{"nbformat": 4, "metadata": {"kernelspec": {"language": "python", "display_name": "Python 3", "name": "python3"}, "language_info": {"version": "3.6.3", "pygments_lexer": "ipython3", "nbconvert_exporter": "python", "file_extension": ".py", "name": "python", "mimetype": "text/x-python", "codemirror_mode": {"version": 3, "name": "ipython"}}}, "nbformat_minor": 1, "cells": [{"cell_type": "markdown", "source": ["## Recipe for training on Google ML Engine\n", "This kernel describes how I got Alex Ozerin's kernel (https://www.kaggle.com/alexozerin/end-to-end-baseline-tf-estimator-lb-0-72 ) running on Google ML Engine.\n", "\n", "Use this recipe at your own risk !"], "metadata": {"_uuid": "1b5eb9139c03d7af400367de9135daf100a2cddf", "_cell_guid": "12c7799d-6216-4ee3-a1f5-21d7f0c6aae7"}}, {"cell_type": "markdown", "source": ["### create a project on Google Cloud\n", "see https://cloud.google.com/resource-manager/docs/creating-managing-projects\n", "* go to https://console.cloud.google.com/\n", "* from the menu on the right select 'IAM & admin' -> 'Manage resources'\n", "* on the newly opened page, click on 'Create Project'\n", "* choose a project id, in my case I selected kaggle-speech-(number)\n", "* for some reason the new project does not appear immediately in the list of projects on the 'Manage resources' page. Check the notification icon at the top right for completion of  the project creation process.\n", "* make sure the new project is the currently selected one (there is a dropdown menu right of 'Google Cloud Platform' at the top right of the Google cloud console at the top right\n", "* make sure that billing is set up correctly\n", "\n", "* install the Google Cloud SDK on your local machine as described here: https://cloud.google.com/sdk/downloads\n", "* initialize the cloud SDK on your local machine (see https://cloud.google.com/sdk/docs/initializing ) by running\n", "\n", "```gcloud init```\n", "\n", "We'll use the gcloud command to submit jobs to the ML engine at the end.\n", "\n", "\n", "   \n"], "metadata": {"_uuid": "94214c1109d0831b45aecd1a00884d826e70007d", "_cell_guid": "78e00324-ba7b-4542-96ae-0ec1ff9480c0"}}, {"attachments": {"image.png": {"image/png": "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"}}, "cell_type": "markdown", "source": ["### create a VM for packing train and test data\n", "It turns out that running training on the ~ 65k files is rather slow compared to packing the training data into one single npy file. In this section I describe how I created such a packed file in a virtual machine running on the Google compute engine thereby avoiding uploading several gigabytes of data through my 1 MBit/s ADSL connection.\n", "\n", "* see also the documentation here: https://cloud.google.com/compute/docs/instances/create-start-instance\n", "* go to https://cloud.google.com/compute/docs/images#os-compute-support\n", "* go to the table row with 'Ubuntu', click on the Start button on the right in this row.\n", "* on the following page, select 'Ubuntu Xenial 16.04' (make sure that this is a virtual machine and not a container)\n", "* click on 'Launch on Compute Engine'\n", "* on the following page ('VM instances') I selected a virtual machine which can deal with the size of the dataset on disk and in memory:\n", "   * select n1-highmem-2 from the 'machine type' drop down menu. This will give you a machine with two virtual CPUs and 13 GByte memory\n", "   * increase the boot disk size to 25 GByte (we need to store the zipped, unzipped and numpy array files on disk for\n", "the train and test dataset -- one could probably do with 20 GByte or even less but 25 GByte works well)\n", "   * The corresponding section on this page should look as follows:\n", "![image.png](attachment:image.png)\n", "   * ensure that the Zone corresponds to where you want to run on GPUs afterwards (zone starts with us-east1 in my case), You can see which zones have GPUs available here: https://cloud.google.com/compute/docs/gpus/ . You can also see the pricing in different zones here: https://cloud.google.com/compute/pricing#gpus\n", "   * under the 'Access scopes' section, select 'Set access for each API'. A list of dropdown menus appears.\n", "      * In order to grant this VM permission to create buckets as part of your project (see https://stackoverflow.com/a/27298944/288875 and https://cloud.google.com/compute/docs/access/create-enable-service-accounts-for-instances):\n", "      * Set the menu item under 'Storage' to 'Read/Write'. \n", "       \n", "Once all settings are as desired:\n", "       \n", "* Click the 'Create' button at the bottom of the page.\n", "    \n", "You then should be directed to a page with an overview of the state of your VM instances (https://console.cloud.google.com/compute/instances ) . The newly created instance should have been started automatically.\n", "* login to this vm: click on 'ssh' on the line corresponding to the newly created VM\n", "   * on first attempt I got a message in Chrome that it blocked a popup but on the second attempt\n", "it opened a window in the browser with a console. "], "metadata": {"_uuid": "940bf4d376e431a492521798449a15d8f9f3e9de", "_cell_guid": "56baa41e-4cb3-4036-8d0b-6ad2f3a34d99"}}, {"cell_type": "markdown", "source": ["### Download train and test datasets to the new VM and pack them \n", "\n", "In this step we'll download the train and test datasets from Kaggle and pack them into one big 2D matrix for each of the train and test datasets.\n", "\n", "* In the shell where you are logged into the VM create a directory (other than /tmp/) to store the Kaggle datasets. In the VM execute the following commands:\n", "   * sudo mkdir /data\n", "   * sudo chown $(whoami) /data\n", "   * cd /data\n", "(make sure NOT to paste the asterisks/bullets at the beginning of the lines into the shell !)\n", "\n", "#### Download the Kaggle dataset from the VM\n", "   * The datasets can't simply be downloaded from the command line as they need to be downloaded from within an authenticated session with kaggle.com\n", "   * To get the necessary URL parameters I installed the Chrome extension CurlWget, see https://chrome.google.com/webstore/detail/curlwget/jmocjfidanebdlinpbcdkcmgdifblncg?hl=en\n", "   * Make sure you are logged in on kaggle.com in the browser on your local machine.\n", "   * Navigate to the dataset page on Kaggle: https://www.kaggle.com/c/tensorflow-speech-recognition-challenge/data. \n", "   * click e.g. on 'train.7z' then click 'Download'. In my case a file/directory chooser dialog opens. Press cancel. If the download starts by itself, cancel it.\n", "    * At the top right I see the 'CurlWget' icon (round yellow icon with a greater than sign). Click on it: you should see a wget command. Make\n", "sure it has the correct type of file you want to download in the command (e.g. 'train.7z'). Copy the entire wget command and paste it into the VM command line.\n", "\n", "In my case the command starts with\n", "\n", "```wget --header=\"```\n", "\n", "and ends with  \n", "\n", "```-O \"train.7z\" -c```\n", "\n", "for the train dataset \n", "\n", "* once the train.7z and test.7z files are downloaded install the 7z utility on the VM and python packages required later.\n", "   * in the VM execute the following commands:\n", "\n", "`sudo apt-get install -y p7zip-full python3-numpy python3-scipy python3-pip python-psutil`\n", "\n", "`pip3 install --user tqdm`\n", "\n", "* then unpack the train and test datasets in the `/data` directory with the following commands:\n", "```7z x train.7z >/dev/null & 7z x test.7z > /dev/null```\n", "\n", "(wait until both processes complete, this can take more than one hour)\n", "\n", "#### Initialize the gcloud environment and create a storage bucket\n", "\n", "* create a storage bucket to download the Kaggle dataset to (see also https://cloud.google.com/compute/docs/disks/gcs-buckets )\n", "    * initialize the glcoud environment on the VM:\n", "\n", "```gcloud init```\n", "\n", "when asked about the account to use, select the account of the form `<number>-compute@developer.gserviceaccount.com`\n", "\n", "When asked about the project id you would like to use, enter the project id (in my case of the form `kaggle-speech-(number)`).\n", "\n", "You can ignore warnings about missing permissions to list all available projects.\n", "\n", "   * set the environment variable PROJECT_ID in the shell on the VM:\n", "\n", "```PROJECT_ID=$(gcloud config list project --format \"value(core.project)\")```\n", "  \n", "   * select a name for the storage bucket, e.g.\n", "\n", "`BUCKET_NAME=${PROJECT_ID}-data`\n", "\n", "* set an environment variable for the Google cloud platform region you want to use, in my case this is:\n", "\n", "`REGION=us-east1`\n", "\n", "   * create the storage bucket:\n", "\n", "`gsutil mb -l $REGION gs://$BUCKET_NAME`\n", "\n", "#### Pack the train and test data\n", "\n", "* get `pack-for-gcp.py` from gist.github.com:\n", "    * In the VM shell execute\n", "    \n", "`wget 'https://gist.githubusercontent.com/andreh7/b444c932b0e6dfda51c2efd19732581c/raw/fb1220c45136a12c9f7530dca27a1cb69863ddf6/pack-for-gcp.py'`\n", "* run pack-for-gcp.py in the directory where you unpacked the train and test datastets (/data), e.g. as follows:\n", "\n", "`cd /data`\n", "`python3 pack-for-gcp.py $PWD $PWD`\n", "\n", "this will search for the input files in the current working directory and will write out .npy (wave data) and .csv files (file names etc.)  to the current working directory.\n", "\n", "#### Copy the packed wave data and meta data to the storage bucket\n", "\n", "* execute the following commands in the shell on the VM:\n", "`cd /data`\n", "\n", "`gsutil -m cp *.npy *.csv gs://$BUCKET_NAME`\n", "\n", "* verify that the files were properly copied to the storage bucket:\n", "\n", "`gsutil ls -l gs://$BUCKET_NAME`\n", "\n", "you should see three .npy files and three .csv files corresponding to the train, test and noise wave files and associated metadata. The .npy files are several gigabytes in size.\n", "\n", "#### shutting down the VM\n", "\n", "* Make sure to shut down the virtual machine when you don't need it anymore to avoid unnecessary costs. You may even want to delete it after having run the above commands and don't need it anymore (e.g. if you plan to train your models by submitting jobs from your local machine to ML Engine).\n", "\n", "You can shutdown the machine from the web interface https://console.cloud.google.com/compute/instances \n", "or by doing \n", "\n", "`sudo /sbin/poweroff`\n", "\n", "from a shell in the VM (note that when shutting the VM down from the command line it takes some time until the VM status is reflected in the web console application)\n", "\n", "\n", "\n", "\n", "   \n", "\n"], "metadata": {"_uuid": "62c79f0907239a2b696222336b771488272990c5", "_cell_guid": "672782b0-38ec-4eff-b902-377c323a6c83"}}, {"cell_type": "markdown", "source": ["### Download and run a modified version of Alex Ozerin's baseline estimator\n", "\n", "* create an (empty) working directory to run the training from\n", "* create a subdirectory `trainer`\n", "* download the adapted kernel from https://www.kaggle.com/holzner/alex-ozerin-s-estimator-for-google-ml-engine/ \n", "   * store it as `trainer/task.py` (i.e. in the subdirectory `trainer/`)\n", "* create an empty file `trainer/__init__.py`. On OSX/Linux you can do this from the command line with the command:\n", "`touch trainer/__init__.py`\n", "\n", "#### Running training locally for a quick test\n", "\n", "* It is recommended to first run the kernel on your local machine before submitting it to Google's ML Engine to see if everything works\n", "   * You'll need a python 2 environment with tensorflow installed. Python 3 is not supported at the moment by the ML Engine as far as I understood.\n", "\n", "   *   set a few environment variables (assuming you have a Bourne like shell such as bash or zsh):\n", "\n", "```\n", "INDIR=... # relative or absolute path of the location of unpacked train and test data directories\n", "OUTDIR=... # relative or absolute path of directory where to store model checkpoints and test set predictions\n", "```\n", "   * run a abbreviated version of the training:\n", "```\n", "gcloud ml-engine local train \\\n", "   --module-name trainer.task \\\n", "   --package-path trainer/ \\\n", "   --job-dir $OUTDIR \\\n", "   -- \\\n", "   --indir $INDIR \\\n", "   --train-steps 100 \\\n", "   --train-steps-per-iteration 10 \\\n", "   --eval-steps 1\n", "```\n", "   * This should produce a file `$OUTDIR/submission.csv`\n", "   \n", "#### Submitting a training job to ML Engine\n", "   \n", "   If the above was successful, you can try submitting your job to the ML Engine:\n", "   \n", "```\n", "# name of the bucket you created above, without the gs:// prefix\n", "BUCKET=... \n", "\n", "# region where you created the VM and bucket above (us-east1 in my case)\n", "REGION=... \n", "\n", "# select a scale tier which has GPUs\n", "# see https://cloud.google.com/ml-engine/docs/training-overview\n", "SCALE_TIER=BASIC_GPU\n", "\n", "# Create a job id. This must be unique (within your project as far as I understood)\n", "# so including the job creation time is a good choice (assuming you are not\n", "# creating more than one job per second)\n", "# note: no special characters (apart from underscores) allowed in job names\n", "JOB_NAME=job_$(date +%Y_%m_%d_%H%M%S)\n", "\n", "# storage location where to store trained model\n", "# files and predictions\n", "# use the input bucket also for writing out submissions\n", "# for simplicity\n", "# (a better design would be to use a separate\n", "#  storage bucket for output)\n", "OUTDIR=gs://$BUCKET/output-$JOB_NAME\n", "\n", "# storage location of the packed train and test wav files\n", "# IMPORTANT: do not leave a trailing / here !\n", "# if the bucket file url has another // in it\n", "# the job may NOT find the files !\n", "INDIR=gs://${BUCKET}\n", "\n", "echo \"test set predictions will be stored in $OUTDIR/submission.csv\"\n", "\n", "# submit the training and prediction job\n", "gcloud ml-engine jobs submit training $JOB_NAME \\\n", "  --region $REGION \\\n", "  --scale-tier $SCALE_TIER \\\n", "  --module-name trainer.task \\\n", "  --package-path trainer/ \\\n", "  --runtime-version 1.4 \\\n", "  --job-dir $OUTDIR \\\n", "  -- \\\n", "  --indir $INDIR\n", "  ```\n", "\n", "* The above lines are best stored in a single script which can be used to submit jobs after generating a job id with the current timestamp.\n", "* To monitor your job, either use the command printed after job submission which is of the form\n", "\n", "`gcloud ml-engine jobs stream-logs $JOBID`\n", "\n", "(the exact command is printed by the gcloud command after successful submission). In my case, running the training and test set prediction took about 41 minutes with the default training parameters from Alex' original kernel.\n", "* Alternatively, you can go to the ML engine job overview page here: https://console.cloud.google.com/mlengine/jobs and follow the job's progress by clicking on the 'View logs' item on the line corresponding to your job.\n", "* During training (or after) you can also run tensorboard on your local machine and monitor training progress by passing the Google storage bucket URL as logdir:\n", "\n", "`tensorboard --logdir $OUTDIR`\n", "\n", "and navigate your browser to http://localhost:6006 . Note that it takes somewhat more time for the dashboard to appear than it does for local directories.\n", "\n", "#### Fetch the predicted output\n", "\n", "* Once  the job has completed successfully, you can retrieve the submission.csv file from the output storage bucket. To copy it to the current directory on your local machine (assuming you still have the environment variables from above set):\n", "\n", "`gsutil cp $OUTDIR/submission.csv $PWD`\n", "\n", "* You can list the output directory contents with\n", "\n", "`gsutil ls $OUTDIR`\n", "\n", "You may want to delete the job output directory when you don't need it anymore to avoid being billed unnecessarily for storage.\n", "\n", "\n", "\n", "\n", "\n", "\n", "\n", "   \n", "\n", "   \n", "\n"], "metadata": {"_uuid": "fad5588ab9c2351b903bb6b255c48f83169a6580", "_cell_guid": "d77040c3-6fef-42ab-a1dd-54393068ae69"}}, {"cell_type": "code", "outputs": [], "metadata": {"collapsed": true, "_uuid": "177ac1af3ed21f0752ba4027b45f835ff5a5c4d5", "_cell_guid": "c8f72ed7-2d8c-446b-ba4e-5c6e03938658"}, "execution_count": null, "source": []}]}