Details here https://www.kaggle.com/getting-started/47096#post271139 It is a Kaggle kernel type environment with 2 cores of Xeon, a single K80 GPU, 13 GB RAM. Programs can run upto 12 hours. Tested it and is true. May be helpful in this competition for people without GPUs.
", "messages": [ { "id": "271317", "postDate": "01/20/2018 06:23:23", "content": "Details here https://www.kaggle.com/getting-started/47096#post271139 It is a Kaggle kernel type environment with 2 cores of Xeon, a single K80 GPU, 13 GB RAM. Programs can run upto 12 hours. Tested it and is true. May be helpful in this competition for people without GPUs.
", "rawMarkdown": "Details here https://www.kaggle.com/getting-started/47096#post271139 It is a Kaggle kernel type environment with 2 cores of Xeon, a single K80 GPU, 13 GB RAM. Programs can run upto 12 hours. Tested it and is true. May be helpful in this competition for people without GPUs.", "votes": null }, { "id": "271500", "postDate": "01/20/2018 15:47:28", "content": "Hello folks - Onepanel is too without any limitations on time.
\n\nI rebuilt the Keras Simple CNN Starter by CVxTz ( https://www.kaggle.com/CVxTz/keras-simple-cnn-starter ) kernel on Onepanel and running on free GPUs. Follow the README.md file on how to clone the fully running env. I will start trying Jobs with incremental experiments in parallel. Looking to build a team on this platform with access to free GPUs... The dataset can be found and mounted within the environment reducing the time between experiments.
\n\nOnepanel is offering to join / support teams by providing parallel GPU optimized job execution engine access on the platform. Just have to provide feedback, bug reports, and feature requests as the platform is still in beta...
\n\n\n\nDatasets can be found and mounted in your notebook here:
\n\n\n\nSORRY BETA IS NOW CLOSED
", "rawMarkdown": "Hello folks - Onepanel is too without any limitations on time. \n\nI rebuilt the Keras Simple CNN Starter by CVxTz ( https://www.kaggle.com/CVxTz/keras-simple-cnn-starter ) kernel on Onepanel and running on free GPUs. Follow the README.md file on how to clone the fully running env. I will start trying Jobs with incremental experiments in parallel. Looking to build a team on this platform with access to free GPUs... The dataset can be found and mounted within the environment reducing the time between experiments.\n\nOnepanel is offering to join / support teams by providing parallel GPU optimized job execution engine access on the platform. Just have to provide feedback, bug reports, and feature requests as the platform is still in beta...\n\nhttps://www.onepanel.io/beta\n\nDatasets can be found and mounted in your notebook here:\n\nhttps://www.onepanel.io/beta\n\nSORRY BETA IS NOW CLOSED", "votes": null }, { "id": "273647", "postDate": "01/24/2018 23:03:49", "content": "Thanks so much for this, was just looking around competitions and stumbled upon this thread
", "rawMarkdown": "Thanks so much for this, was just looking around competitions and stumbled upon this thread", "votes": null }, { "id": "273891", "postDate": "01/25/2018 12:37:09", "content": "Hi Abhishek,
\n\nHow did you import Kaggle's data in the notebook?
", "rawMarkdown": "Hi Abhishek,\n\nHow did you import Kaggle's data in the notebook?", "votes": null }, { "id": "273896", "postDate": "01/25/2018 12:49:14", "content": "I guess via Gdrive. I'm also struggling for the same.
", "rawMarkdown": "I guess via Gdrive. I'm also struggling for the same.", "votes": null }, { "id": "274164", "postDate": "01/26/2018 00:53:17", "content": "You can use kaggle-cli to load the data directly from Kaggle into Colab:
\n\n# Install\n!pip install kaggle-cli\n\n# Fill in your Kaggle credentials\nkaggle_id=''\nkaggle_password=''\nkaggle_challenge='sp-society-camera-model-identification'\n\n# Get an authentication token from Kaggle. It will be saved in a file.\n!kg config -g -u {kaggle_id} -p {kaggle_password} -c {kaggle_challenge}\n\n# Download the zips from Kaggle\n!kg download\n\n# Unzip\n!unzip train.zip\n!unzip test.zip\n!unzip sample_submission.csv.zip\n\n# Clean up\n!rm train.zip\n!rm test.zip\n!rm sample_submission.csv.zip\nSince Colab instances are experimental, be sure to back up your work and data. I use Google Cloud Storage because it's fast and cheap.
\n\n# Fill in your project ID and bucket name\nkaggle_project_id = ''\nkaggle_bucket_name = ''\n\n# Get authentication token from Google\nfrom google.colab import auth\nauth.authenticate_user()\n!gcloud config set project {kaggle_project_id}\n\n# Back up files to GCS\n!gsutil -m cp -r test gs://{kaggle_bucket_name}\n!gsutil -m cp -r train gs://{kaggle_bucket_name}\n\n# Restore files from GCS\n!gsutil -m cp gs://{kaggle_bucket_name}/train . \n!gsutil -m cp gs://{kaggle_bucket_name}/test .\nRather than restoring from GCS, you can also access your data by API calls. That's what I do, mostly for esthetic reasons. Under the covers, there is no file system. API calls are the ground truth, so why pretend?
", "rawMarkdown": "You can use kaggle-cli to load the data directly from Kaggle into Colab:\n\n # Install\n !pip install kaggle-cli\n\n # Fill in your Kaggle credentials\n kaggle_id=''\n kaggle_password=''\n kaggle_challenge='sp-society-camera-model-identification'\n\n # Get an authentication token from Kaggle. It will be saved in a file.\n !kg config -g -u {kaggle_id} -p {kaggle_password} -c {kaggle_challenge}\n\n # Download the zips from Kaggle\n !kg download\n\n # Unzip\n !unzip train.zip\n !unzip test.zip\n !unzip sample_submission.csv.zip\n\n # Clean up\n !rm train.zip\n !rm test.zip\n !rm sample_submission.csv.zip\n \nSince Colab instances are experimental, be sure to back up your work and data. I use Google Cloud Storage because it's fast and cheap.\n\n # Fill in your project ID and bucket name\n kaggle_project_id = ''\n kaggle_bucket_name = ''\n \n # Get authentication token from Google\n from google.colab import auth\n auth.authenticate_user()\n !gcloud config set project {kaggle_project_id}\n \n # Back up files to GCS\n !gsutil -m cp -r test gs://{kaggle_bucket_name}\n !gsutil -m cp -r train gs://{kaggle_bucket_name}\n \n # Restore files from GCS\n !gsutil -m cp gs://{kaggle_bucket_name}/train . \n !gsutil -m cp gs://{kaggle_bucket_name}/test .\n\nRather than restoring from GCS, you can also access your data by API calls. That's what I do, mostly for esthetic reasons. Under the covers, _there is no file system_. API calls are the ground truth, so why pretend?", "votes": null } ], "comments": [ { "id": 271500, "author_name": "onepanel", "author_url": "", "post_date": "01/20/2018 15:47:28", "content": "Hello folks - Onepanel is too without any limitations on time.
\n\nI rebuilt the Keras Simple CNN Starter by CVxTz ( https://www.kaggle.com/CVxTz/keras-simple-cnn-starter ) kernel on Onepanel and running on free GPUs. Follow the README.md file on how to clone the fully running env. I will start trying Jobs with incremental experiments in parallel. Looking to build a team on this platform with access to free GPUs... The dataset can be found and mounted within the environment reducing the time between experiments.
\n\nOnepanel is offering to join / support teams by providing parallel GPU optimized job execution engine access on the platform. Just have to provide feedback, bug reports, and feature requests as the platform is still in beta...
\n\n\n\nDatasets can be found and mounted in your notebook here:
\n\n\n\nSORRY BETA IS NOW CLOSED
", "votes": null, "replies": [] }, { "id": 273647, "author_name": "livingprogram", "author_url": "", "post_date": "01/24/2018 23:03:49", "content": "Thanks so much for this, was just looking around competitions and stumbled upon this thread
", "votes": null, "replies": [] }, { "id": 273891, "author_name": "igormunizims", "author_url": "", "post_date": "01/25/2018 12:37:09", "content": "Hi Abhishek,
\n\nHow did you import Kaggle's data in the notebook?
", "votes": null, "replies": [ { "id": 273896, "author_name": "nuhsikander", "author_url": "", "post_date": "01/25/2018 12:49:14", "content": "I guess via Gdrive. I'm also struggling for the same.
", "votes": null, "replies": [] } ] }, { "id": 274164, "author_name": "versalog", "author_url": "", "post_date": "01/26/2018 00:53:17", "content": "You can use kaggle-cli to load the data directly from Kaggle into Colab:
\n\n# Install\n!pip install kaggle-cli\n\n# Fill in your Kaggle credentials\nkaggle_id=''\nkaggle_password=''\nkaggle_challenge='sp-society-camera-model-identification'\n\n# Get an authentication token from Kaggle. It will be saved in a file.\n!kg config -g -u {kaggle_id} -p {kaggle_password} -c {kaggle_challenge}\n\n# Download the zips from Kaggle\n!kg download\n\n# Unzip\n!unzip train.zip\n!unzip test.zip\n!unzip sample_submission.csv.zip\n\n# Clean up\n!rm train.zip\n!rm test.zip\n!rm sample_submission.csv.zip\nSince Colab instances are experimental, be sure to back up your work and data. I use Google Cloud Storage because it's fast and cheap.
\n\n# Fill in your project ID and bucket name\nkaggle_project_id = ''\nkaggle_bucket_name = ''\n\n# Get authentication token from Google\nfrom google.colab import auth\nauth.authenticate_user()\n!gcloud config set project {kaggle_project_id}\n\n# Back up files to GCS\n!gsutil -m cp -r test gs://{kaggle_bucket_name}\n!gsutil -m cp -r train gs://{kaggle_bucket_name}\n\n# Restore files from GCS\n!gsutil -m cp gs://{kaggle_bucket_name}/train . \n!gsutil -m cp gs://{kaggle_bucket_name}/test .\nRather than restoring from GCS, you can also access your data by API calls. That's what I do, mostly for esthetic reasons. Under the covers, there is no file system. API calls are the ground truth, so why pretend?
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "271317": "Details here https://www.kaggle.com/getting-started/47096#post271139 It is a Kaggle kernel type environment with 2 cores of Xeon, a single K80 GPU, 13 GB RAM. Programs can run upto 12 hours. Tested it and is true. May be helpful in this competition for people without GPUs.", "271500": "Hello folks - Onepanel is too without any limitations on time. \n\nI rebuilt the Keras Simple CNN Starter by CVxTz ( https://www.kaggle.com/CVxTz/keras-simple-cnn-starter ) kernel on Onepanel and running on free GPUs. Follow the README.md file on how to clone the fully running env. I will start trying Jobs with incremental experiments in parallel. Looking to build a team on this platform with access to free GPUs... The dataset can be found and mounted within the environment reducing the time between experiments.\n\nOnepanel is offering to join / support teams by providing parallel GPU optimized job execution engine access on the platform. Just have to provide feedback, bug reports, and feature requests as the platform is still in beta...\n\nhttps://www.onepanel.io/beta\n\nDatasets can be found and mounted in your notebook here:\n\nhttps://www.onepanel.io/beta\n\nSORRY BETA IS NOW CLOSED", "273647": "Thanks so much for this, was just looking around competitions and stumbled upon this thread", "273891": "Hi Abhishek,\n\nHow did you import Kaggle's data in the notebook?", "273896": "I guess via Gdrive. I'm also struggling for the same.", "274164": "You can use kaggle-cli to load the data directly from Kaggle into Colab:\n\n # Install\n !pip install kaggle-cli\n\n # Fill in your Kaggle credentials\n kaggle_id=''\n kaggle_password=''\n kaggle_challenge='sp-society-camera-model-identification'\n\n # Get an authentication token from Kaggle. It will be saved in a file.\n !kg config -g -u {kaggle_id} -p {kaggle_password} -c {kaggle_challenge}\n\n # Download the zips from Kaggle\n !kg download\n\n # Unzip\n !unzip train.zip\n !unzip test.zip\n !unzip sample_submission.csv.zip\n\n # Clean up\n !rm train.zip\n !rm test.zip\n !rm sample_submission.csv.zip\n \nSince Colab instances are experimental, be sure to back up your work and data. I use Google Cloud Storage because it's fast and cheap.\n\n # Fill in your project ID and bucket name\n kaggle_project_id = ''\n kaggle_bucket_name = ''\n \n # Get authentication token from Google\n from google.colab import auth\n auth.authenticate_user()\n !gcloud config set project {kaggle_project_id}\n \n # Back up files to GCS\n !gsutil -m cp -r test gs://{kaggle_bucket_name}\n !gsutil -m cp -r train gs://{kaggle_bucket_name}\n \n # Restore files from GCS\n !gsutil -m cp gs://{kaggle_bucket_name}/train . \n !gsutil -m cp gs://{kaggle_bucket_name}/test .\n\nRather than restoring from GCS, you can also access your data by API calls. That's what I do, mostly for esthetic reasons. Under the covers, _there is no file system_. API calls are the ground truth, so why pretend?" }, "source": "meta" }