Updates to tutorial #846
Updates to tutorial #846
Conversation
| 1. **Create or identify a kernel on Kaggle.com.** | ||
| * Log in to kaggle.com. | ||
| * Find an existing notebook (or create one). For this tutorial, let's assume its title is "My CLI Test Kernel". | ||
| * If the notebook has not been saved before, make a small change and **save a version** of the notebook on Kaggle.com (e.g., click "Save Version" and choose "Save & Run All (Commit)"). You cannot pull or push a kernel that is only in draft form. |
There was a problem hiding this comment.
If this is still true, it feels like a launch blocker. It might have been fixed though.
jplotts
left a comment
jplotts
left a comment
There was a problem hiding this comment.
One other confusion in this document is that it splits up creating Models vs Model Instances vs Model Instance Versions. I'm not sure what is the usefulness of creating a Model or Model Instance without creating a Model Instance Version. Even though it's possible, we should consider streamlining the tutorial to just go all the way to Model Instance Version, since that's what most users will want.
| 1. Installed the Kaggle CLI. | ||
| 2. Downloaded your `kaggle.json` API token from your Kaggle account page (e.g., `https://www.kaggle.com/settings`) and placed it in the `~/.kaggle/` directory (or `C:\Users\<Windows-username>\.kaggle\` on Windows). | ||
| 3. Logged in to your kaggle.com account in a web browser. This will allow you to easily verify the results of the CLI commands in the "Your Work" section of your Kaggle profile. | ||
| 1. Installed the Kaggle CLI, following the instructions [here](https://github.com/Kaggle/kaggle-api/blob/main/documentation/intro.md#installation). |
There was a problem hiding this comment.
This document uses CLI, but it's in a repo called "Kaggle API". Let's sync with Abhishek on the naming.
| * Once you are happy with your model, you can submit your prediction to the competition. You can do this using the `kaggle competitions submit` command: | ||
| ```bash | ||
| kaggle competitions submit -c <competition-name> -f <submission-file> -m <message> | ||
| kaggle competitions submit -c <competition-name> -k <username>/<notebook-slug> -m <message> |
There was a problem hiding this comment.
This looked right based on the code, but I didn't test it. Steve - can you confirm?
Creating a model instance creates the first version. Similar to how creating a dataset creates the first version for that dataset and then you need to create new dataset version. |
| 5. **Create the model.** | ||
|
|
||
| ```bash | ||
| kaggle models create -p . |
There was a problem hiding this comment.
Let's remove the section with sed commands above ^^.
I don't see much value. Users can use whatever they like to edit this file.
I think it just creates more confusion for users not familiar with sed.
That's good, so let's update the tutorial to be the minimum the user needs to do to get a working Model. Is it a model + model instance? |
Yes. And if they want to upload new versions of an existing model instance, then they need to use the new model version flow. Agree, to keep the tutorial short, we can get rid of the new version flow and maybe just include a note about it and a link to the documentation to create a new model instance version? |
| * Once you are happy with your model, you can submit your prediction to the competition. You can do this using the `kaggle competitions submit` command: | ||
| ```bash | ||
| kaggle competitions submit -c <competition-name> -f <submission-file> -m <message> | ||
| kaggle competitions submit -c <competition-name> -k <username>/<notebook-slug> -m <message> |
3 participants
No description provided.