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.87597185, shape=(), dtype=float32)\n","Step 87 : train loss tf.Tensor(0.008669192, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.87742406, shape=(), dtype=float32)\n","Step 88 : train loss tf.Tensor(0.010606245, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.87884265, shape=(), dtype=float32)\n","Step 89 : train loss tf.Tensor(0.005252361, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8802288, shape=(), dtype=float32)\n","Step 90 : train loss tf.Tensor(0.004174004, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8815836, shape=(), dtype=float32)\n","Step 91 : train loss tf.Tensor(0.19738376, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.882209, shape=(), dtype=float32)\n","Step 92 : train loss tf.Tensor(0.3021383, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.88212794, shape=(), dtype=float32)\n","val loss 0.0306860805 ; val accuracy 0.934146345\n","Step 93 : train loss tf.Tensor(0.026486779, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.88342506, shape=(), dtype=float32)\n","Step 94 : train loss tf.Tensor(0.06341301, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.88435376, shape=(), dtype=float32)\n","Step 95 : train loss tf.Tensor(0.049923487, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8852624, shape=(), dtype=float32)\n","Step 96 : train loss tf.Tensor(0.011926647, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8864847, shape=(), dtype=float32)\n","Step 97 : train loss tf.Tensor(0.0122311665, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8876812, shape=(), dtype=float32)\n","Step 98 : train loss tf.Tensor(0.049387, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.88852674, shape=(), dtype=float32)\n","Step 99 : train loss tf.Tensor(0.010879425, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8896774, shape=(), dtype=float32)\n","Step 100 : train loss tf.Tensor(0.013766019, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8908046, shape=(), dtype=float32)\n","Step 101 : train loss tf.Tensor(0.055225134, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.891909, shape=(), dtype=float32)\n","Step 102 : train loss tf.Tensor(0.009503381, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.89299124, shape=(), dtype=float32)\n","Step 103 : train loss tf.Tensor(0.04900175, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.89374226, shape=(), dtype=float32)\n","Step 104 : train loss tf.Tensor(0.0055221883, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8947853, shape=(), dtype=float32)\n","Step 105 : train loss tf.Tensor(0.008871604, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.89580804, shape=(), dtype=float32)\n","Step 106 : train loss tf.Tensor(0.0135488175, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.89681107, shape=(), dtype=float32)\n","Step 107 : train loss tf.Tensor(0.007147815, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.897795, shape=(), dtype=float32)\n","Step 108 : train loss tf.Tensor(0.019875165, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8987603, shape=(), dtype=float32)\n","Step 109 : train loss tf.Tensor(0.0103829615, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.8997076, shape=(), dtype=float32)\n","Step 110 : train loss tf.Tensor(0.04186872, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.9006373, shape=(), dtype=float32)\n","Step 111 : train loss tf.Tensor(0.018286305, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.90154994, shape=(), dtype=float32)\n","Step 112 : train loss tf.Tensor(0.018773295, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.902446, shape=(), dtype=float32)\n","Step 113 : train loss tf.Tensor(0.013197573, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.9033258, shape=(), dtype=float32)\n","Step 114 : train loss tf.Tensor(0.04214125, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.90391064, shape=(), dtype=float32)\n","Step 115 : train loss tf.Tensor(0.012165627, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.90431154, shape=(), dtype=float32)\n","val loss 0.0309781805 ; val accuracy 0.943414629\n","Step 116 : train loss tf.Tensor(0.042548846, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.90488005, shape=(), dtype=float32)\n","Step 117 : train loss tf.Tensor(0.007450414, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.90571195, shape=(), dtype=float32)\n","Step 118 : train loss tf.Tensor(0.0011495077, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.90652937, shape=(), dtype=float32)\n","Step 119 : train loss tf.Tensor(0.0010706608, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.9073328, shape=(), dtype=float32)\n","Step 120 : train loss tf.Tensor(0.03999934, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.90785617, shape=(), dtype=float32)\n","Step 121 : train loss tf.Tensor(0.01527014, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.9086348, shape=(), dtype=float32)\n","Step 122 : train loss tf.Tensor(0.008442381, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.90940034, shape=(), dtype=float32)\n","Step 123 : train loss tf.Tensor(0.01833167, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.9101532, shape=(), dtype=float32)\n","Step 124 : train loss tf.Tensor(0.011410604, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.9108936, shape=(), dtype=float32)\n","Step 125 : train loss tf.Tensor(0.028856765, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.911622, shape=(), dtype=float32)\n","Step 126 : train loss tf.Tensor(0.0012145615, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.9123385, shape=(), dtype=float32)\n","Step 127 : train loss tf.Tensor(0.014175143, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.9130435, shape=(), dtype=float32)\n","Step 128 : train loss tf.Tensor(0.016063452, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.91373724, shape=(), dtype=float32)\n","Step 129 : train loss tf.Tensor(0.0039662374, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.91442, shape=(), dtype=float32)\n","Step 130 : train loss tf.Tensor(0.027492257, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.91509205, shape=(), dtype=float32)\n","Step 131 : train loss tf.Tensor(0.013436513, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.9157536, shape=(), dtype=float32)\n","Step 132 : train loss tf.Tensor(0.005885317, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.9164049, shape=(), dtype=float32)\n","Step 133 : train loss tf.Tensor(0.0019990709, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.91704625, shape=(), dtype=float32)\n","Step 134 : train loss tf.Tensor(0.00565638, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.9176778, shape=(), dtype=float32)\n","Step 135 : train loss tf.Tensor(0.0073682107, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.91829985, shape=(), dtype=float32)\n","Step 136 : train loss tf.Tensor(0.004102971, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.9189126, shape=(), dtype=float32)\n","Step 137 : train loss tf.Tensor(0.002065517, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.91951615, shape=(), dtype=float32)\n","Step 138 : train loss tf.Tensor(0.012864536, shape=(), dtype=float32) ; train accuracy tf.Tensor(0.919796, shape=(), dtype=float32)\n","val loss 0.0110942367 ; val accuracy 0.951219499\n","20.611680507659912\n"]}],"source":["#Solving the model in eager mode, and printing the time elapsed. \n","st = time.time()\n","train_eager(model, optimizer, 6, train_ds_eager, \n","            train_loss, train_accuracy, valid_ds_eager, \n","            val_loss, val_accuracy)\n","et = time.time()\n","print(et - st)"]},{"cell_type":"code","execution_count":23,"id":"9d84688d","metadata":{"_kg_hide-output":true,"execution":{"iopub.execute_input":"2023-04-04T21:12:01.788209Z","iopub.status.busy":"2023-04-04T21:12:01.787873Z","iopub.status.idle":"2023-04-04T21:12:12.708294Z","shell.execute_reply":"2023-04-04T21:12:12.707362Z"},"papermill":{"duration":10.938798,"end_time":"2023-04-04T21:12:12.711571","exception":false,"start_time":"2023-04-04T21:12:01.772773","status":"completed"},"scrolled":true,"tags":[]},"outputs":[{"name":"stdout","output_type":"stream","text":["Step 1 : train loss 0.0113857221 ; train accuracy 0.920386553\n","Step 2 : train loss 0.00245724199 ; train accuracy 0.920968473\n","Step 3 : train loss 0.00123249646 ; train accuracy 0.921541929\n","Step 4 : train loss 0.000821825757 ; train accuracy 0.92210716\n","Step 5 : train loss 0.0082052676 ; train accuracy 0.922664285\n","Step 6 : train loss 0.00161436095 ; train accuracy 0.923213482\n","Step 7 : train loss 0.000934211712 ; train accuracy 0.92375493\n","Step 8 : train loss 0.000865034759 ; train accuracy 0.924288869\n","Step 9 : train loss 0.000767617952 ; train accuracy 0.924815297\n","Step 10 : train loss 0.000860867789 ; train accuracy 0.925334513\n","Step 11 : train loss 0.000501754519 ; train accuracy 0.925846577\n","Step 12 : train loss 0.0135465879 ; train accuracy 0.926351666\n","Step 13 : train loss 0.003770323 ; train accuracy 0.926849902\n","Step 14 : train loss 0.00104648573 ; train accuracy 0.927341461\n","Step 15 : train loss 0.000132593064 ; train accuracy 0.927826464\n","Step 16 : train loss 0.00304113864 ; train accuracy 0.92830503\n","Step 17 : train loss 0.00137540977 ; train accuracy 0.928777277\n","Step 18 : train loss 0.000462238851 ; train accuracy 0.929243326\n","Step 19 : train loss 0.000844962895 ; train accuracy 0.929703355\n","Step 20 : train loss 0.000505100528 ; train accuracy 0.930157423\n","Step 21 : train loss 9.94947841e-05 ; train accuracy 0.93060571\n","Step 22 : train loss 0.000356109638 ; train accuracy 0.931048214\n","Step 23 : train loss 0.000136836374 ; train accuracy 0.931253731\n","val loss 0.000637792866 ; val accuracy 0.957491279\n","Step 24 : train loss 0.00512903975 ; train accuracy 0.93168807\n","Step 25 : train loss 0.00125513913 ; train accuracy 0.932116926\n","Step 26 : train loss 0.000573234749 ; train accuracy 0.932540476\n","Step 27 : train loss 0.000842907233 ; train accuracy 0.932958722\n","Step 28 : train loss 0.000508245837 ; train accuracy 0.933371842\n","Step 29 : train loss 0.00010523705 ; train accuracy 0.933779895\n","Step 30 : train loss 0.000220375718 ; train accuracy 0.934183\n","Step 31 : train loss 0.0003720893 ; train accuracy 0.93458122\n","Step 32 : train loss 0.00086398155 ; train accuracy 0.934974611\n","Step 33 : train loss 0.00178087933 ; train accuracy 0.935363352\n","Step 34 : train loss 0.00159227906 ; train accuracy 0.935747445\n","Step 35 : train loss 0.000457554619 ; train accuracy 0.936127\n","Step 36 : train loss 7.57289235e-05 ; train accuracy 0.936502099\n","Step 37 : train loss 0.00130333973 ; train accuracy 0.93687284\n","Step 38 : train loss 0.000483620504 ; train accuracy 0.93723923\n","Step 39 : train loss 0.000945189095 ; train accuracy 0.937601447\n","Step 40 : train loss 0.000179715 ; train accuracy 0.937959492\n","Step 41 : train loss 0.000737593102 ; train accuracy 0.938313425\n","Step 42 : train loss 0.000352769857 ; train accuracy 0.938663363\n","Step 43 : train loss 8.5948006e-05 ; train accuracy 0.939009368\n","Step 44 : train loss 0.00132431439 ; train accuracy 0.939351439\n","Step 45 : train loss 0.00120488729 ; train accuracy 0.939689755\n","Step 46 : train loss 0.00015455221 ; train accuracy 0.939847\n","val loss 0.00109769509 ; val accuracy 0.962195098\n","Step 47 : train loss 6.30287104e-05 ; train accuracy 0.940179825\n","Step 48 : train loss 0.00151787803 ; train accuracy 0.940508962\n","Step 49 : train loss 0.000954453542 ; train accuracy 0.940834463\n","Step 50 : train loss 0.000288817537 ; train accuracy 0.941156447\n","Step 51 : train loss 0.00104669528 ; train accuracy 0.941475\n","Step 52 : train loss 0.000495524844 ; train accuracy 0.941790044\n","Step 53 : train loss 0.000372508832 ; train accuracy 0.942101717\n","Step 54 : train loss 0.00116532866 ; train accuracy 0.942410111\n","Step 55 : train loss 0.000684132741 ; train accuracy 0.942715228\n","Step 56 : train loss 0.000159341216 ; train accuracy 0.943017125\n","Step 57 : train loss 0.000342590385 ; train accuracy 0.943315864\n","Step 58 : train loss 0.00023495272 ; train accuracy 0.943611503\n","Step 59 : train loss 0.000761616626 ; train accuracy 0.943904042\n","Step 60 : train loss 5.5139546e-05 ; train accuracy 0.944193542\n","Step 61 : train loss 7.12754845e-05 ; train accuracy 0.944480121\n","Step 62 : train loss 0.000158147217 ; train accuracy 0.94476372\n","Step 63 : train loss 0.000330876879 ; train accuracy 0.945044458\n","Step 64 : train loss 0.00102841749 ; train accuracy 0.945322394\n","Step 65 : train loss 0.00134395983 ; train accuracy 0.94559747\n","Step 66 : train loss 0.000176211965 ; train accuracy 0.945869863\n","Step 67 : train loss 0.000176591522 ; train accuracy 0.946139455\n","Step 68 : train loss 0.000414764188 ; train accuracy 0.946406424\n","Step 69 : train loss 0.000205743141 ; train accuracy 0.9465307\n","val loss 0.000732547196 ; val accuracy 0.965853631\n","Step 70 : train loss 0.00025877048 ; train accuracy 0.946793795\n","Step 71 : train loss 0.000607891416 ; train accuracy 0.947054327\n","Step 72 : train loss 0.000705531449 ; train accuracy 0.947312295\n","Step 73 : train loss 4.31984445e-05 ; train accuracy 0.947567821\n","Step 74 : train loss 0.000537542452 ; train accuracy 0.947820842\n","Step 75 : train loss 0.00026413289 ; train accuracy 0.94807142\n","Step 76 : train loss 5.39932553e-05 ; train accuracy 0.948319614\n","Step 77 : train loss 0.000783159339 ; train accuracy 0.948565483\n","Step 78 : train loss 0.00017090961 ; train accuracy 0.948808968\n","Step 79 : train loss 0.000107409811 ; train accuracy 0.949050188\n","Step 80 : train loss 0.000214120344 ; train accuracy 0.949289143\n","Step 81 : train loss 0.00014793915 ; train accuracy 0.949525893\n","Step 82 : train loss 6.31787407e-05 ; train accuracy 0.949760437\n","Step 83 : train loss 0.000286711671 ; train accuracy 0.949992776\n","Step 84 : train loss 0.00047388603 ; train accuracy 0.950222969\n","Step 85 : train loss 0.000401714409 ; train accuracy 0.950451076\n","Step 86 : train loss 0.000412946101 ; train accuracy 0.950677097\n","Step 87 : train loss 5.63694557e-05 ; train accuracy 0.950901091\n","Step 88 : train loss 0.000272833102 ; train accuracy 0.951123059\n","Step 89 : train loss 0.000355581549 ; train accuracy 0.951343\n","Step 90 : train loss 0.000152462395 ; train accuracy 0.951561\n","Step 91 : train loss 0.000106109335 ; train accuracy 0.951777\n","Step 92 : train loss 0.00264014653 ; train accuracy 0.951877594\n","val loss 0.000589644769 ; val accuracy 0.968780518\n","Step 93 : train loss 7.50482723e-05 ; train accuracy 0.952090859\n","Step 94 : train loss 0.000188287551 ; train accuracy 0.952302158\n","Step 95 : train loss 0.000545131625 ; train accuracy 0.952511668\n","Step 96 : train loss 0.000160120064 ; train accuracy 0.952719331\n","Step 97 : train loss 8.99732549e-05 ; train accuracy 0.952925146\n","Step 98 : train loss 0.000232733131 ; train accuracy 0.953129232\n","Step 99 : train loss 0.000143363635 ; train accuracy 0.95333153\n","Step 100 : train loss 7.47525046e-05 ; train accuracy 0.9535321\n","Step 101 : train loss 0.000379662612 ; train accuracy 0.953730941\n","Step 102 : train loss 3.4478373e-05 ; train accuracy 0.953928113\n","Step 103 : train loss 0.000460924872 ; train accuracy 0.954123557\n","Step 104 : train loss 0.000665000523 ; train accuracy 0.954317391\n","Step 105 : train loss 8.67476265e-05 ; train accuracy 0.954509616\n","Step 106 : train loss 9.34798954e-05 ; train accuracy 0.954700172\n","Step 107 : train loss 0.000477583584 ; train accuracy 0.954889178\n","Step 108 : train loss 0.000163505931 ; train accuracy 0.955076575\n","Step 109 : train loss 0.000121157413 ; train accuracy 0.955262482\n","Step 110 : train loss 6.30576687e-05 ; train accuracy 0.955446839\n","Step 111 : train loss 0.000185035984 ; train accuracy 0.955629647\n","Step 112 : train loss 0.000324040942 ; train accuracy 0.955810964\n","Step 113 : train loss 0.000325833942 ; train accuracy 0.955990851\n","Step 114 : train loss 9.81943231e-06 ; train accuracy 0.956169248\n","Step 115 : train loss 0.000645672611 ; train accuracy 0.956252396\n","val loss 0.000346768793 ; val accuracy 0.971175194\n","Step 116 : train loss 2.05100805e-05 ; train accuracy 0.956428647\n","Step 117 : train loss 0.000102631617 ; train accuracy 0.956603527\n","Step 118 : train loss 7.14799244e-05 ; train accuracy 0.956777036\n","Step 119 : train loss 0.000274408056 ; train accuracy 0.956949115\n","Step 120 : train loss 0.000345972134 ; train accuracy 0.957119823\n","Step 121 : train loss 0.000188704726 ; train accuracy 0.957289219\n","Step 122 : train loss 0.000532618491 ; train accuracy 0.957457244\n","Step 123 : train loss 0.000281851302 ; train accuracy 0.957624\n","Step 124 : train loss 4.072e-05 ; train accuracy 0.957789421\n","Step 125 : train loss 0.000208775891 ; train accuracy 0.957953572\n","Step 126 : train loss 4.88994119e-05 ; train accuracy 0.958116472\n","Step 127 : train loss 8.91265227e-05 ; train accuracy 0.95827806\n","Step 128 : train loss 1.35742921e-05 ; train accuracy 0.958438456\n","Step 129 : train loss 0.00020502643 ; train accuracy 0.9585976\n","Step 130 : train loss 0.000183166208 ; train accuracy 0.958755493\n","Step 131 : train loss 3.231554e-05 ; train accuracy 0.958912253\n","Step 132 : train loss 3.53331343e-05 ; train accuracy 0.959067762\n","Step 133 : train loss 9.75470539e-05 ; train accuracy 0.959222138\n","Step 134 : train loss 0.000594421814 ; train accuracy 0.959375381\n","Step 135 : train loss 5.18982124e-05 ; train accuracy 0.959527433\n","Step 136 : train loss 0.000146486418 ; train accuracy 0.959678352\n","Step 137 : train loss 9.89961409e-05 ; train accuracy 0.959828138\n","Step 138 : train loss 3.19964965e-05 ; train accuracy 0.959898\n","val loss 0.000220190108 ; val accuracy 0.973170757\n","10.911803722381592\n"]}],"source":["#Solving the model in graph mode, and printing the time elapsed. \n","st = time.time()\n","train_graph(model, optimizer, 6, train_ds_graph, \n","            train_loss, train_accuracy, valid_ds_graph, \n","            val_loss, val_accuracy)\n","et = time.time()\n","print(et - st)"]},{"cell_type":"markdown","id":"1012c9c6","metadata":{"papermill":{"duration":0.030365,"end_time":"2023-04-04T21:12:12.775219","exception":false,"start_time":"2023-04-04T21:12:12.744854","status":"completed"},"tags":[]},"source":["The execution time is quite variable between different executions, but at the time of this writing, **the model in eager mode took 21 seconds, while in graph mode it took only 10. That is, it took only half of the time**. \n","\n","It should be noted **that not only are there differences in performance, but an improvement in the accuracy and loss of the model is also achieved**. Not huge, but we can't ignore it.\n","\n","With eager mode, we obtain an accuracy of 0.974 while with Graph mode, the accuracy obtained is 0.987. **It would be necessary to study if this difference is maintained between executions, in different machines, and with different models.**\n","\n","These numbers have been obtained by training for only two epochs and using Kaggle's P100 GPUs, with the horse vs human dataset.\n","\n","If you download the notebook you can try it with any of the two TensorFlow hub models that have been prepared, I recommend resnet_50, and with the cats vs dogs dataset, which is much larger.\n","\n","You will see that the times are very different, it is even possible that just by lowering the notebook, without changing anything, both the final time and the percentage of performance improvement change, which can be even better.\n","\n","On my laptop, **the performance improvement is about 70%**. This may be due to a parallelization improvement or the type of GPU, no idea. \n"]},{"cell_type":"markdown","id":"099a20a5","metadata":{"papermill":{"duration":0.030157,"end_time":"2023-04-04T21:12:12.837023","exception":false,"start_time":"2023-04-04T21:12:12.806866","status":"completed"},"tags":[]},"source":["# Check the Model "]},{"cell_type":"code","execution_count":24,"id":"6c2cde6c","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:12:12.904249Z","iopub.status.busy":"2023-04-04T21:12:12.903756Z","iopub.status.idle":"2023-04-04T21:12:12.914434Z","shell.execute_reply":"2023-04-04T21:12:12.913131Z"},"papermill":{"duration":0.04857,"end_time":"2023-04-04T21:12:12.917131","exception":false,"start_time":"2023-04-04T21:12:12.868561","status":"completed"},"tags":[]},"outputs":[],"source":["import matplotlib.pyplot as plt\n","import matplotlib.image as mpimg\n","from matplotlib import rcParams"]},{"cell_type":"code","execution_count":25,"id":"39ce2dd9","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:12:12.985874Z","iopub.status.busy":"2023-04-04T21:12:12.984913Z","iopub.status.idle":"2023-04-04T21:12:13.007914Z","shell.execute_reply":"2023-04-04T21:12:13.007031Z"},"papermill":{"duration":0.058663,"end_time":"2023-04-04T21:12:13.010573","exception":false,"start_time":"2023-04-04T21:12:12.95191","status":"completed"},"tags":[]},"outputs":[],"source":["class_names = ['cats', 'dogs']\n","class_names = ['horse', 'human']\n","def plot_image(i, predictions_array, true_label, img):\n","    predictions_array, true_label, img = predictions_array[i], true_label[i], img[i]\n","    plt.grid(False)\n","    plt.xticks([])\n","    plt.yticks([])\n","\n","    img = np.squeeze(img)\n","\n","    plt.imshow(img, cmap=plt.cm.binary)\n","\n","    predicted_label = np.argmax(predictions_array)\n","    \n","    # green-colored annotations will mark correct predictions. red otherwise.\n","    if predicted_label == true_label:\n","        color = 'green'\n","    else:\n","        color = 'red'\n","    \n","    # print the true label first\n","    print(true_label)\n","  \n","    # show the image and overlay the prediction\n","    plt.xlabel(\"{} {:2.0f}% ({})\".format(class_names[predicted_label],\n","                                100*np.max(predictions_array),\n","                                class_names[true_label]),\n","                                color=color)"]},{"cell_type":"code","execution_count":26,"id":"378095fc","metadata":{"_kg_hide-output":true,"execution":{"iopub.execute_input":"2023-04-04T21:12:13.076048Z","iopub.status.busy":"2023-04-04T21:12:13.075635Z","iopub.status.idle":"2023-04-04T21:12:14.388291Z","shell.execute_reply":"2023-04-04T21:12:14.387133Z"},"papermill":{"duration":1.348896,"end_time":"2023-04-04T21:12:14.391465","exception":false,"start_time":"2023-04-04T21:12:13.042569","status":"completed"},"scrolled":true,"tags":[]},"outputs":[{"name":"stderr","output_type":"stream","text":["Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n","Cleanup called...\n"]}],"source":["test_imgs = []\n","test_labels = []\n","\n","predictions = []\n","for images, labels in test_ds_eager:\n","    preds = model(images)\n","    preds = preds.numpy()\n","    predictions.extend(preds)\n","\n","    test_imgs.extend(images.numpy())\n","    test_labels.extend(labels.numpy())"]},{"cell_type":"code","execution_count":27,"id":"49b05432","metadata":{"execution":{"iopub.execute_input":"2023-04-04T21:12:14.441828Z","iopub.status.busy":"2023-04-04T21:12:14.440838Z","iopub.status.idle":"2023-04-04T21:12:14.585343Z","shell.execute_reply":"2023-04-04T21:12:14.584218Z"},"papermill":{"duration":0.173035,"end_time":"2023-04-04T21:12:14.589433","exception":false,"start_time":"2023-04-04T21:12:14.416398","status":"completed"},"tags":[]},"outputs":[{"name":"stdout","output_type":"stream","text":["0\n"]},{"data":{"image/png":"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\n","text/plain":["<Figure size 432x216 with 1 Axes>"]},"metadata":{},"output_type":"display_data"}],"source":["# Visualize the outputs \n","plt.figure(figsize=(6,3))\n","plt.subplot(1,2,1)\n","plot_image(100, predictions, test_labels, test_imgs)\n","plt.show()"]},{"cell_type":"markdown","id":"f2f69c1a","metadata":{"execution":{"iopub.execute_input":"2022-10-28T13:55:16.232353Z","iopub.status.busy":"2022-10-28T13:55:16.231982Z","iopub.status.idle":"2022-10-28T13:55:16.236735Z","shell.execute_reply":"2022-10-28T13:55:16.235515Z","shell.execute_reply.started":"2022-10-28T13:55:16.232322Z"},"papermill":{"duration":0.022384,"end_time":"2023-04-04T21:12:14.669544","exception":false,"start_time":"2023-04-04T21:12:14.64716","status":"completed"},"tags":[]},"source":["# Conclusions & Fork & Play\n","We have seen how with little effort we can achieve a significant improvement in our notebooks.\n","\n","It's as simple as decorating the function we want to speed up with tf.function, and complying with some development rules.\n","\n","It is true that many examples are missing from the notebook and the code will need more modifications. But the improvements achieved are worth it.\n","\n","So dedicating a little more effort to development to obtain such a significant gain in training and execution time is something that surely pays off.\n","\n","## Experiments to do\n","If you have a powerful machine, try changing the Dataset or the Model, so you can see if the gain percentage increases or decreases as the process becomes more complicated.\n","\n","A good place to try this technique is on NLP models. Treating text to make it ready to use in an NLP model is usually a huge resource hog, surely Autograph can be used to improve this performance.\n","\n","I plan to try it later, but if you do before I do, I'd love for you to tell me how it worked. I wait for you in the comments!"]},{"cell_type":"markdown","id":"0f7b5dc2","metadata":{"papermill":{"duration":0.021789,"end_time":"2023-04-04T21:12:14.713923","exception":false,"start_time":"2023-04-04T21:12:14.692134","status":"completed"},"tags":[]},"source":["# More Notebooks in the TensorFlow Serie.\n","I'm working in a serie of notebooks with some interesting techniques in Tensorflow: *Tensorflow beyond the basics*.\n","\n","How to create a Siamese Network to compare images. https://www.kaggle.com/code/peremartramanonellas/how-to-create-a-siamese-network-to-compare-images\n","\n","Multiple outputs with Keras Functional API. \n","https://www.kaggle.com/code/peremartramanonellas/guide-multiple-outputs-with-keras-functional-api/edit/run/109206893\n","\n","In my medium profile you can find the articles where each of the notebooks is explained.\n","https://medium.com/@peremartra\n","\n","#### **Please, if you liked the notebook, consider upvoting it. It really encourages me a lot to write more notebooks like that.**"]},{"cell_type":"code","execution_count":null,"id":"a21b6c2c","metadata":{"papermill":{"duration":0.021184,"end_time":"2023-04-04T21:12:14.757006","exception":false,"start_time":"2023-04-04T21:12:14.735822","status":"completed"},"tags":[]},"outputs":[],"source":[]}],"metadata":{"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"codemirror_mode":{"name":"ipython","version":3},"file_extension":".py","mimetype":"text/x-python","name":"python","nbconvert_exporter":"python","pygments_lexer":"ipython3","version":"3.7.12"},"papermill":{"default_parameters":{},"duration":59.347801,"end_time":"2023-04-04T21:12:18.054435","environment_variables":{},"exception":null,"input_path":"__notebook__.ipynb","output_path":"__notebook__.ipynb","parameters":{},"start_time":"2023-04-04T21:11:18.706634","version":"2.3.4"},"widgets":{"application/vnd.jupyter.widget-state+json":{"state":{"0088dc0365024443939b9a4374cdbe41":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"0206fd6df80f4e1cbcfbc239c6092717":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"DescriptionStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"026ba2fd747f456aa1c755f4c470eb19":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"02f3758e1ebb460eaa0eb1c20185655d":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"0346cc08825c455b83c383747f27df02":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HBoxModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_56d8d1fe3d6b4214abeada285ac8ac33","IPY_MODEL_fb3e1b84f51640d784f0e2267b7975ad","IPY_MODEL_5accea8bbcb5449f94b095d731aa0d85"],"layout":"IPY_MODEL_ca109c6932ef436e891fcb8dd923f913"}},"0517829364a6446e839c277e847c1e0b":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"DescriptionStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"0696ada07e5541f08e39253659eb71e4":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HBoxModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_51d85951288c4e5f9f60d245c1c4749f","IPY_MODEL_e432fa5ea88543d3bfc786f9354d9d28","IPY_MODEL_bbb14d7eeeeb477785be4c9c18b3f94f"],"layout":"IPY_MODEL_74741409affa44019994b8ae8d2009c7"}},"0dcbe94fc2e3413a8754680001f8219b":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"FloatProgressModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"","description":"","description_tooltip":null,"layout":"IPY_MODEL_2de178a12f484487b05d20c58d10dfd8","max":2.0,"min":0.0,"orientation":"horizontal","style":"IPY_MODEL_acccf8aea49942319a2111de32729524","value":2.0}},"0df06983918844df86e3b37bc30c1824":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"DescriptionStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"1252b21abc114f40a831a29a3c4d1623":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"FloatProgressModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"success","description":"","description_tooltip":null,"layout":"IPY_MODEL_676b516479084e3486fcc7ee0cfa762c","max":1.0,"min":0.0,"orientation":"horizontal","style":"IPY_MODEL_4e778d7cb0fd4b02ae317e389b1d0d6b","value":1.0}},"1453f8f1286d467e969866bc07c668eb":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"ProgressStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"17f17fb1d1fd492a9ead5bfc4b93f09d":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"DescriptionStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"1e8defb4e6ab498d9bf52b21a84b72a2":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_7e163da1f3134f71ba6aa34efa12a4b8","placeholder":"​","style":"IPY_MODEL_9a898c7485404656ac18939369b63243","value":" 1017/1027 [00:00&lt;00:00, 1636.56 examples/s]"}},"21e3518820e346dea5aac5b913927541":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_0088dc0365024443939b9a4374cdbe41","placeholder":"​","style":"IPY_MODEL_17f17fb1d1fd492a9ead5bfc4b93f09d","value":"Generating test examples...:  56%"}},"22b6e721426d42adb21fe1a6ef36f6ae":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"DescriptionStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"261476e6a9474610b4f924c00cb44bd0":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":"20px"}},"2a87c9dc4fed481d86a1bb208c52b5f2":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HBoxModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_c0d6ee57998c436c996afed65c66e7ee","IPY_MODEL_1252b21abc114f40a831a29a3c4d1623","IPY_MODEL_8e08a3efaddd437694edfa77490e2d66"],"layout":"IPY_MODEL_552d3d8c1c5944e19db8732bfbed4ba5"}},"2b5a289987c842d3b66cba36acb36030":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"2d67ed67527344378fac6e685753d360":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"DescriptionStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"2de178a12f484487b05d20c58d10dfd8":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"309b990e1e394d788f086e215718f121":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"47bf20d0f91842fdb3bc5e28ec19fc80":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"48ca196eb75d4c3282bdedeb2406008d":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"DescriptionStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"4ce1c0f861cb48b89a008caa2798b3a5":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"DescriptionStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"4e778d7cb0fd4b02ae317e389b1d0d6b":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"ProgressStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"4f255448ad35413c812c3b4f097ddd5c":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"DescriptionStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"4f6b781132af4370aecd0091a3f31a89":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"5150908d89124a47a86074ad22a47183":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"ProgressStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"51d85951288c4e5f9f60d245c1c4749f":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_5bae49f04b42479ebc9de13cca1bb71d","placeholder":"​","style":"IPY_MODEL_ddeddb603f8a480b910d4e82e293826e","value":"Dl Size...: 100%"}},"552d3d8c1c5944e19db8732bfbed4ba5":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"56871d39c41f4d8d8a2e077834e53565":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"ProgressStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"56d8d1fe3d6b4214abeada285ac8ac33":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_8b95055bc4db421b9520c20a090b5c23","placeholder":"​","style":"IPY_MODEL_4f255448ad35413c812c3b4f097ddd5c","value":"Shuffling horses_or_humans-train.tfrecord...:  54%"}},"5accea8bbcb5449f94b095d731aa0d85":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_aee58b352cb941b59017d54827fbef0d","placeholder":"​","style":"IPY_MODEL_4ce1c0f861cb48b89a008caa2798b3a5","value":" 550/1027 [00:00&lt;00:00, 5498.73 examples/s]"}},"5bae49f04b42479ebc9de13cca1bb71d":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"615ccdf39f504c66a377dcc4b32f631c":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"64d19d80c2ef4925a90ce4e72ecf604a":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HBoxModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_d44544138bb94128b421deb3b6d72291","IPY_MODEL_9825fabe7c4d462a9652639d37d95ea0","IPY_MODEL_1e8defb4e6ab498d9bf52b21a84b72a2"],"layout":"IPY_MODEL_ee7a478af2784e11afed3f5e9c8e82aa"}},"6641bff7acb0434690cff4ebd85d421b":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"676b516479084e3486fcc7ee0cfa762c":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":"20px"}},"6808e9ac18ab47d2ac722177892748fb":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"ProgressStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"6d2f518348c746e78e8af10ff476e380":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"FloatProgressModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"","description":"","description_tooltip":null,"layout":"IPY_MODEL_bbd3da42574042e7bd71718e089de852","max":256.0,"min":0.0,"orientation":"horizontal","style":"IPY_MODEL_7d906ae3673e41adaf0481bbf8313a11","value":256.0}},"74741409affa44019994b8ae8d2009c7":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"75e8207bf3fb4308be3880b59178afaa":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"7a348ff7539d43148df94fc14f81353e":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_615ccdf39f504c66a377dcc4b32f631c","placeholder":"​","style":"IPY_MODEL_0df06983918844df86e3b37bc30c1824","value":"Generating splits...: 100%"}},"7b7d7ce31c3948c19ead18334c876fc5":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"7d906ae3673e41adaf0481bbf8313a11":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"ProgressStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"7e163da1f3134f71ba6aa34efa12a4b8":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"8b95055bc4db421b9520c20a090b5c23":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"8e08a3efaddd437694edfa77490e2d66":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_026ba2fd747f456aa1c755f4c470eb19","placeholder":"​","style":"IPY_MODEL_22b6e721426d42adb21fe1a6ef36f6ae","value":" 2/2 [00:01&lt;00:00,  1.07s/ url]"}},"9825fabe7c4d462a9652639d37d95ea0":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"FloatProgressModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"","description":"","description_tooltip":null,"layout":"IPY_MODEL_4f6b781132af4370aecd0091a3f31a89","max":1027.0,"min":0.0,"orientation":"horizontal","style":"IPY_MODEL_5150908d89124a47a86074ad22a47183","value":1027.0}},"98581a536e964517a7f29e8dd62284d7":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"9a898c7485404656ac18939369b63243":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"DescriptionStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"acccf8aea49942319a2111de32729524":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"ProgressStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"ad5e38f1f7d14480ad5c23be151746c2":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"ada125bc5c6e43a8b15ed3f219fbc263":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"FloatProgressModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"","description":"","description_tooltip":null,"layout":"IPY_MODEL_6641bff7acb0434690cff4ebd85d421b","max":256.0,"min":0.0,"orientation":"horizontal","style":"IPY_MODEL_1453f8f1286d467e969866bc07c668eb","value":256.0}},"aee58b352cb941b59017d54827fbef0d":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"b06398c563c3484ba20b5ffa239356a9":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"b2b36d9468844e049379edbfa726ba92":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_2b5a289987c842d3b66cba36acb36030","placeholder":"​","style":"IPY_MODEL_0517829364a6446e839c277e847c1e0b","value":" 2/2 [00:01&lt;00:00,  2.03 splits/s]"}},"b4db46091dc6438ab8dfb43ac2eae42f":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HBoxModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_bfa6c3ad84934231a3c4bb5743d89da9","IPY_MODEL_ada125bc5c6e43a8b15ed3f219fbc263","IPY_MODEL_bcb753162b9a4f8f9532bc2732dd13a3"],"layout":"IPY_MODEL_02f3758e1ebb460eaa0eb1c20185655d"}},"bbb14d7eeeeb477785be4c9c18b3f94f":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_309b990e1e394d788f086e215718f121","placeholder":"​","style":"IPY_MODEL_e7447328fdfb4a7c8c4105db899ebf27","value":" 152/152 [00:01&lt;00:00, 91.49 MiB/s]"}},"bbd3da42574042e7bd71718e089de852":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"bcb753162b9a4f8f9532bc2732dd13a3":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_e010375ece8a4bf0accd59e248015349","placeholder":"​","style":"IPY_MODEL_2d67ed67527344378fac6e685753d360","value":" 0/256 [00:00&lt;?, ? examples/s]"}},"bf2c9539d5df4b47a7024dd3374f76e9":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"bfa6c3ad84934231a3c4bb5743d89da9":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_98581a536e964517a7f29e8dd62284d7","placeholder":"​","style":"IPY_MODEL_0206fd6df80f4e1cbcfbc239c6092717","value":"Shuffling horses_or_humans-test.tfrecord...:   0%"}},"c0d6ee57998c436c996afed65c66e7ee":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_75e8207bf3fb4308be3880b59178afaa","placeholder":"​","style":"IPY_MODEL_c607b4ff89554d1ab28dc49d0a7d4449","value":"Dl Completed...: 100%"}},"c607b4ff89554d1ab28dc49d0a7d4449":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"DescriptionStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"c8d80826eef54a7ca1315542c3936eac":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_7b7d7ce31c3948c19ead18334c876fc5","placeholder":"​","style":"IPY_MODEL_48ca196eb75d4c3282bdedeb2406008d","value":" 144/256 [00:00&lt;00:00, 1439.53 examples/s]"}},"ca109c6932ef436e891fcb8dd923f913":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"d3e5a3b930e145f0bbbda44bae0972e0":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HBoxModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_7a348ff7539d43148df94fc14f81353e","IPY_MODEL_0dcbe94fc2e3413a8754680001f8219b","IPY_MODEL_b2b36d9468844e049379edbfa726ba92"],"layout":"IPY_MODEL_bf2c9539d5df4b47a7024dd3374f76e9"}},"d44544138bb94128b421deb3b6d72291":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_ad5e38f1f7d14480ad5c23be151746c2","placeholder":"​","style":"IPY_MODEL_e15033d48e27445fa947d695b3327933","value":"Generating train examples...:  99%"}},"ddeddb603f8a480b910d4e82e293826e":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"DescriptionStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"e010375ece8a4bf0accd59e248015349":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"e15033d48e27445fa947d695b3327933":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"DescriptionStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"e432fa5ea88543d3bfc786f9354d9d28":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"FloatProgressModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"success","description":"","description_tooltip":null,"layout":"IPY_MODEL_261476e6a9474610b4f924c00cb44bd0","max":1.0,"min":0.0,"orientation":"horizontal","style":"IPY_MODEL_6808e9ac18ab47d2ac722177892748fb","value":1.0}},"e630f4e736344ab488cd4d9685d35cff":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"HBoxModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_21e3518820e346dea5aac5b913927541","IPY_MODEL_6d2f518348c746e78e8af10ff476e380","IPY_MODEL_c8d80826eef54a7ca1315542c3936eac"],"layout":"IPY_MODEL_b06398c563c3484ba20b5ffa239356a9"}},"e7447328fdfb4a7c8c4105db899ebf27":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"DescriptionStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"ee7a478af2784e11afed3f5e9c8e82aa":{"model_module":"@jupyter-widgets/base","model_module_version":"1.2.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"fb3e1b84f51640d784f0e2267b7975ad":{"model_module":"@jupyter-widgets/controls","model_module_version":"1.5.0","model_name":"FloatProgressModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"","description":"","description_tooltip":null,"layout":"IPY_MODEL_47bf20d0f91842fdb3bc5e28ec19fc80","max":1027.0,"min":0.0,"orientation":"horizontal","style":"IPY_MODEL_56871d39c41f4d8d8a2e077834e53565","value":1027.0}}},"version_major":2,"version_minor":0}}},"nbformat":4,"nbformat_minor":5}