![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2920583%2F08b10bbec1ec869f00e4c9c4bf67308f%2FScreenshot_3.png?generation=1566662210200322&alt=media\")
There's another problem, I bet, for instance,
\n\n!conda install -c rdkit rdkit -y
it comes from a notebook cell, it's fine.
\n\nAs soon as the session fades or the context is reset, you have to put it anew - it's wildly annoying. It is said that it is possible to put a new image through the gui interface or through pip directly, and it is supposedly possible to choose a new image in the list with Docker - but it has never been possible to install it through the interface yet. Installation via pip rarely gives a drop-down list with new images.\n
Questions:\nWhat am I doing wrong?\nI want to get such a drop-down list of containers all the time, and put it with conda, how?\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2920583%2Fcaa6cdf39447a26565978e77b38e6bfa%2FScreenshot_4.png?generation=1566662343855081&alt=media\")
Is there any way to update my docker image using conda?\nHow do I make my own copy of the poker image for my notebook computer so that new images appear? I'll be glad to have tutorials or explanations.\nNow the problem is that the utility script does not commit because there is not enough lib, but to put it without a laptop (in the script itself something does not seem like it). It remains a variant to make in the laptop the docker-image and to use it for a script, but the new image does not appear
", "messages": [ { "id": "607101", "postDate": "08/24/2019 15:59:07", "content": "There's another problem, I bet, for instance,
\n\n!conda install -c rdkit rdkit -y
it comes from a notebook cell, it's fine.
\n\nAs soon as the session fades or the context is reset, you have to put it anew - it's wildly annoying. It is said that it is possible to put a new image through the gui interface or through pip directly, and it is supposedly possible to choose a new image in the list with Docker - but it has never been possible to install it through the interface yet. Installation via pip rarely gives a drop-down list with new images.\n
Questions:\nWhat am I doing wrong?\nI want to get such a drop-down list of containers all the time, and put it with conda, how?\n
Is there any way to update my docker image using conda?\nHow do I make my own copy of the poker image for my notebook computer so that new images appear? I'll be glad to have tutorials or explanations.\nNow the problem is that the utility script does not commit because there is not enough lib, but to put it without a laptop (in the script itself something does not seem like it). It remains a variant to make in the laptop the docker-image and to use it for a script, but the new image does not appear
", "rawMarkdown": "There's another problem, I bet, for instance,\n\n`!conda install -c rdkit rdkit -y`\n\nit comes from a notebook cell, it's fine. \n\nAs soon as the session fades or the context is reset, you have to put it anew - it's wildly annoying. It is said that it is possible to put a new image through the gui interface or through pip directly, and it is supposedly possible to choose a new image in the list with Docker - but it has never been possible to install it through the interface yet. Installation via pip rarely gives a drop-down list with new images.\n\n\nQuestions:\nWhat am I doing wrong?\nI want to get such a drop-down list of containers all the time, and put it with conda, how?\n\n\nIs there any way to update my docker image using conda?\nHow do I make my own copy of the poker image for my notebook computer so that new images appear? I'll be glad to have tutorials or explanations.\nNow the problem is that the utility script does not commit because there is not enough lib, but to put it without a laptop (in the script itself something does not seem like it). It remains a variant to make in the laptop the docker-image and to use it for a script, but the new image does not appear", "votes": null }, { "id": "607125", "postDate": "08/24/2019 16:32:29", "content": "I'm afraid you mix docker images and docker containers. A docker image is typically created with a docker file that contains all the updates needed from a base docker image. A container is a runtime created from a docker image. Kaggle noiebooks run in a container. You can update the python environment in it with pip (or conda if conda is installed). You can update the container used by your notebook in Kaggle, but I doubt you can build your own docker image. This is to say that unfortunately you'll have to update each time you create a new notebook.
", "rawMarkdown": "I'm afraid you mix docker images and docker containers. A docker image is typically created with a docker file that contains all the updates needed from a base docker image. A container is a runtime created from a docker image. Kaggle noiebooks run in a container. You can update the python environment in it with pip (or conda if conda is installed). You can update the container used by your notebook in Kaggle, but I doubt you can build your own docker image. This is to say that unfortunately you'll have to update each time you create a new notebook.", "votes": null }, { "id": "607142", "postDate": "08/24/2019 16:53:58", "content": "@cpmpml The Jupiter runs in the container - yes, if you look at the list above (screens) - it is my list of docker images that I can use to run the container based on it. The problem is that this item on the scallop works very curtly - this list does not always appear. Ideally, I put a piped commit and I should have a new image, and next time I can choose this image, say with a library already installed. But after updating, I can't get this list at all. And through conda, this feeling does not work at all. So I would like to understand how to do it correctly, but it seems that this part of the conda needs a serious revision
\n\nP.S. sorry I write through a translator (deepl)
", "rawMarkdown": "cpmpml The Jupiter runs in the container - yes, if you look at the list above (screens) - it is my list of docker images that I can use to run the container based on it. The problem is that this item on the scallop works very curtly - this list does not always appear. Ideally, I put a piped commit and I should have a new image, and next time I can choose this image, say with a library already installed. But after updating, I can't get this list at all. And through conda, this feeling does not work at all. So I would like to understand how to do it correctly, but it seems that this part of the conda needs a serious revision \n\nP.S. sorry I write through a translator (deepl)", "votes": null } ], "comments": [ { "id": 607125, "author_name": "cpmpml", "author_url": "", "post_date": "08/24/2019 16:32:29", "content": "I'm afraid you mix docker images and docker containers. A docker image is typically created with a docker file that contains all the updates needed from a base docker image. A container is a runtime created from a docker image. Kaggle noiebooks run in a container. You can update the python environment in it with pip (or conda if conda is installed). You can update the container used by your notebook in Kaggle, but I doubt you can build your own docker image. This is to say that unfortunately you'll have to update each time you create a new notebook.
", "votes": null, "replies": [ { "id": 607142, "author_name": "langsys", "author_url": "", "post_date": "08/24/2019 16:53:58", "content": "@cpmpml The Jupiter runs in the container - yes, if you look at the list above (screens) - it is my list of docker images that I can use to run the container based on it. The problem is that this item on the scallop works very curtly - this list does not always appear. Ideally, I put a piped commit and I should have a new image, and next time I can choose this image, say with a library already installed. But after updating, I can't get this list at all. And through conda, this feeling does not work at all. So I would like to understand how to do it correctly, but it seems that this part of the conda needs a serious revision
\n\nP.S. sorry I write through a translator (deepl)
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "607101": "There's another problem, I bet, for instance,\n\n`!conda install -c rdkit rdkit -y`\n\nit comes from a notebook cell, it's fine. \n\nAs soon as the session fades or the context is reset, you have to put it anew - it's wildly annoying. It is said that it is possible to put a new image through the gui interface or through pip directly, and it is supposedly possible to choose a new image in the list with Docker - but it has never been possible to install it through the interface yet. Installation via pip rarely gives a drop-down list with new images.\n\n\nQuestions:\nWhat am I doing wrong?\nI want to get such a drop-down list of containers all the time, and put it with conda, how?\n\n\nIs there any way to update my docker image using conda?\nHow do I make my own copy of the poker image for my notebook computer so that new images appear? I'll be glad to have tutorials or explanations.\nNow the problem is that the utility script does not commit because there is not enough lib, but to put it without a laptop (in the script itself something does not seem like it). It remains a variant to make in the laptop the docker-image and to use it for a script, but the new image does not appear", "607125": "I'm afraid you mix docker images and docker containers. A docker image is typically created with a docker file that contains all the updates needed from a base docker image. A container is a runtime created from a docker image. Kaggle noiebooks run in a container. You can update the python environment in it with pip (or conda if conda is installed). You can update the container used by your notebook in Kaggle, but I doubt you can build your own docker image. This is to say that unfortunately you'll have to update each time you create a new notebook.", "607142": "cpmpml The Jupiter runs in the container - yes, if you look at the list above (screens) - it is my list of docker images that I can use to run the container based on it. The problem is that this item on the scallop works very curtly - this list does not always appear. Ideally, I put a piped commit and I should have a new image, and next time I can choose this image, say with a library already installed. But after updating, I can't get this list at all. And through conda, this feeling does not work at all. So I would like to understand how to do it correctly, but it seems that this part of the conda needs a serious revision \n\nP.S. sorry I write through a translator (deepl)" }, "source": "meta" }