{
  "id": 41478,
  "title": "End-to-end image classification solution... and free $100 in gpu ",
  "url": "/competitions/cdiscount-image-classification-challenge/discussion/41478",
  "author_name": "",
  "post_date": "2017-10-18T15:51:49.813466100Z",
  "votes": 34,
  "comment_count": 64,
  "views": 0,
  "content": "<p>Hi everybody!\nAt deepsense.ai we love competitions and this time we really want to spread our love with the community.</p>\n\n<p>We have prepared a nice end-to-end image classification pipeline that is easily extendable and tweakable. \nCheck the code here:\n<a href=\"https://github.com/deepsense-ai/cdiscount-starter\">https://github.com/deepsense-ai/cdiscount-starter</a>\nIt gets your from raw data to submission but it's highly modular so you can build your solution upon that.</p>\n\n<p>As you well know gpu is going to be burned left and right during this competition.\nSo we have decided that we will give the community an opportunity to create and train amazing models for free!\nThat is correct. \nSign up to <a href=\"https://neptune.ml/\">https://neptune.ml</a> and get $100 to train your models. You can get a really nice submission with that so use it wisely.</p>\n\n<p>You are welcome and good luck!</p>",
  "messages": [
    {
      "id": "232854",
      "postDate": "10/18/2017 15:51:49",
      "content": "<p>Hi everybody!\nAt deepsense.ai we love competitions and this time we really want to spread our love with the community.</p>\n\n<p>We have prepared a nice end-to-end image classification pipeline that is easily extendable and tweakable. \nCheck the code here:\n<a href=\"https://github.com/deepsense-ai/cdiscount-starter\">https://github.com/deepsense-ai/cdiscount-starter</a>\nIt gets your from raw data to submission but it's highly modular so you can build your solution upon that.</p>\n\n<p>As you well know gpu is going to be burned left and right during this competition.\nSo we have decided that we will give the community an opportunity to create and train amazing models for free!\nThat is correct. \nSign up to <a href=\"https://neptune.ml/\">https://neptune.ml</a> and get $100 to train your models. You can get a really nice submission with that so use it wisely.</p>\n\n<p>You are welcome and good luck!</p>",
      "rawMarkdown": "Hi everybody!\nAt deepsense.ai we love competitions and this time we really want to spread our love with the community.\n\nWe have prepared a nice end-to-end image classification pipeline that is easily extendable and tweakable. \nCheck the code here:\nhttps://github.com/deepsense-ai/cdiscount-starter\nIt gets your from raw data to submission but it's highly modular so you can build your solution upon that.\n\nAs you well know gpu is going to be burned left and right during this competition.\nSo we have decided that we will give the community an opportunity to create and train amazing models for free!\nThat is correct. \nSign up to https://neptune.ml and get $100 to train your models. You can get a really nice submission with that so use it wisely.\n\nYou are welcome and good luck!",
      "votes": null
    },
    {
      "id": "232864",
      "postDate": "10/18/2017 16:24:36",
      "content": "<p>Hi Jakub, deepsense.ai is awesome! For sure I will take a look at it. Thanks!</p>",
      "rawMarkdown": "Hi Jakub, deepsense.ai is awesome! For sure I will take a look at it. Thanks!",
      "votes": null
    },
    {
      "id": "232898",
      "postDate": "10/18/2017 18:10:05",
      "content": "<p>@Jakub, do you believe the free account can handle this competition datasets? </p>",
      "rawMarkdown": "Jakub, do you believe the free account can handle this competition datasets?",
      "votes": null
    },
    {
      "id": "232976",
      "postDate": "10/18/2017 23:20:05",
      "content": "<p>We mounted the dataset for you (/public/Cdiscount) so you can go ahead and start training right away! </p>",
      "rawMarkdown": "We mounted the dataset for you (/public/Cdiscount) so you can go ahead and start training right away!",
      "votes": null
    },
    {
      "id": "232998",
      "postDate": "10/19/2017 01:05:55",
      "content": "<p>Thanks Jakub. I am trying to run some experiments and i find module bson from pymongo is missing...</p>",
      "rawMarkdown": "Thanks Jakub. I am trying to run some experiments and i find module bson from pymongo is missing...",
      "votes": null
    },
    {
      "id": "233028",
      "postDate": "10/19/2017 02:45:52",
      "content": "<p>There is missing meta data Error:\nUsing TensorFlow backend.\nTraceback (most recent call last):\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/deepsense/neptune/job_wrapper.py\", line 138, in \n    execute()\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/deepsense/neptune/job_wrapper.py\", line 134, in execute\n    execfile(job_filepath, job_globals)\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/past/builtins/misc.py\", line 82, in execfile\n    exec_(code, myglobals, mylocals)\n  File \"experiment_manager.py\", line 273, in \n    registered_actionsargs.action\n  File \"experiment_manager.py\", line 20, in run_pipeline\n    train_valid_split(args)\n  File \"/home/yanchao/cdiscount-starter/utils.py\", line 40, in wrapper\n    function(*args, **kwargs)\n  File \"experiment_manager.py\", line 154, in train_valid_split\n    meta_data = pd.read_csv(meta_data_filepath)\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/pandas/io/parsers.py\", line 655, in parser_f\n    return _read(filepath_or_buffer, kwds)\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/pandas/io/parsers.py\", line 405, in _read\n    parser = TextFileReader(filepath_or_buffer, **kwds)\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/pandas/io/parsers.py\", line 764, in <strong>init</strong>\n    self._make_engine(self.engine)\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/pandas/io/parsers.py\", line 985, in _make_engine\n    self._engine = CParserWrapper(self.f, **self.options)\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/pandas/io/parsers.py\", line 1605, in <strong>init</strong>\n    self._reader = parsers.TextReader(src, **kwds)\n  File \"pandas/_libs/parsers.pyx\", line 394, in pandas._libs.parsers.TextReader.<strong>cinit</strong> (pandas/_libs/parsers.c:4209)\n  File \"pandas/_libs/parsers.pyx\", line 710, in pandas._libs.parsers.TextReader._setup_parser_source (pandas/_libs/parsers.c:8873)\nFileNotFoundError: File b'/public/Cdiscount/meta/meta_train.csv' does not exist</p>",
      "rawMarkdown": "There is missing meta data Error:\nUsing TensorFlow backend.\nTraceback (most recent call last):\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/deepsense/neptune/job_wrapper.py\", line 138, in",
      "votes": null
    },
    {
      "id": "233036",
      "postDate": "10/19/2017 02:55:09",
      "content": "<p>My bad , I was running on local machine.</p>",
      "rawMarkdown": "My bad , I was running on local machine.",
      "votes": null
    },
    {
      "id": "233080",
      "postDate": "10/19/2017 06:28:00",
      "content": "<p>Hi Aloisio,\npymongo is installed on neptune just before running it and I have just confirmed on another machine that cloning the repo and running <code>source run_neptune_command.sh</code> works just fine.\nAre you running it with neptune run or neptune send? What is your os?\nIf you are running neptune run then you should first install all the requirements locally.</p>",
      "rawMarkdown": "Hi Aloisio,\npymongo is installed on neptune just before running it and I have just confirmed on another machine that cloning the repo and running `source run_neptune_command.sh` works just fine.\nAre you running it with neptune run or neptune send? What is your os?\nIf you are running neptune run then you should first install all the requirements locally.",
      "votes": null
    },
    {
      "id": "233161",
      "postDate": "10/19/2017 13:42:18",
      "content": "<p>I am running with neptune send. Its funny because when I commented the BSON part of the code and the module import it runs OK.</p>",
      "rawMarkdown": "I am running with neptune send. Its funny because when I commented the BSON part of the code and the module import it runs OK.",
      "votes": null
    },
    {
      "id": "233163",
      "postDate": "10/19/2017 13:42:53",
      "content": "<p>I will try again later. Thanks !</p>",
      "rawMarkdown": "I will try again later. Thanks !",
      "votes": null
    },
    {
      "id": "233174",
      "postDate": "10/19/2017 14:24:14",
      "content": "<p>Hi Jakub, thanks a lot for sharing! May I get some hints on how to resolve the error in the attached screenshot? Thanks!</p>",
      "rawMarkdown": "Hi Jakub, thanks a lot for sharing! May I get some hints on how to resolve the error in the attached screenshot? Thanks!",
      "votes": null
    },
    {
      "id": "233280",
      "postDate": "10/19/2017 18:58:10",
      "content": "<p>.</p>",
      "rawMarkdown": ".",
      "votes": null
    },
    {
      "id": "233281",
      "postDate": "10/19/2017 19:01:16",
      "content": "<p>Have you done this?</p>\n\n<p>from deepsense import neptune</p>",
      "rawMarkdown": "Have you done this?\n\nfrom deepsense import neptune",
      "votes": null
    },
    {
      "id": "233346",
      "postDate": "10/19/2017 22:30:22",
      "content": "<p>Can someone who's run the baseline tell me how long the baseline takes to run? I can't tell if it's stuck or just predicting.</p>",
      "rawMarkdown": "Can someone who's run the baseline tell me how long the baseline takes to run? I can't tell if it's stuck or just predicting.",
      "votes": null
    },
    {
      "id": "233415",
      "postDate": "10/20/2017 06:29:39",
      "content": "<p>Hi Steven. It takes a while to predict on this test set (I will time it and post the info on that). So I think you  just need to be patient.</p>",
      "rawMarkdown": "Hi Steven. It takes a while to predict on this test set (I will time it and post the info on that). So I think you  just need to be patient.",
      "votes": null
    },
    {
      "id": "233416",
      "postDate": "10/20/2017 06:31:40",
      "content": "<p>You are correct neptune needs to be installed. I merged PR that changed neptune to hard requirement to make it explicit.</p>",
      "rawMarkdown": "You are correct neptune needs to be installed. I merged PR that changed neptune to hard requirement to make it explicit.",
      "votes": null
    },
    {
      "id": "233418",
      "postDate": "10/20/2017 06:34:36",
      "content": "<p>Some people asked me to be able to contribute to this repo. Of course you can! That is the point,\nSo if you want to contribute in any way just make a pull request, we will review and merge it.</p>",
      "rawMarkdown": "Some people asked me to be able to contribute to this repo. Of course you can! That is the point,\nSo if you want to contribute in any way just make a pull request, we will review and merge it.",
      "votes": null
    },
    {
      "id": "233530",
      "postDate": "10/20/2017 13:19:52",
      "content": "<p>Thanks Jakub!</p>",
      "rawMarkdown": "Thanks Jakub!",
      "votes": null
    },
    {
      "id": "233547",
      "postDate": "10/20/2017 14:02:24",
      "content": "<p>I am not running the full dataset yet. Instead, I am using only images of less frequent classes to tweak LR, Decay and some aspect of my classifier design with neptune grid search. It´s very cool! Trainnig 10 epochs of InceptionV3 with Keras takes only 10 minutes!    </p>",
      "rawMarkdown": "I am not running the full dataset yet. Instead, I am using only images of less frequent classes to tweak LR, Decay and some aspect of my classifier design with neptune grid search. It´s very cool! Trainnig 10 epochs of InceptionV3 with Keras takes only 10 minutes!",
      "votes": null
    },
    {
      "id": "233549",
      "postDate": "10/20/2017 14:06:20",
      "content": "<p>InceptionV3, 400 less frequent classes gridsearch in neptune. 10 minutes, 10 epochs.</p>",
      "rawMarkdown": "InceptionV3, 400 less frequent classes gridsearch in neptune. 10 minutes, 10 epochs.",
      "votes": null
    },
    {
      "id": "233701",
      "postDate": "10/20/2017 21:43:08",
      "content": "<p>How exactly does neptune grid search work? I'm just starting to play around with the platform and its pretty cool! I like how I don't need to monitor VMs and their prices are very competitive.</p>",
      "rawMarkdown": "How exactly does neptune grid search work? I'm just starting to play around with the platform and its pretty cool! I like how I don't need to monitor VMs and their prices are very competitive.",
      "votes": null
    },
    {
      "id": "233706",
      "postDate": "10/20/2017 21:58:52",
      "content": "<p>It's nice and simple. \nIt creates a job for each node of the grid search and simply runs them as if they were single experiments. After it justs pools the results.  </p>",
      "rawMarkdown": "It's nice and simple. \nIt creates a job for each node of the grid search and simply runs them as if they were single experiments. After it justs pools the results.",
      "votes": null
    },
    {
      "id": "234342",
      "postDate": "10/23/2017 02:52:46",
      "content": "<p>.</p>",
      "rawMarkdown": ".",
      "votes": null
    },
    {
      "id": "234830",
      "postDate": "10/24/2017 07:03:29",
      "content": "<p>Thanks a lot for your offer. As I am currently searching a suitable cloud environment for our company, it comes at the perfect time.</p>\n\n<p>However, when I use the CDiscount Script, the job gets queued for a really long time (16 hours right now). Is there something I can do against it?</p>",
      "rawMarkdown": "Thanks a lot for your offer. As I am currently searching a suitable cloud environment for our company, it comes at the perfect time.\n\nHowever, when I use the CDiscount Script, the job gets queued for a really long time (16 hours right now). Is there something I can do against it?",
      "votes": null
    },
    {
      "id": "234993",
      "postDate": "10/24/2017 14:41:29",
      "content": "<p>Did you find a solution? I am having the same issue.</p>",
      "rawMarkdown": "Did you find a solution? I am having the same issue.",
      "votes": null
    },
    {
      "id": "235025",
      "postDate": "10/24/2017 16:25:55",
      "content": "<p>Hi Maximilian and sorry for the trouble. There was a lot of interest and we had a temporary problem with the resources. Everything should be working smoothly now.</p>",
      "rawMarkdown": "Hi Maximilian and sorry for the trouble. There was a lot of interest and we had a temporary problem with the resources. Everything should be working smoothly now.",
      "votes": null
    },
    {
      "id": "235026",
      "postDate": "10/24/2017 16:27:31",
      "content": "<p>We made some improvements to the starter code. I would reccomend that you download the new version and try with it. </p>",
      "rawMarkdown": "We made some improvements to the starter code. I would reccomend that you download the new version and try with it.",
      "votes": null
    },
    {
      "id": "235036",
      "postDate": "10/24/2017 16:42:17",
      "content": "<p>Really no need for excuses here :) Just wanted to know whether I did something wrong. The job starts directly now.</p>",
      "rawMarkdown": "Really no need for excuses here :) Just wanted to know whether I did something wrong. The job starts directly now.",
      "votes": null
    },
    {
      "id": "235066",
      "postDate": "10/24/2017 17:20:09",
      "content": "<p>Happy to hear that!</p>",
      "rawMarkdown": "Happy to hear that!",
      "votes": null
    },
    {
      "id": "235518",
      "postDate": "10/25/2017 16:35:28",
      "content": "<p>Hi!</p>\n\n<p>Can you please explain how can I get meta_ files for local run? \nI tried to look it up in README or your post, but couldn't find anything.</p>",
      "rawMarkdown": "Hi!\n\nCan you please explain how can I get meta_ files for local run? \nI tried to look it up in README or your post, but couldn't find anything.",
      "votes": null
    },
    {
      "id": "236174",
      "postDate": "10/26/2017 19:49:52",
      "content": "<p>Hi Steven. Here an submission command example:</p>\n\n<p>neptune send --environment keras-2.0-cpu-py3 --worker gcp-gpu-large --input train.json.tar.gz/train.json -- '--lr %[0.001, 0.002, 0.003] --decay %[0.001, 0.005, 0.009] --n_epochs %[20]'</p>",
      "rawMarkdown": "Hi Steven. Here an submission command example:\n\nneptune send --environment keras-2.0-cpu-py3 --worker gcp-gpu-large --input train.json.tar.gz/train.json -- '--lr %[0.001, 0.002, 0.003] --decay %[0.001, 0.005, 0.009] --n_epochs %[20]'",
      "votes": null
    },
    {
      "id": "236175",
      "postDate": "10/26/2017 19:53:31",
      "content": "<p>And here a piece of code you should have in your main.py to manage the parameters:</p>\n\n<pre><code>params_parser = argparse.ArgumentParser()\nparams_parser.add_argument('--lr', type=float, default=0.001)\nparams_parser.add_argument('--decay', type=float, default=0.005)\nparams_parser.add_argument('--n_epochs', type=int, default=10)\n\nparams = params_parser.parse_args()\n##################################\nDECAY = params.decay\nLR = params.lr\nN_EPOCHS = params.n_epochs\n</code></pre>",
      "rawMarkdown": "And here a piece of code you should have in your main.py to manage the parameters:\n\n    params_parser = argparse.ArgumentParser()\n    params_parser.add_argument('--lr', type=float, default=0.001)\n    params_parser.add_argument('--decay', type=float, default=0.005)\n    params_parser.add_argument('--n_epochs', type=int, default=10)\n    \n    params = params_parser.parse_args()\n    ##################################\n    DECAY = params.decay\n    LR = params.lr\n    N_EPOCHS = params.n_epochs",
      "votes": null
    },
    {
      "id": "236329",
      "postDate": "10/27/2017 04:39:13",
      "content": "<p>Hi Jakub, \nThank You for the free 100$ credit on neptune. \nI tried to run locally. But I am seeing this issue. \n<code>Started job execution, id: ae4398ce-7e32-4bfa-86f0-eee06e7b1770</code>\n<code>To browse the job, follow:</code>\n<code>https://boole.neptune.ml/#dashboard/job/ae4398ce-7e32-4bfa-86f0-eee06e7b1770?getStartedState=folded</code>\n<code>Traceback (most recent call last):</code>\n  <code>File \"/Users/yarlab/neptune/lib/python2.7/site-packages/deepsense/neptune/job_wrapper.py\", line 138, in</code>\n<code>execute()</code>\n  <code>File \"/Users/yarlab/neptune/lib/python2.7/site-packages/deepsense/neptune/job_wrapper.py\", line 134, in execute</code>\n    execfile(job_filepath, job_globals)<code>\n</code>File \"run_manager.py\", line 252<code>\n</code>config_merged = {**exp_config, **data_config}<code>\n</code>                  ^<code>\n</code>SyntaxError: invalid syntax<code>\n</code>Process exited with return code 1.`</p>",
      "rawMarkdown": "Hi Jakub, \nThank You for the free 100$ credit on neptune. \nI tried to run locally. But I am seeing this issue. \n`Started job execution, id: ae4398ce-7e32-4bfa-86f0-eee06e7b1770`\n`To browse the job, follow:`\n`https://boole.neptune.ml/#dashboard/job/ae4398ce-7e32-4bfa-86f0-eee06e7b1770?getStartedState=folded`\n`Traceback (most recent call last):`\n  `File \"/Users/yarlab/neptune/lib/python2.7/site-packages/deepsense/neptune/job_wrapper.py\", line 138, in `\n`execute()`\n  `File \"/Users/yarlab/neptune/lib/python2.7/site-packages/deepsense/neptune/job_wrapper.py\", line 134, in execute`\n    execfile(job_filepath, job_globals)`\n  `File \"run_manager.py\", line 252`\n    `config_merged = {**exp_config, **data_config}`\n    `                  ^`\n`SyntaxError: invalid syntax`\n`Process exited with return code 1.`",
      "votes": null
    },
    {
      "id": "236343",
      "postDate": "10/27/2017 04:47:53",
      "content": "<p>nevermind fixed it. Thank you</p>",
      "rawMarkdown": "nevermind fixed it. Thank you",
      "votes": null
    },
    {
      "id": "236584",
      "postDate": "10/27/2017 17:43:59",
      "content": "<p>Command given is same as mentioned in the readme File : source run_neptune_command.sh</p>\n\n<p>Fails with this error :</p>\n\n<p>2017-10-27 17:37:27.984764: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.\n2017-10-27 17:37:27.984773: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.\n2017-10-27 17:37:28.758937: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:893] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n2017-10-27 17:37:28.760243: I tensorflow/core/common_runtime/gpu/gpu_device.cc:940] Found device 0 with properties:\nname: Tesla K80\nmajor: 3 minor: 7 memoryClockRate (GHz) 0.8235\npciBusID 0000:00:04.0\nTotal memory: 11.92GiB\nFree memory: 11.86GiB\n2017-10-27 17:37:28.760275: I tensorflow/core/common_runtime/gpu/gpu_device.cc:961] DMA: 0\n2017-10-27 17:37:28.760283: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0:   Y\n2017-10-27 17:37:28.760292: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Creating TensorFlow device (/gpu:0) -&gt; (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0)\nException in thread Thread-11:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in <strong>next</strong>\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'\nException in thread Thread-8:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in <strong>next</strong>\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'</p>\n\n<p>Exception in thread Thread-9:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in <strong>next</strong>\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'</p>\n\n<p>Exception in thread Thread-10:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in <strong>next</strong>\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'</p>\n\n<p>Traceback (most recent call last):\n  File \"/usr/local/lib/python3.5/dist-packages/deepsense/neptune/job_wrapper.py\", line 138, in \n  File \"/usr/local/lib/python3.5/dist-packages/deepsense/neptune/job_wrapper.py\", line 134, in execute\n  File \"/usr/local/lib/python3.5/dist-packages/past/builtins/misc.py\", line 82, in execfile\n    exec_(code, myglobals, mylocals)\n  File \"run_manager.py\", line 263, in \n    registered_actionsargs.action\n  File \"run_manager.py\", line 21, in run_pipeline\n    train_pipeline(args)\n  File \"run_manager.py\", line 51, in train_pipeline\n    img_dataset_filepath=train_filepath)\n  File \"/neptune/preprocessing.py\", line 50, in fit\n    img_dataset_filepath=img_dataset_filepath)\n  File \"/neptune/postprocessing.py\", line 10, in fit\n    step.fit(X, y, validation_data=validation_data, img_dataset_filepath=img_dataset_filepath)\n  File \"/neptune/pipelines.py\", line 77, in fit\n    self.deep_model.fit(datagens)\n  File \"/neptune/models.py\", line 37, in fit\n    **self.training_cfg)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/legacy/interfaces.py\", line 87, in wrapper\n    return func(*args, **kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/engine/training.py\", line 2011, in fit_generator\n    generator_output = next(output_generator)\nStopIteration</p>",
      "rawMarkdown": "Command given is same as mentioned in the readme File : source run_neptune_command.sh\n\nFails with this error :\n\n2017-10-27 17:37:27.984764: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.\n2017-10-27 17:37:27.984773: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.\n2017-10-27 17:37:28.758937: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:893] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n2017-10-27 17:37:28.760243: I tensorflow/core/common_runtime/gpu/gpu_device.cc:940] Found device 0 with properties:\nname: Tesla K80\nmajor: 3 minor: 7 memoryClockRate (GHz) 0.8235\npciBusID 0000:00:04.0\nTotal memory: 11.92GiB\nFree memory: 11.86GiB\n2017-10-27 17:37:28.760275: I tensorflow/core/common_runtime/gpu/gpu_device.cc:961] DMA: 0\n2017-10-27 17:37:28.760283: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0:   Y\n2017-10-27 17:37:28.760292: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Creating TensorFlow device (/gpu:0) -&gt; (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0)\nException in thread Thread-11:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in __next__\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'\nException in thread Thread-8:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in __next__\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'\n\n\nException in thread Thread-9:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in __next__\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'\n\nException in thread Thread-10:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in __next__\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'\n\nTraceback (most recent call last):\n  File \"/usr/local/lib/python3.5/dist-packages/deepsense/neptune/job_wrapper.py\", line 138, in",
      "votes": null
    },
    {
      "id": "236739",
      "postDate": "10/28/2017 00:51:11",
      "content": "<p>Thanks for the promo, I started playing with neptune platform and it's fun !\nBut can I only run jobs on single GPU Tesla 80K ?\nIs there a way to run a job on multiple faster GPUs ?\nCheers !</p>",
      "rawMarkdown": "Thanks for the promo, I started playing with neptune platform and it's fun !\nBut can I only run jobs on single GPU Tesla 80K ?\nIs there a way to run a job on multiple faster GPUs ?\nCheers !",
      "votes": null
    },
    {
      "id": "236746",
      "postDate": "10/28/2017 01:39:36",
      "content": "<p>It looks like the neptune platform is built on top of GCP with clusters for either 4 or 8 k80s using the argument <code>--worker gcp-gpu-medium</code> or <code>--worker gcp-gpu-large</code>. In my experience with GCP you can use it like a single GPU with expected diminishing returns.</p>",
      "rawMarkdown": "It looks like the neptune platform is built on top of GCP with clusters for either 4 or 8 k80s using the argument `--worker gcp-gpu-medium` or `--worker gcp-gpu-large`. In my experience with GCP you can use it like a single GPU with expected diminishing returns.",
      "votes": null
    },
    {
      "id": "237408",
      "postDate": "10/30/2017 08:56:18",
      "content": "<p>Hi Phillip,\nYou simply need to run </p>\n\n<p><code>\npython run_manager.py create_metadata\n</code></p>\n\n<p>Meta data will be stored in your <code>meta_data_dir</code> specified in the data_config.yaml</p>\n\n<p>Thank you for pointing that out, I will add this to the readme.</p>",
      "rawMarkdown": "Hi Phillip,\nYou simply need to run \n\n```\npython run_manager.py create_metadata\n```\n\nMeta data will be stored in your `meta_data_dir` specified in the data_config.yaml\n\nThank you for pointing that out, I will add this to the readme.",
      "votes": null
    },
    {
      "id": "237414",
      "postDate": "10/30/2017 09:16:03",
      "content": "<p>It is fixed now.  Sorry for the trouble.</p>",
      "rawMarkdown": "It is fixed now.  Sorry for the trouble.",
      "votes": null
    },
    {
      "id": "237455",
      "postDate": "10/30/2017 10:56:30",
      "content": "<p>Unfortunately at the moment we don't have workers that support multi-gpu and <code>gcp-gpu-large</code> is the strongest machine. That will likely change in the future so keep your finger on the pulse.</p>",
      "rawMarkdown": "Unfortunately at the moment we don't have workers that support multi-gpu and `gcp-gpu-large` is the strongest machine. That will likely change in the future so keep your finger on the pulse.",
      "votes": null
    },
    {
      "id": "237960",
      "postDate": "10/31/2017 11:44:51",
      "content": "<p>Hi, check our <a href=\"https://blog.deepsense.ai/image-classification-sample-solution-kaggle/\">blog post featuring starter code</a> prepared by Jakub. Post has short video, where Jakub gives quick code overview.</p>",
      "rawMarkdown": "Hi, check our [blog post featuring starter code][1] prepared by Jakub. Post has short video, where Jakub gives quick code overview.\n\n\n  [1]: https://blog.deepsense.ai/image-classification-sample-solution-kaggle/",
      "votes": null
    },
    {
      "id": "242691",
      "postDate": "11/12/2017 11:50:56",
      "content": "<p>Hey! Thanks for the share! Does this include storage?</p>",
      "rawMarkdown": "Hey! Thanks for the share! Does this include storage?",
      "votes": null
    },
    {
      "id": "242849",
      "postDate": "11/12/2017 19:27:26",
      "content": "<p>Hi Sinish. The data is mounted in neptune so you can start experimenting right of the bat.</p>",
      "rawMarkdown": "Hi Sinish. The data is mounted in neptune so you can start experimenting right of the bat.",
      "votes": null
    },
    {
      "id": "243881",
      "postDate": "11/15/2017 03:56:29",
      "content": "<p>Did you get a chance to post the time to run?  I can't seem to find it posted in the discussions.  Thanks for any help.</p>",
      "rawMarkdown": "Did you get a chance to post the time to run?  I can't seem to find it posted in the discussions.  Thanks for any help.",
      "votes": null
    },
    {
      "id": "245861",
      "postDate": "11/19/2017 22:08:23",
      "content": "<p>I run the code using neptune as per instructions, it gets stuck after 10 epochs, I do not see any file in the output folder. Please help. </p>",
      "rawMarkdown": "I run the code using neptune as per instructions, it gets stuck after 10 epochs, I do not see any file in the output folder. Please help.",
      "votes": null
    },
    {
      "id": "246689",
      "postDate": "11/21/2017 16:15:50",
      "content": "<p>I am having the same issue.  Any help would be much appreciated.  Thanks!</p>",
      "rawMarkdown": "I am having the same issue.  Any help would be much appreciated.  Thanks!",
      "votes": null
    },
    {
      "id": "246711",
      "postDate": "11/21/2017 16:57:10",
      "content": "<p>How would we access the data in Neptune using a Jupyter Notebook?  Would the \"path to inputs visible to notebook\" = \"/public/Cdiscount/test.bson\" for the test.bson file?  Thanks for any help!</p>",
      "rawMarkdown": "How would we access the data in Neptune using a Jupyter Notebook?  Would the \"path to inputs visible to notebook\" = \"/public/Cdiscount/test.bson\" for the test.bson file?  Thanks for any help!",
      "votes": null
    },
    {
      "id": "247531",
      "postDate": "11/23/2017 09:49:33",
      "content": "<p>Having the same issue here, anyone got a solution for this? Thanks.</p>",
      "rawMarkdown": "Having the same issue here, anyone got a solution for this? Thanks.",
      "votes": null
    },
    {
      "id": "247837",
      "postDate": "11/24/2017 03:50:08",
      "content": "<p>I have the same issue and it has been running for 3h 42min. On the dash board for the job I got to by copy pasting the link given after running the command \"source run_neptune_command.sh\", it shows an orange triangle at the top left corner just below \"starter_100_100, Running for 4 hours\".</p>",
      "rawMarkdown": "I have the same issue and it has been running for 3h 42min. On the dash board for the job I got to by copy pasting the link given after running the command \"source run_neptune_command.sh\", it shows an orange triangle at the top left corner just below \"starter_100_100, Running for 4 hours\".",
      "votes": null
    },
    {
      "id": "249693",
      "postDate": "11/29/2017 01:45:27",
      "content": "<p>No improvement, I run source run_neptune_command.sh again. Same it hangs after 2 hours.</p>",
      "rawMarkdown": "No improvement, I run source run_neptune_command.sh again. Same it hangs after 2 hours.",
      "votes": null
    },
    {
      "id": "256418",
      "postDate": "12/11/2017 22:40:15",
      "content": "<p>Hi Pradeep. I have just tried to reproduce your problem but after 3 hours the 100 classes 100 images starter has succeeded getting 0.21 on the LB. The submission is in the output/project_data/submissions folder. </p>",
      "rawMarkdown": "Hi Pradeep. I have just tried to reproduce your problem but after 3 hours the 100 classes 100 images starter has succeeded getting 0.21 on the LB. The submission is in the output/project_data/submissions folder.",
      "votes": null
    },
    {
      "id": "256907",
      "postDate": "12/12/2017 22:39:25",
      "content": "<p>@Jakub Czakon, the point is running the starter code on the neptune.ml system keeps giving lost connection error so many times, it eats up so much valuable time and money. In the end mine did run but could not finish before I ran out of the $100. Going by my short experience with it, I felt it would be a waste of money to subscribe. I still feel bad that I wasted so much time trying it out. Plus if you check the post on the very top here by @Pradeep, 3 other people including me complained over 20 days ago and you did not respond until today. To me, that just does not seem right.</p>",
      "rawMarkdown": "Jakub Czakon, the point is running the starter code on the neptune.ml system keeps giving lost connection error so many times, it eats up so much valuable time and money. In the end mine did run but could not finish before I ran out of the $100. Going by my short experience with it, I felt it would be a waste of money to subscribe. I still feel bad that I wasted so much time trying it out. Plus if you check the post on the very top here by @Pradeep, 3 other people including me complained over 20 days ago and you did not respond until today. To me, that just does not seem right.",
      "votes": null
    },
    {
      "id": "256913",
      "postDate": "12/12/2017 22:56:25",
      "content": "<p>You are right YaGana I should have answered sooner and I really apologize for that. In the future, I will make sure to attend any problems swiftly. </p>",
      "rawMarkdown": "You are right YaGana I should have answered sooner and I really apologize for that. In the future, I will make sure to attend any problems swiftly.",
      "votes": null
    },
    {
      "id": "257499",
      "postDate": "12/14/2017 11:08:20",
      "content": "<p>Hello Jakub,</p>\n\n<p>I have an error with experience_config.yaml </p>\n\n<pre><code>The provided job configuration /cdiscount-starter-master/experiment_config.yaml is invalid! Validation errors: 1. Value '100' is not of type 'str'. Path: '/properties/0/value', 2. Value '100' is not of type 'str'. Path: '/properties/1/value', 3. Value '10' is not of type 'str'. Path: '/properties/2/value'\n</code></pre>\n\n<p>If i changed the type to string, i can send the process but i have another error on neptune.ml</p>\n\n<pre><code>75.934818   TypeError: cannot do slice indexing on &lt;class 'pandas.core.indexes.range.RangeIndex'&gt; with these indexers [100] of &lt;class 'str'&gt;\n</code></pre>\n\n<p>Im using neptune-cli (2.4.3)</p>",
      "rawMarkdown": "Hello Jakub,\n\nI have an error with experience_config.yaml \n\n    The provided job configuration /cdiscount-starter-master/experiment_config.yaml is invalid! Validation errors: 1. Value '100' is not of type 'str'. Path: '/properties/0/value', 2. Value '100' is not of type 'str'. Path: '/properties/1/value', 3. Value '10' is not of type 'str'. Path: '/properties/2/value'\n\nIf i changed the type to string, i can send the process but i have another error on neptune.ml\n\n    75.934818\tTypeError: cannot do slice indexing on",
      "votes": null
    },
    {
      "id": "257609",
      "postDate": "12/14/2017 16:04:23",
      "content": "<p>Hi Chris,</p>\n\n<p>You may want to start Notebook directly from the <a href=\"https://neptune.ml/\">Neptune</a>. Simply log in via web, then at the top bar you can see link \"Start Notebook\".</p>\n\n<p>From inside the notebook, you can type: <code>!ls -la /public/Cdiscount/</code> to see Cdiscount data (read only).</p>\n\n<p>Let me mention that bson is not pre-installed, however, you can install it from inside the notebook, via:\n<code>!pip install your_package</code>.</p>",
      "rawMarkdown": "Hi Chris,\n\nYou may want to start Notebook directly from the [Neptune](https://neptune.ml/). Simply log in via web, then at the top bar you can see link \"Start Notebook\".\n\nFrom inside the notebook, you can type: `!ls -la /public/Cdiscount/` to see Cdiscount data (read only).\n\nLet me mention that bson is not pre-installed, however, you can install it from inside the notebook, via:\n`!pip install your_package`.",
      "votes": null
    },
    {
      "id": "257612",
      "postDate": "12/14/2017 16:10:38",
      "content": "<p>Hi YaGana,</p>\n\n<p>Thanks for pointing this out.</p>\n\n<p>Can you post your experiment ID, so that I can check what is going on with your job, especially your 'lost connection error'.</p>\n\n<p>You also may want to re-run your experiment. In this case let me know and we will discuss best way to do it.</p>\n\n<p>Best,</p>\n\n<p>Kamil (who works with Jakub ;-) )</p>",
      "rawMarkdown": "Hi YaGana,\n\nThanks for pointing this out.\n\nCan you post your experiment ID, so that I can check what is going on with your job, especially your 'lost connection error'.\n\nYou also may want to re-run your experiment. In this case let me know and we will discuss best way to do it.\n\nBest,\n\nKamil (who works with Jakub ;-) )",
      "votes": null
    },
    {
      "id": "257613",
      "postDate": "12/14/2017 16:12:37",
      "content": "<p>Hi @Pradeep,</p>\n\n<p>Please post you experiment ID / link to the experiment, so that I can check what is going on.</p>\n\n<p>Also, make sure that you are using the latest version of the starter code.</p>\n\n<p>Best,</p>\n\n<p>Kamil (on behalf of @Jakub)</p>",
      "rawMarkdown": "Hi @Pradeep,\n\nPlease post you experiment ID / link to the experiment, so that I can check what is going on.\n\nAlso, make sure that you are using the latest version of the starter code.\n\nBest,\n\nKamil (on behalf of @Jakub)",
      "votes": null
    },
    {
      "id": "257614",
      "postDate": "12/14/2017 16:25:55",
      "content": "<p>Hi @Pradeep, @Chris, @Yunfeng, @YaGana,</p>\n\n<p>We need your experiment ID in order to check what caused the job to get stuck after some training time. Also, please double check <code>output</code> directory, since temporary connection issues do not cause an experiment to crash.</p>\n\n<p>Please feel free to post your experiment ID.</p>\n\n<p>Cheers,</p>\n\n<p>Kamil (on behalf of @Jakub)</p>",
      "rawMarkdown": "Hi @Pradeep, @Chris, @Yunfeng, @YaGana,\n\nWe need your experiment ID in order to check what caused the job to get stuck after some training time. Also, please double check `output` directory, since temporary connection issues do not cause an experiment to crash.\n\nPlease feel free to post your experiment ID.\n\nCheers,\n\nKamil (on behalf of @Jakub)",
      "votes": null
    },
    {
      "id": "257622",
      "postDate": "12/14/2017 16:38:53",
      "content": "<p>@kamil, I could not see any id but here is the end of the line of the link from my experiment that aborted by itself after running for 9h 43min. .../457f27c2-3d94-46ce-92b8-f5a42a058701\n Or the whole link to it is:\n<a href=\"https://galois.neptune.ml/#dashboard/job/457f27c2-3d94-46ce-92b8-f5a42a058701\">https://galois.neptune.ml/#dashboard/job/457f27c2-3d94-46ce-92b8-f5a42a058701</a></p>\n\n<p>Then when I ru the same thing again, it ran for <strong>3days 8h</strong> until I ran out of credit. I could only see the save model on the output/project_data/..   directory. Here is the link to that one.</p>\n\n<p>...job/af030724-7366-4919-9a69-5e4f8233253b</p>\n\n<p>About re-running my experiment, I am not sure the best way to go about it since I have only few hours left before the contest ends. I am open to suggestions if you can help.</p>\n\n<p>In fact I would not mind just getting a prediction done on the test data with my saved model at this point.</p>",
      "rawMarkdown": "kamil, I could not see any id but here is the end of the line of the link from my experiment that aborted by itself after running for 9h 43min. .../457f27c2-3d94-46ce-92b8-f5a42a058701\n Or the whole link to it is:\nhttps://galois.neptune.ml/#dashboard/job/457f27c2-3d94-46ce-92b8-f5a42a058701\n\nThen when I ru the same thing again, it ran for **3days 8h** until I ran out of credit. I could only see the save model on the output/project_data/..   directory. Here is the link to that one.\n\n...job/af030724-7366-4919-9a69-5e4f8233253b\n\nAbout re-running my experiment, I am not sure the best way to go about it since I have only few hours left before the contest ends. I am open to suggestions if you can help.\n\nIn fact I would not mind just getting a prediction done on the test data with my saved model at this point.",
      "votes": null
    },
    {
      "id": "257630",
      "postDate": "12/14/2017 17:11:27",
      "content": "<p>Hi @Pierre,</p>\n\n<p>Thanks for sharing this with us. I will take a closer look your error and I will contact you with the update.</p>\n\n<p>Best,</p>\n\n<p>Kamil (on behalf of Jakub)</p>",
      "rawMarkdown": "Hi @Pierre,\n\nThanks for sharing this with us. I will take a closer look your error and I will contact you with the update.\n\nBest,\n\nKamil (on behalf of Jakub)",
      "votes": null
    },
    {
      "id": "257639",
      "postDate": "12/14/2017 17:38:32",
      "content": "<p>Hi @YaGana,</p>\n\n<p>Thanks, this is what I need: ID is this part that comes after the <code>job/</code> part of the link. I will take a closer look at your job, and get back to you.</p>\n\n<p>Some ideas:</p>\n\n<ol>\n<li><p>Quick workaround would be to download your saved model and run prediction locally. You should be able to make it before contest ends.</p></li>\n<li><p>I will try to grant you some resources to make prediction, however, I cannot make it within next couple of hours.</p></li>\n</ol>\n\n<p>Best,</p>\n\n<p>Kamil</p>",
      "rawMarkdown": "Hi @YaGana,\n\nThanks, this is what I need: ID is this part that comes after the `job/` part of the link. I will take a closer look at your job, and get back to you.\n\nSome ideas:\n\n1.  Quick workaround would be to download your saved model and run prediction locally. You should be able to make it before contest ends.\n\n2. I will try to grant you some resources to make prediction, however, I cannot make it within next couple of hours.\n\nBest,\n\nKamil",
      "votes": null
    },
    {
      "id": "257645",
      "postDate": "12/14/2017 17:49:45",
      "content": "<p>I have a CPU machine and prediction takes 5.5days on my local machine hence my using a cloud service. My other models are run on FloydHub. I have a prediction running there right now so I cannot use it for my Neptune model in time for submission.</p>",
      "rawMarkdown": "I have a CPU machine and prediction takes 5.5days on my local machine hence my using a cloud service. My other models are run on FloydHub. I have a prediction running there right now so I cannot use it for my Neptune model in time for submission.",
      "votes": null
    },
    {
      "id": "257760",
      "postDate": "12/14/2017 22:55:30",
      "content": "<p>Hi @Pierre,</p>\n\n<p>Your issue is now solved. You can pull new version from the <a href=\"https://github.com/deepsense-ai/cdiscount-starter\">Cdiscount-starter</a> repository.</p>\n\n<p>Best,</p>\n\n<p>K</p>",
      "rawMarkdown": "Hi @Pierre,\n\nYour issue is now solved. You can pull new version from the [Cdiscount-starter](https://github.com/deepsense-ai/cdiscount-starter) repository.\n\nBest,\n\nK",
      "votes": null
    },
    {
      "id": "257761",
      "postDate": "12/14/2017 22:57:03",
      "content": "<p>YaGana,</p>\n\n<p>I am sorry to hear that.</p>\n\n<p>Best of luck with your final submissions!</p>",
      "rawMarkdown": "YaGana,\n\nI am sorry to hear that.\n\nBest of luck with your final submissions!",
      "votes": null
    },
    {
      "id": "956849",
      "postDate": "08/03/2020 20:45:15",
      "content": "<p>Hello Guys. I choose ResNet for İmage Classification. I thınk you should check thıs code\n<a href=\"https://github.com/batuhan3526/ResNet50_on_Cifar_100_Without_Transfer_Learning\">https://github.com/batuhan3526/ResNet50_on_Cifar_100_Without_Transfer_Learning</a></p>",
      "rawMarkdown": "Hello Guys. I choose ResNet for İmage Classification. I thınk you should check thıs code\nhttps://github.com/batuhan3526/ResNet50_on_Cifar_100_Without_Transfer_Learning",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 232864,
      "author_name": "titericz",
      "author_url": "",
      "post_date": "10/18/2017 16:24:36",
      "content": "<p>Hi Jakub, deepsense.ai is awesome! For sure I will take a look at it. Thanks!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 232898,
      "author_name": "aloisiodn",
      "author_url": "",
      "post_date": "10/18/2017 18:10:05",
      "content": "<p>@Jakub, do you believe the free account can handle this competition datasets? </p>",
      "votes": null,
      "replies": [
        {
          "id": 232976,
          "author_name": "jakubczakon",
          "author_url": "",
          "post_date": "10/18/2017 23:20:05",
          "content": "<p>We mounted the dataset for you (/public/Cdiscount) so you can go ahead and start training right away! </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 246711,
          "author_name": "neuralnetworks",
          "author_url": "",
          "post_date": "11/21/2017 16:57:10",
          "content": "<p>How would we access the data in Neptune using a Jupyter Notebook?  Would the \"path to inputs visible to notebook\" = \"/public/Cdiscount/test.bson\" for the test.bson file?  Thanks for any help!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 257609,
          "author_name": "kkaczmarek",
          "author_url": "",
          "post_date": "12/14/2017 16:04:23",
          "content": "<p>Hi Chris,</p>\n\n<p>You may want to start Notebook directly from the <a href=\"https://neptune.ml/\">Neptune</a>. Simply log in via web, then at the top bar you can see link \"Start Notebook\".</p>\n\n<p>From inside the notebook, you can type: <code>!ls -la /public/Cdiscount/</code> to see Cdiscount data (read only).</p>\n\n<p>Let me mention that bson is not pre-installed, however, you can install it from inside the notebook, via:\n<code>!pip install your_package</code>.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 232998,
      "author_name": "aloisiodn",
      "author_url": "",
      "post_date": "10/19/2017 01:05:55",
      "content": "<p>Thanks Jakub. I am trying to run some experiments and i find module bson from pymongo is missing...</p>",
      "votes": null,
      "replies": [
        {
          "id": 233080,
          "author_name": "jakubczakon",
          "author_url": "",
          "post_date": "10/19/2017 06:28:00",
          "content": "<p>Hi Aloisio,\npymongo is installed on neptune just before running it and I have just confirmed on another machine that cloning the repo and running <code>source run_neptune_command.sh</code> works just fine.\nAre you running it with neptune run or neptune send? What is your os?\nIf you are running neptune run then you should first install all the requirements locally.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 233161,
          "author_name": "aloisiodn",
          "author_url": "",
          "post_date": "10/19/2017 13:42:18",
          "content": "<p>I am running with neptune send. Its funny because when I commented the BSON part of the code and the module import it runs OK.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 233163,
          "author_name": "aloisiodn",
          "author_url": "",
          "post_date": "10/19/2017 13:42:53",
          "content": "<p>I will try again later. Thanks !</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 234993,
          "author_name": "maximilianhahn",
          "author_url": "",
          "post_date": "10/24/2017 14:41:29",
          "content": "<p>Did you find a solution? I am having the same issue.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 235026,
          "author_name": "jakubczakon",
          "author_url": "",
          "post_date": "10/24/2017 16:27:31",
          "content": "<p>We made some improvements to the starter code. I would reccomend that you download the new version and try with it. </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 249693,
          "author_name": "ppujari",
          "author_url": "",
          "post_date": "11/29/2017 01:45:27",
          "content": "<p>No improvement, I run source run_neptune_command.sh again. Same it hangs after 2 hours.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 256418,
          "author_name": "jakubczakon",
          "author_url": "",
          "post_date": "12/11/2017 22:40:15",
          "content": "<p>Hi Pradeep. I have just tried to reproduce your problem but after 3 hours the 100 classes 100 images starter has succeeded getting 0.21 on the LB. The submission is in the output/project_data/submissions folder. </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 256907,
          "author_name": "sheriytm",
          "author_url": "",
          "post_date": "12/12/2017 22:39:25",
          "content": "<p>@Jakub Czakon, the point is running the starter code on the neptune.ml system keeps giving lost connection error so many times, it eats up so much valuable time and money. In the end mine did run but could not finish before I ran out of the $100. Going by my short experience with it, I felt it would be a waste of money to subscribe. I still feel bad that I wasted so much time trying it out. Plus if you check the post on the very top here by @Pradeep, 3 other people including me complained over 20 days ago and you did not respond until today. To me, that just does not seem right.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 256913,
          "author_name": "jakubczakon",
          "author_url": "",
          "post_date": "12/12/2017 22:56:25",
          "content": "<p>You are right YaGana I should have answered sooner and I really apologize for that. In the future, I will make sure to attend any problems swiftly. </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 257612,
          "author_name": "kkaczmarek",
          "author_url": "",
          "post_date": "12/14/2017 16:10:38",
          "content": "<p>Hi YaGana,</p>\n\n<p>Thanks for pointing this out.</p>\n\n<p>Can you post your experiment ID, so that I can check what is going on with your job, especially your 'lost connection error'.</p>\n\n<p>You also may want to re-run your experiment. In this case let me know and we will discuss best way to do it.</p>\n\n<p>Best,</p>\n\n<p>Kamil (who works with Jakub ;-) )</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 257613,
          "author_name": "kkaczmarek",
          "author_url": "",
          "post_date": "12/14/2017 16:12:37",
          "content": "<p>Hi @Pradeep,</p>\n\n<p>Please post you experiment ID / link to the experiment, so that I can check what is going on.</p>\n\n<p>Also, make sure that you are using the latest version of the starter code.</p>\n\n<p>Best,</p>\n\n<p>Kamil (on behalf of @Jakub)</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 257622,
          "author_name": "sheriytm",
          "author_url": "",
          "post_date": "12/14/2017 16:38:53",
          "content": "<p>@kamil, I could not see any id but here is the end of the line of the link from my experiment that aborted by itself after running for 9h 43min. .../457f27c2-3d94-46ce-92b8-f5a42a058701\n Or the whole link to it is:\n<a href=\"https://galois.neptune.ml/#dashboard/job/457f27c2-3d94-46ce-92b8-f5a42a058701\">https://galois.neptune.ml/#dashboard/job/457f27c2-3d94-46ce-92b8-f5a42a058701</a></p>\n\n<p>Then when I ru the same thing again, it ran for <strong>3days 8h</strong> until I ran out of credit. I could only see the save model on the output/project_data/..   directory. Here is the link to that one.</p>\n\n<p>...job/af030724-7366-4919-9a69-5e4f8233253b</p>\n\n<p>About re-running my experiment, I am not sure the best way to go about it since I have only few hours left before the contest ends. I am open to suggestions if you can help.</p>\n\n<p>In fact I would not mind just getting a prediction done on the test data with my saved model at this point.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 257639,
          "author_name": "kkaczmarek",
          "author_url": "",
          "post_date": "12/14/2017 17:38:32",
          "content": "<p>Hi @YaGana,</p>\n\n<p>Thanks, this is what I need: ID is this part that comes after the <code>job/</code> part of the link. I will take a closer look at your job, and get back to you.</p>\n\n<p>Some ideas:</p>\n\n<ol>\n<li><p>Quick workaround would be to download your saved model and run prediction locally. You should be able to make it before contest ends.</p></li>\n<li><p>I will try to grant you some resources to make prediction, however, I cannot make it within next couple of hours.</p></li>\n</ol>\n\n<p>Best,</p>\n\n<p>Kamil</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 257645,
          "author_name": "sheriytm",
          "author_url": "",
          "post_date": "12/14/2017 17:49:45",
          "content": "<p>I have a CPU machine and prediction takes 5.5days on my local machine hence my using a cloud service. My other models are run on FloydHub. I have a prediction running there right now so I cannot use it for my Neptune model in time for submission.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 257761,
          "author_name": "kkaczmarek",
          "author_url": "",
          "post_date": "12/14/2017 22:57:03",
          "content": "<p>YaGana,</p>\n\n<p>I am sorry to hear that.</p>\n\n<p>Best of luck with your final submissions!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 233028,
      "author_name": "yanchao727",
      "author_url": "",
      "post_date": "10/19/2017 02:45:52",
      "content": "<p>There is missing meta data Error:\nUsing TensorFlow backend.\nTraceback (most recent call last):\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/deepsense/neptune/job_wrapper.py\", line 138, in \n    execute()\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/deepsense/neptune/job_wrapper.py\", line 134, in execute\n    execfile(job_filepath, job_globals)\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/past/builtins/misc.py\", line 82, in execfile\n    exec_(code, myglobals, mylocals)\n  File \"experiment_manager.py\", line 273, in \n    registered_actionsargs.action\n  File \"experiment_manager.py\", line 20, in run_pipeline\n    train_valid_split(args)\n  File \"/home/yanchao/cdiscount-starter/utils.py\", line 40, in wrapper\n    function(*args, **kwargs)\n  File \"experiment_manager.py\", line 154, in train_valid_split\n    meta_data = pd.read_csv(meta_data_filepath)\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/pandas/io/parsers.py\", line 655, in parser_f\n    return _read(filepath_or_buffer, kwds)\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/pandas/io/parsers.py\", line 405, in _read\n    parser = TextFileReader(filepath_or_buffer, **kwds)\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/pandas/io/parsers.py\", line 764, in <strong>init</strong>\n    self._make_engine(self.engine)\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/pandas/io/parsers.py\", line 985, in _make_engine\n    self._engine = CParserWrapper(self.f, **self.options)\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/pandas/io/parsers.py\", line 1605, in <strong>init</strong>\n    self._reader = parsers.TextReader(src, **kwds)\n  File \"pandas/_libs/parsers.pyx\", line 394, in pandas._libs.parsers.TextReader.<strong>cinit</strong> (pandas/_libs/parsers.c:4209)\n  File \"pandas/_libs/parsers.pyx\", line 710, in pandas._libs.parsers.TextReader._setup_parser_source (pandas/_libs/parsers.c:8873)\nFileNotFoundError: File b'/public/Cdiscount/meta/meta_train.csv' does not exist</p>",
      "votes": null,
      "replies": [
        {
          "id": 233036,
          "author_name": "yanchao727",
          "author_url": "",
          "post_date": "10/19/2017 02:55:09",
          "content": "<p>My bad , I was running on local machine.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 233174,
      "author_name": "terryli",
      "author_url": "",
      "post_date": "10/19/2017 14:24:14",
      "content": "<p>Hi Jakub, thanks a lot for sharing! May I get some hints on how to resolve the error in the attached screenshot? Thanks!</p>",
      "votes": null,
      "replies": [
        {
          "id": 233280,
          "author_name": "aloisiodn",
          "author_url": "",
          "post_date": "10/19/2017 18:58:10",
          "content": "<p>.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 233281,
          "author_name": "aloisiodn",
          "author_url": "",
          "post_date": "10/19/2017 19:01:16",
          "content": "<p>Have you done this?</p>\n\n<p>from deepsense import neptune</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 233416,
          "author_name": "jakubczakon",
          "author_url": "",
          "post_date": "10/20/2017 06:31:40",
          "content": "<p>You are correct neptune needs to be installed. I merged PR that changed neptune to hard requirement to make it explicit.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 234342,
          "author_name": "kauelm",
          "author_url": "",
          "post_date": "10/23/2017 02:52:46",
          "content": "<p>.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 233346,
      "author_name": "stevenknguyen",
      "author_url": "",
      "post_date": "10/19/2017 22:30:22",
      "content": "<p>Can someone who's run the baseline tell me how long the baseline takes to run? I can't tell if it's stuck or just predicting.</p>",
      "votes": null,
      "replies": [
        {
          "id": 233415,
          "author_name": "jakubczakon",
          "author_url": "",
          "post_date": "10/20/2017 06:29:39",
          "content": "<p>Hi Steven. It takes a while to predict on this test set (I will time it and post the info on that). So I think you  just need to be patient.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 233530,
          "author_name": "stevenknguyen",
          "author_url": "",
          "post_date": "10/20/2017 13:19:52",
          "content": "<p>Thanks Jakub!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 233547,
          "author_name": "aloisiodn",
          "author_url": "",
          "post_date": "10/20/2017 14:02:24",
          "content": "<p>I am not running the full dataset yet. Instead, I am using only images of less frequent classes to tweak LR, Decay and some aspect of my classifier design with neptune grid search. It´s very cool! Trainnig 10 epochs of InceptionV3 with Keras takes only 10 minutes!    </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 233701,
          "author_name": "stevenknguyen",
          "author_url": "",
          "post_date": "10/20/2017 21:43:08",
          "content": "<p>How exactly does neptune grid search work? I'm just starting to play around with the platform and its pretty cool! I like how I don't need to monitor VMs and their prices are very competitive.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 233706,
          "author_name": "jakubczakon",
          "author_url": "",
          "post_date": "10/20/2017 21:58:52",
          "content": "<p>It's nice and simple. \nIt creates a job for each node of the grid search and simply runs them as if they were single experiments. After it justs pools the results.  </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 236174,
          "author_name": "aloisiodn",
          "author_url": "",
          "post_date": "10/26/2017 19:49:52",
          "content": "<p>Hi Steven. Here an submission command example:</p>\n\n<p>neptune send --environment keras-2.0-cpu-py3 --worker gcp-gpu-large --input train.json.tar.gz/train.json -- '--lr %[0.001, 0.002, 0.003] --decay %[0.001, 0.005, 0.009] --n_epochs %[20]'</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 236175,
          "author_name": "aloisiodn",
          "author_url": "",
          "post_date": "10/26/2017 19:53:31",
          "content": "<p>And here a piece of code you should have in your main.py to manage the parameters:</p>\n\n<pre><code>params_parser = argparse.ArgumentParser()\nparams_parser.add_argument('--lr', type=float, default=0.001)\nparams_parser.add_argument('--decay', type=float, default=0.005)\nparams_parser.add_argument('--n_epochs', type=int, default=10)\n\nparams = params_parser.parse_args()\n##################################\nDECAY = params.decay\nLR = params.lr\nN_EPOCHS = params.n_epochs\n</code></pre>",
          "votes": null,
          "replies": []
        },
        {
          "id": 243881,
          "author_name": "neuralnetworks",
          "author_url": "",
          "post_date": "11/15/2017 03:56:29",
          "content": "<p>Did you get a chance to post the time to run?  I can't seem to find it posted in the discussions.  Thanks for any help.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 233418,
      "author_name": "jakubczakon",
      "author_url": "",
      "post_date": "10/20/2017 06:34:36",
      "content": "<p>Some people asked me to be able to contribute to this repo. Of course you can! That is the point,\nSo if you want to contribute in any way just make a pull request, we will review and merge it.</p>",
      "votes": null,
      "replies": [
        {
          "id": 234830,
          "author_name": "maximilianhahn",
          "author_url": "",
          "post_date": "10/24/2017 07:03:29",
          "content": "<p>Thanks a lot for your offer. As I am currently searching a suitable cloud environment for our company, it comes at the perfect time.</p>\n\n<p>However, when I use the CDiscount Script, the job gets queued for a really long time (16 hours right now). Is there something I can do against it?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 235025,
          "author_name": "jakubczakon",
          "author_url": "",
          "post_date": "10/24/2017 16:25:55",
          "content": "<p>Hi Maximilian and sorry for the trouble. There was a lot of interest and we had a temporary problem with the resources. Everything should be working smoothly now.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 235036,
          "author_name": "maximilianhahn",
          "author_url": "",
          "post_date": "10/24/2017 16:42:17",
          "content": "<p>Really no need for excuses here :) Just wanted to know whether I did something wrong. The job starts directly now.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 235066,
          "author_name": "jakubczakon",
          "author_url": "",
          "post_date": "10/24/2017 17:20:09",
          "content": "<p>Happy to hear that!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 233549,
      "author_name": "aloisiodn",
      "author_url": "",
      "post_date": "10/20/2017 14:06:20",
      "content": "<p>InceptionV3, 400 less frequent classes gridsearch in neptune. 10 minutes, 10 epochs.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 235518,
      "author_name": "sinitsyn",
      "author_url": "",
      "post_date": "10/25/2017 16:35:28",
      "content": "<p>Hi!</p>\n\n<p>Can you please explain how can I get meta_ files for local run? \nI tried to look it up in README or your post, but couldn't find anything.</p>",
      "votes": null,
      "replies": [
        {
          "id": 237408,
          "author_name": "jakubczakon",
          "author_url": "",
          "post_date": "10/30/2017 08:56:18",
          "content": "<p>Hi Phillip,\nYou simply need to run </p>\n\n<p><code>\npython run_manager.py create_metadata\n</code></p>\n\n<p>Meta data will be stored in your <code>meta_data_dir</code> specified in the data_config.yaml</p>\n\n<p>Thank you for pointing that out, I will add this to the readme.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 236329,
      "author_name": "bharathyarlagadda",
      "author_url": "",
      "post_date": "10/27/2017 04:39:13",
      "content": "<p>Hi Jakub, \nThank You for the free 100$ credit on neptune. \nI tried to run locally. But I am seeing this issue. \n<code>Started job execution, id: ae4398ce-7e32-4bfa-86f0-eee06e7b1770</code>\n<code>To browse the job, follow:</code>\n<code>https://boole.neptune.ml/#dashboard/job/ae4398ce-7e32-4bfa-86f0-eee06e7b1770?getStartedState=folded</code>\n<code>Traceback (most recent call last):</code>\n  <code>File \"/Users/yarlab/neptune/lib/python2.7/site-packages/deepsense/neptune/job_wrapper.py\", line 138, in</code>\n<code>execute()</code>\n  <code>File \"/Users/yarlab/neptune/lib/python2.7/site-packages/deepsense/neptune/job_wrapper.py\", line 134, in execute</code>\n    execfile(job_filepath, job_globals)<code>\n</code>File \"run_manager.py\", line 252<code>\n</code>config_merged = {**exp_config, **data_config}<code>\n</code>                  ^<code>\n</code>SyntaxError: invalid syntax<code>\n</code>Process exited with return code 1.`</p>",
      "votes": null,
      "replies": [
        {
          "id": 236343,
          "author_name": "bharathyarlagadda",
          "author_url": "",
          "post_date": "10/27/2017 04:47:53",
          "content": "<p>nevermind fixed it. Thank you</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 236584,
      "author_name": "razerx",
      "author_url": "",
      "post_date": "10/27/2017 17:43:59",
      "content": "<p>Command given is same as mentioned in the readme File : source run_neptune_command.sh</p>\n\n<p>Fails with this error :</p>\n\n<p>2017-10-27 17:37:27.984764: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.\n2017-10-27 17:37:27.984773: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.\n2017-10-27 17:37:28.758937: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:893] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n2017-10-27 17:37:28.760243: I tensorflow/core/common_runtime/gpu/gpu_device.cc:940] Found device 0 with properties:\nname: Tesla K80\nmajor: 3 minor: 7 memoryClockRate (GHz) 0.8235\npciBusID 0000:00:04.0\nTotal memory: 11.92GiB\nFree memory: 11.86GiB\n2017-10-27 17:37:28.760275: I tensorflow/core/common_runtime/gpu/gpu_device.cc:961] DMA: 0\n2017-10-27 17:37:28.760283: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0:   Y\n2017-10-27 17:37:28.760292: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Creating TensorFlow device (/gpu:0) -&gt; (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0)\nException in thread Thread-11:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in <strong>next</strong>\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'\nException in thread Thread-8:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in <strong>next</strong>\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'</p>\n\n<p>Exception in thread Thread-9:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in <strong>next</strong>\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'</p>\n\n<p>Exception in thread Thread-10:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in <strong>next</strong>\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'</p>\n\n<p>Traceback (most recent call last):\n  File \"/usr/local/lib/python3.5/dist-packages/deepsense/neptune/job_wrapper.py\", line 138, in \n  File \"/usr/local/lib/python3.5/dist-packages/deepsense/neptune/job_wrapper.py\", line 134, in execute\n  File \"/usr/local/lib/python3.5/dist-packages/past/builtins/misc.py\", line 82, in execfile\n    exec_(code, myglobals, mylocals)\n  File \"run_manager.py\", line 263, in \n    registered_actionsargs.action\n  File \"run_manager.py\", line 21, in run_pipeline\n    train_pipeline(args)\n  File \"run_manager.py\", line 51, in train_pipeline\n    img_dataset_filepath=train_filepath)\n  File \"/neptune/preprocessing.py\", line 50, in fit\n    img_dataset_filepath=img_dataset_filepath)\n  File \"/neptune/postprocessing.py\", line 10, in fit\n    step.fit(X, y, validation_data=validation_data, img_dataset_filepath=img_dataset_filepath)\n  File \"/neptune/pipelines.py\", line 77, in fit\n    self.deep_model.fit(datagens)\n  File \"/neptune/models.py\", line 37, in fit\n    **self.training_cfg)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/legacy/interfaces.py\", line 87, in wrapper\n    return func(*args, **kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/engine/training.py\", line 2011, in fit_generator\n    generator_output = next(output_generator)\nStopIteration</p>",
      "votes": null,
      "replies": [
        {
          "id": 237414,
          "author_name": "jakubczakon",
          "author_url": "",
          "post_date": "10/30/2017 09:16:03",
          "content": "<p>It is fixed now.  Sorry for the trouble.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 236739,
      "author_name": "randombishop",
      "author_url": "",
      "post_date": "10/28/2017 00:51:11",
      "content": "<p>Thanks for the promo, I started playing with neptune platform and it's fun !\nBut can I only run jobs on single GPU Tesla 80K ?\nIs there a way to run a job on multiple faster GPUs ?\nCheers !</p>",
      "votes": null,
      "replies": [
        {
          "id": 236746,
          "author_name": "stevenknguyen",
          "author_url": "",
          "post_date": "10/28/2017 01:39:36",
          "content": "<p>It looks like the neptune platform is built on top of GCP with clusters for either 4 or 8 k80s using the argument <code>--worker gcp-gpu-medium</code> or <code>--worker gcp-gpu-large</code>. In my experience with GCP you can use it like a single GPU with expected diminishing returns.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 237455,
          "author_name": "jakubczakon",
          "author_url": "",
          "post_date": "10/30/2017 10:56:30",
          "content": "<p>Unfortunately at the moment we don't have workers that support multi-gpu and <code>gcp-gpu-large</code> is the strongest machine. That will likely change in the future so keep your finger on the pulse.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 237960,
      "author_name": "kkaczmarek",
      "author_url": "",
      "post_date": "10/31/2017 11:44:51",
      "content": "<p>Hi, check our <a href=\"https://blog.deepsense.ai/image-classification-sample-solution-kaggle/\">blog post featuring starter code</a> prepared by Jakub. Post has short video, where Jakub gives quick code overview.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 242691,
      "author_name": "skinish",
      "author_url": "",
      "post_date": "11/12/2017 11:50:56",
      "content": "<p>Hey! Thanks for the share! Does this include storage?</p>",
      "votes": null,
      "replies": [
        {
          "id": 242849,
          "author_name": "jakubczakon",
          "author_url": "",
          "post_date": "11/12/2017 19:27:26",
          "content": "<p>Hi Sinish. The data is mounted in neptune so you can start experimenting right of the bat.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 245861,
      "author_name": "ppujari",
      "author_url": "",
      "post_date": "11/19/2017 22:08:23",
      "content": "<p>I run the code using neptune as per instructions, it gets stuck after 10 epochs, I do not see any file in the output folder. Please help. </p>",
      "votes": null,
      "replies": [
        {
          "id": 246689,
          "author_name": "neuralnetworks",
          "author_url": "",
          "post_date": "11/21/2017 16:15:50",
          "content": "<p>I am having the same issue.  Any help would be much appreciated.  Thanks!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 247531,
          "author_name": "ymcdull",
          "author_url": "",
          "post_date": "11/23/2017 09:49:33",
          "content": "<p>Having the same issue here, anyone got a solution for this? Thanks.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 247837,
          "author_name": "sheriytm",
          "author_url": "",
          "post_date": "11/24/2017 03:50:08",
          "content": "<p>I have the same issue and it has been running for 3h 42min. On the dash board for the job I got to by copy pasting the link given after running the command \"source run_neptune_command.sh\", it shows an orange triangle at the top left corner just below \"starter_100_100, Running for 4 hours\".</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 257614,
          "author_name": "kkaczmarek",
          "author_url": "",
          "post_date": "12/14/2017 16:25:55",
          "content": "<p>Hi @Pradeep, @Chris, @Yunfeng, @YaGana,</p>\n\n<p>We need your experiment ID in order to check what caused the job to get stuck after some training time. Also, please double check <code>output</code> directory, since temporary connection issues do not cause an experiment to crash.</p>\n\n<p>Please feel free to post your experiment ID.</p>\n\n<p>Cheers,</p>\n\n<p>Kamil (on behalf of @Jakub)</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 257499,
      "author_name": "psardin",
      "author_url": "",
      "post_date": "12/14/2017 11:08:20",
      "content": "<p>Hello Jakub,</p>\n\n<p>I have an error with experience_config.yaml </p>\n\n<pre><code>The provided job configuration /cdiscount-starter-master/experiment_config.yaml is invalid! Validation errors: 1. Value '100' is not of type 'str'. Path: '/properties/0/value', 2. Value '100' is not of type 'str'. Path: '/properties/1/value', 3. Value '10' is not of type 'str'. Path: '/properties/2/value'\n</code></pre>\n\n<p>If i changed the type to string, i can send the process but i have another error on neptune.ml</p>\n\n<pre><code>75.934818   TypeError: cannot do slice indexing on &lt;class 'pandas.core.indexes.range.RangeIndex'&gt; with these indexers [100] of &lt;class 'str'&gt;\n</code></pre>\n\n<p>Im using neptune-cli (2.4.3)</p>",
      "votes": null,
      "replies": [
        {
          "id": 257630,
          "author_name": "kkaczmarek",
          "author_url": "",
          "post_date": "12/14/2017 17:11:27",
          "content": "<p>Hi @Pierre,</p>\n\n<p>Thanks for sharing this with us. I will take a closer look your error and I will contact you with the update.</p>\n\n<p>Best,</p>\n\n<p>Kamil (on behalf of Jakub)</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 257760,
          "author_name": "kkaczmarek",
          "author_url": "",
          "post_date": "12/14/2017 22:55:30",
          "content": "<p>Hi @Pierre,</p>\n\n<p>Your issue is now solved. You can pull new version from the <a href=\"https://github.com/deepsense-ai/cdiscount-starter\">Cdiscount-starter</a> repository.</p>\n\n<p>Best,</p>\n\n<p>K</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 956849,
      "author_name": "batuhan35",
      "author_url": "",
      "post_date": "08/03/2020 20:45:15",
      "content": "<p>Hello Guys. I choose ResNet for İmage Classification. I thınk you should check thıs code\n<a href=\"https://github.com/batuhan3526/ResNet50_on_Cifar_100_Without_Transfer_Learning\">https://github.com/batuhan3526/ResNet50_on_Cifar_100_Without_Transfer_Learning</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "232854": "Hi everybody!\nAt deepsense.ai we love competitions and this time we really want to spread our love with the community.\n\nWe have prepared a nice end-to-end image classification pipeline that is easily extendable and tweakable. \nCheck the code here:\nhttps://github.com/deepsense-ai/cdiscount-starter\nIt gets your from raw data to submission but it's highly modular so you can build your solution upon that.\n\nAs you well know gpu is going to be burned left and right during this competition.\nSo we have decided that we will give the community an opportunity to create and train amazing models for free!\nThat is correct. \nSign up to https://neptune.ml and get $100 to train your models. You can get a really nice submission with that so use it wisely.\n\nYou are welcome and good luck!",
    "232864": "Hi Jakub, deepsense.ai is awesome! For sure I will take a look at it. Thanks!",
    "232898": "Jakub, do you believe the free account can handle this competition datasets?",
    "232976": "We mounted the dataset for you (/public/Cdiscount) so you can go ahead and start training right away!",
    "232998": "Thanks Jakub. I am trying to run some experiments and i find module bson from pymongo is missing...",
    "233028": "There is missing meta data Error:\nUsing TensorFlow backend.\nTraceback (most recent call last):\n  File \"/home/yanchao/anaconda3/lib/python3.6/site-packages/deepsense/neptune/job_wrapper.py\", line 138, in",
    "233036": "My bad , I was running on local machine.",
    "233080": "Hi Aloisio,\npymongo is installed on neptune just before running it and I have just confirmed on another machine that cloning the repo and running `source run_neptune_command.sh` works just fine.\nAre you running it with neptune run or neptune send? What is your os?\nIf you are running neptune run then you should first install all the requirements locally.",
    "233161": "I am running with neptune send. Its funny because when I commented the BSON part of the code and the module import it runs OK.",
    "233163": "I will try again later. Thanks !",
    "233174": "Hi Jakub, thanks a lot for sharing! May I get some hints on how to resolve the error in the attached screenshot? Thanks!",
    "233280": ".",
    "233281": "Have you done this?\n\nfrom deepsense import neptune",
    "233346": "Can someone who's run the baseline tell me how long the baseline takes to run? I can't tell if it's stuck or just predicting.",
    "233415": "Hi Steven. It takes a while to predict on this test set (I will time it and post the info on that). So I think you  just need to be patient.",
    "233416": "You are correct neptune needs to be installed. I merged PR that changed neptune to hard requirement to make it explicit.",
    "233418": "Some people asked me to be able to contribute to this repo. Of course you can! That is the point,\nSo if you want to contribute in any way just make a pull request, we will review and merge it.",
    "233530": "Thanks Jakub!",
    "233547": "I am not running the full dataset yet. Instead, I am using only images of less frequent classes to tweak LR, Decay and some aspect of my classifier design with neptune grid search. It´s very cool! Trainnig 10 epochs of InceptionV3 with Keras takes only 10 minutes!",
    "233549": "InceptionV3, 400 less frequent classes gridsearch in neptune. 10 minutes, 10 epochs.",
    "233701": "How exactly does neptune grid search work? I'm just starting to play around with the platform and its pretty cool! I like how I don't need to monitor VMs and their prices are very competitive.",
    "233706": "It's nice and simple. \nIt creates a job for each node of the grid search and simply runs them as if they were single experiments. After it justs pools the results.",
    "234342": ".",
    "234830": "Thanks a lot for your offer. As I am currently searching a suitable cloud environment for our company, it comes at the perfect time.\n\nHowever, when I use the CDiscount Script, the job gets queued for a really long time (16 hours right now). Is there something I can do against it?",
    "234993": "Did you find a solution? I am having the same issue.",
    "235025": "Hi Maximilian and sorry for the trouble. There was a lot of interest and we had a temporary problem with the resources. Everything should be working smoothly now.",
    "235026": "We made some improvements to the starter code. I would reccomend that you download the new version and try with it.",
    "235036": "Really no need for excuses here :) Just wanted to know whether I did something wrong. The job starts directly now.",
    "235066": "Happy to hear that!",
    "235518": "Hi!\n\nCan you please explain how can I get meta_ files for local run? \nI tried to look it up in README or your post, but couldn't find anything.",
    "236174": "Hi Steven. Here an submission command example:\n\nneptune send --environment keras-2.0-cpu-py3 --worker gcp-gpu-large --input train.json.tar.gz/train.json -- '--lr %[0.001, 0.002, 0.003] --decay %[0.001, 0.005, 0.009] --n_epochs %[20]'",
    "236175": "And here a piece of code you should have in your main.py to manage the parameters:\n\n    params_parser = argparse.ArgumentParser()\n    params_parser.add_argument('--lr', type=float, default=0.001)\n    params_parser.add_argument('--decay', type=float, default=0.005)\n    params_parser.add_argument('--n_epochs', type=int, default=10)\n    \n    params = params_parser.parse_args()\n    ##################################\n    DECAY = params.decay\n    LR = params.lr\n    N_EPOCHS = params.n_epochs",
    "236329": "Hi Jakub, \nThank You for the free 100$ credit on neptune. \nI tried to run locally. But I am seeing this issue. \n`Started job execution, id: ae4398ce-7e32-4bfa-86f0-eee06e7b1770`\n`To browse the job, follow:`\n`https://boole.neptune.ml/#dashboard/job/ae4398ce-7e32-4bfa-86f0-eee06e7b1770?getStartedState=folded`\n`Traceback (most recent call last):`\n  `File \"/Users/yarlab/neptune/lib/python2.7/site-packages/deepsense/neptune/job_wrapper.py\", line 138, in `\n`execute()`\n  `File \"/Users/yarlab/neptune/lib/python2.7/site-packages/deepsense/neptune/job_wrapper.py\", line 134, in execute`\n    execfile(job_filepath, job_globals)`\n  `File \"run_manager.py\", line 252`\n    `config_merged = {**exp_config, **data_config}`\n    `                  ^`\n`SyntaxError: invalid syntax`\n`Process exited with return code 1.`",
    "236343": "nevermind fixed it. Thank you",
    "236584": "Command given is same as mentioned in the readme File : source run_neptune_command.sh\n\nFails with this error :\n\n2017-10-27 17:37:27.984764: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.\n2017-10-27 17:37:27.984773: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.\n2017-10-27 17:37:28.758937: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:893] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n2017-10-27 17:37:28.760243: I tensorflow/core/common_runtime/gpu/gpu_device.cc:940] Found device 0 with properties:\nname: Tesla K80\nmajor: 3 minor: 7 memoryClockRate (GHz) 0.8235\npciBusID 0000:00:04.0\nTotal memory: 11.92GiB\nFree memory: 11.86GiB\n2017-10-27 17:37:28.760275: I tensorflow/core/common_runtime/gpu/gpu_device.cc:961] DMA: 0\n2017-10-27 17:37:28.760283: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0:   Y\n2017-10-27 17:37:28.760292: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Creating TensorFlow device (/gpu:0) -&gt; (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0)\nException in thread Thread-11:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in __next__\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'\nException in thread Thread-8:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in __next__\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'\n\n\nException in thread Thread-9:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in __next__\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'\n\nException in thread Thread-10:\nTraceback (most recent call last):\n  File \"/usr/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n    self.run()\n  File \"/usr/lib/python3.5/threading.py\", line 862, in run\n    self._target(*self._args, **self._kwargs)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/utils/data_utils.py\", line 568, in data_generator_task\n    generator_output = next(self._generator)\n  File \"/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py\", line 737, in __next__\n    return self.next(*args, **kwargs)\n  File \"/neptune/preprocessing.py\", line 182, in next\n    return self._get_batches_of_transformed_samples(index_array)\n  File \"/neptune/preprocessing.py\", line 156, in _get_batches_of_transformed_samples\n    batch_y = to_categorical(batch_y_id, num_classes=self.num_classes)\nAttributeError: 'bsonIterator' object has no attribute 'num_classes'\n\nTraceback (most recent call last):\n  File \"/usr/local/lib/python3.5/dist-packages/deepsense/neptune/job_wrapper.py\", line 138, in",
    "236739": "Thanks for the promo, I started playing with neptune platform and it's fun !\nBut can I only run jobs on single GPU Tesla 80K ?\nIs there a way to run a job on multiple faster GPUs ?\nCheers !",
    "236746": "It looks like the neptune platform is built on top of GCP with clusters for either 4 or 8 k80s using the argument `--worker gcp-gpu-medium` or `--worker gcp-gpu-large`. In my experience with GCP you can use it like a single GPU with expected diminishing returns.",
    "237408": "Hi Phillip,\nYou simply need to run \n\n```\npython run_manager.py create_metadata\n```\n\nMeta data will be stored in your `meta_data_dir` specified in the data_config.yaml\n\nThank you for pointing that out, I will add this to the readme.",
    "237414": "It is fixed now.  Sorry for the trouble.",
    "237455": "Unfortunately at the moment we don't have workers that support multi-gpu and `gcp-gpu-large` is the strongest machine. That will likely change in the future so keep your finger on the pulse.",
    "237960": "Hi, check our [blog post featuring starter code][1] prepared by Jakub. Post has short video, where Jakub gives quick code overview.\n\n\n  [1]: https://blog.deepsense.ai/image-classification-sample-solution-kaggle/",
    "242691": "Hey! Thanks for the share! Does this include storage?",
    "242849": "Hi Sinish. The data is mounted in neptune so you can start experimenting right of the bat.",
    "243881": "Did you get a chance to post the time to run?  I can't seem to find it posted in the discussions.  Thanks for any help.",
    "245861": "I run the code using neptune as per instructions, it gets stuck after 10 epochs, I do not see any file in the output folder. Please help.",
    "246689": "I am having the same issue.  Any help would be much appreciated.  Thanks!",
    "246711": "How would we access the data in Neptune using a Jupyter Notebook?  Would the \"path to inputs visible to notebook\" = \"/public/Cdiscount/test.bson\" for the test.bson file?  Thanks for any help!",
    "247531": "Having the same issue here, anyone got a solution for this? Thanks.",
    "247837": "I have the same issue and it has been running for 3h 42min. On the dash board for the job I got to by copy pasting the link given after running the command \"source run_neptune_command.sh\", it shows an orange triangle at the top left corner just below \"starter_100_100, Running for 4 hours\".",
    "249693": "No improvement, I run source run_neptune_command.sh again. Same it hangs after 2 hours.",
    "256418": "Hi Pradeep. I have just tried to reproduce your problem but after 3 hours the 100 classes 100 images starter has succeeded getting 0.21 on the LB. The submission is in the output/project_data/submissions folder.",
    "256907": "Jakub Czakon, the point is running the starter code on the neptune.ml system keeps giving lost connection error so many times, it eats up so much valuable time and money. In the end mine did run but could not finish before I ran out of the $100. Going by my short experience with it, I felt it would be a waste of money to subscribe. I still feel bad that I wasted so much time trying it out. Plus if you check the post on the very top here by @Pradeep, 3 other people including me complained over 20 days ago and you did not respond until today. To me, that just does not seem right.",
    "256913": "You are right YaGana I should have answered sooner and I really apologize for that. In the future, I will make sure to attend any problems swiftly.",
    "257499": "Hello Jakub,\n\nI have an error with experience_config.yaml \n\n    The provided job configuration /cdiscount-starter-master/experiment_config.yaml is invalid! Validation errors: 1. Value '100' is not of type 'str'. Path: '/properties/0/value', 2. Value '100' is not of type 'str'. Path: '/properties/1/value', 3. Value '10' is not of type 'str'. Path: '/properties/2/value'\n\nIf i changed the type to string, i can send the process but i have another error on neptune.ml\n\n    75.934818\tTypeError: cannot do slice indexing on",
    "257609": "Hi Chris,\n\nYou may want to start Notebook directly from the [Neptune](https://neptune.ml/). Simply log in via web, then at the top bar you can see link \"Start Notebook\".\n\nFrom inside the notebook, you can type: `!ls -la /public/Cdiscount/` to see Cdiscount data (read only).\n\nLet me mention that bson is not pre-installed, however, you can install it from inside the notebook, via:\n`!pip install your_package`.",
    "257612": "Hi YaGana,\n\nThanks for pointing this out.\n\nCan you post your experiment ID, so that I can check what is going on with your job, especially your 'lost connection error'.\n\nYou also may want to re-run your experiment. In this case let me know and we will discuss best way to do it.\n\nBest,\n\nKamil (who works with Jakub ;-) )",
    "257613": "Hi @Pradeep,\n\nPlease post you experiment ID / link to the experiment, so that I can check what is going on.\n\nAlso, make sure that you are using the latest version of the starter code.\n\nBest,\n\nKamil (on behalf of @Jakub)",
    "257614": "Hi @Pradeep, @Chris, @Yunfeng, @YaGana,\n\nWe need your experiment ID in order to check what caused the job to get stuck after some training time. Also, please double check `output` directory, since temporary connection issues do not cause an experiment to crash.\n\nPlease feel free to post your experiment ID.\n\nCheers,\n\nKamil (on behalf of @Jakub)",
    "257622": "kamil, I could not see any id but here is the end of the line of the link from my experiment that aborted by itself after running for 9h 43min. .../457f27c2-3d94-46ce-92b8-f5a42a058701\n Or the whole link to it is:\nhttps://galois.neptune.ml/#dashboard/job/457f27c2-3d94-46ce-92b8-f5a42a058701\n\nThen when I ru the same thing again, it ran for **3days 8h** until I ran out of credit. I could only see the save model on the output/project_data/..   directory. Here is the link to that one.\n\n...job/af030724-7366-4919-9a69-5e4f8233253b\n\nAbout re-running my experiment, I am not sure the best way to go about it since I have only few hours left before the contest ends. I am open to suggestions if you can help.\n\nIn fact I would not mind just getting a prediction done on the test data with my saved model at this point.",
    "257630": "Hi @Pierre,\n\nThanks for sharing this with us. I will take a closer look your error and I will contact you with the update.\n\nBest,\n\nKamil (on behalf of Jakub)",
    "257639": "Hi @YaGana,\n\nThanks, this is what I need: ID is this part that comes after the `job/` part of the link. I will take a closer look at your job, and get back to you.\n\nSome ideas:\n\n1.  Quick workaround would be to download your saved model and run prediction locally. You should be able to make it before contest ends.\n\n2. I will try to grant you some resources to make prediction, however, I cannot make it within next couple of hours.\n\nBest,\n\nKamil",
    "257645": "I have a CPU machine and prediction takes 5.5days on my local machine hence my using a cloud service. My other models are run on FloydHub. I have a prediction running there right now so I cannot use it for my Neptune model in time for submission.",
    "257760": "Hi @Pierre,\n\nYour issue is now solved. You can pull new version from the [Cdiscount-starter](https://github.com/deepsense-ai/cdiscount-starter) repository.\n\nBest,\n\nK",
    "257761": "YaGana,\n\nI am sorry to hear that.\n\nBest of luck with your final submissions!",
    "956849": "Hello Guys. I choose ResNet for İmage Classification. I thınk you should check thıs code\nhttps://github.com/batuhan3526/ResNet50_on_Cifar_100_Without_Transfer_Learning"
  },
  "source": "meta"
}