I'm new to Kaggle and not yet familiar to all its capabilities. I'm curious… do most people participating in competitions like this run the code directly in a Kaggle notebook, or do they download the data, run the model locally in their own computer and then upload the result back to Kaggle? I'm asking b/c these are fairly large size files involved and uploading the answer to the kaggle server from my machine is a slow process. So how do most people do it? Thanks!
", "messages": [ { "id": 2358428, "postDate": "2023-07-25T13:48:34.410Z", "content": "well as u mentioned in your last line , local machine specifications can slow up the entire process , so why to approach the same thing in a two way process (from Kaggle -> local , then local -> Kaggle)
\nu can go in one way like run directly in your Kaggle notebook and add data in right hand side panel if your model needs it and the output data of your model will also be there in right hand , u can download it and then submit it
\nhope u understand
", "rawMarkdown": "well as u mentioned in your last line , local machine specifications can slow up the entire process , so why to approach the same thing in a two way process (from Kaggle -> local , then local -> Kaggle)\n\nu can go in one way like run directly in your Kaggle notebook and add data in right hand side panel if your model needs it and the output data of your model will also be there in right hand , u can download it and then submit it \n\nhope u understand \n\n\n\n", "votes": 1, "replies": [ { "id": 2358475, "postDate": "2023-07-25T14:28:58.107Z", "content": "DT, I got it, however, at times even the Kaggle notebook moves pretty slow. Is that always the case, or is it only here due to the large files involved?
", "rawMarkdown": "DT, I got it, however, at times even the Kaggle notebook moves pretty slow. Is that always the case, or is it only here due to the large files involved?", "replies": [ { "id": 2358491, "postDate": "2023-07-25T14:40:44.100Z", "content": "Hey @eduardastefanescu
\nHow about you check it yourself , like some say execution time changes every time ,
here is the code with the help of which you can check execution time of the cell ,
\nimport time
\nstart_time = time.time()
end_time = time.time()
\nprint(end_time - start_time)
I'm new to Kaggle and not yet familiar to all its capabilities. I'm curious… do most people participating in competitions like this run the code directly in a Kaggle notebook, or do they download the data, run the model locally in their own computer and then upload the result back to Kaggle? I'm asking b/c these are fairly large size files involved and uploading the answer to the kaggle server from my machine is a slow process. So how do most people do it? Thanks!
", "rawMarkdown": "I'm new to Kaggle and not yet familiar to all its capabilities. I'm curious... do most people participating in competitions like this run the code directly in a Kaggle notebook, or do they download the data, run the model locally in their own computer and then upload the result back to Kaggle? I'm asking b/c these are fairly large size files involved and uploading the answer to the kaggle server from my machine is a slow process. So how do most people do it? Thanks!", "votes": 1 }, { "id": 2358977, "postDate": "2023-07-25T22:02:49.737Z", "content": "In such comps, it is better to train the model locally and add it to a Kaggle kernel as a model/ dataset and write an inference and prediction code for submission @eduardastefanescu
", "rawMarkdown": "In such comps, it is better to train the model locally and add it to a Kaggle kernel as a model/ dataset and write an inference and prediction code for submission @eduardastefanescu ", "votes": 2, "replies": [ { "id": 2368958, "postDate": "2023-08-01T13:21:42.317Z", "content": "@ravi20076 May I ask where did you train the model? Was it on a cloud platform and how many GPU/TPU you needed?Many thanks!
", "rawMarkdown": "@ravi20076 May I ask where did you train the model? Was it on a cloud platform and how many GPU/TPU you needed?Many thanks!" } ] }, { "id": 2358473, "postDate": "2023-07-25T14:28:28.990Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 2358428, "author_name": "_student", "author_url": "", "post_date": "2023-07-25T13:48:34.410000", "content": "well as u mentioned in your last line , local machine specifications can slow up the entire process , so why to approach the same thing in a two way process (from Kaggle -> local , then local -> Kaggle)
\nu can go in one way like run directly in your Kaggle notebook and add data in right hand side panel if your model needs it and the output data of your model will also be there in right hand , u can download it and then submit it
\nhope u understand
", "votes": 1, "replies": [ { "id": 2358475, "author_name": "Eduard Stefanescu", "author_url": "", "post_date": "2023-07-25T14:28:58.107000", "content": "DT, I got it, however, at times even the Kaggle notebook moves pretty slow. Is that always the case, or is it only here due to the large files involved?
", "votes": 0, "replies": [ { "id": 2358491, "author_name": "_student", "author_url": "", "post_date": "2023-07-25T14:40:44.100000", "content": "Hey @eduardastefanescu
\nHow about you check it yourself , like some say execution time changes every time ,
here is the code with the help of which you can check execution time of the cell ,
\nimport time
\nstart_time = time.time()
end_time = time.time()
\nprint(end_time - start_time)
In such comps, it is better to train the model locally and add it to a Kaggle kernel as a model/ dataset and write an inference and prediction code for submission @eduardastefanescu
", "votes": 2, "replies": [ { "id": 2368958, "author_name": "PapaBougie", "author_url": "", "post_date": "2023-08-01T13:21:42.317000", "content": "@ravi20076 May I ask where did you train the model? Was it on a cloud platform and how many GPU/TPU you needed?Many thanks!
", "votes": 0, "replies": [] } ] }, { "id": 2358473, "author_name": "", "author_url": "", "post_date": "2023-07-25T14:28:28.990000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "2358428": "well as u mentioned in your last line , local machine specifications can slow up the entire process , so why to approach the same thing in a two way process (from Kaggle -> local , then local -> Kaggle)\n\nu can go in one way like run directly in your Kaggle notebook and add data in right hand side panel if your model needs it and the output data of your model will also be there in right hand , u can download it and then submit it \n\nhope u understand \n\n\n\n", "2358344": "I'm new to Kaggle and not yet familiar to all its capabilities. I'm curious... do most people participating in competitions like this run the code directly in a Kaggle notebook, or do they download the data, run the model locally in their own computer and then upload the result back to Kaggle? I'm asking b/c these are fairly large size files involved and uploading the answer to the kaggle server from my machine is a slow process. So how do most people do it? Thanks!", "2358977": "In such comps, it is better to train the model locally and add it to a Kaggle kernel as a model/ dataset and write an inference and prediction code for submission @eduardastefanescu ", "2358473": "" } }