![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F712087%2Faafeebcf8cde181eaa82e281e1e4aea0%2Fgpu_toggle.png?generation=1578958404950599&alt=media\")
I used the innate Kaggle GPU to train my model for 9 epochs. I started the training on 12 Jan, 7:45 PM IST. The GPU timer started with 00:00. The training was to be done for approx. 8 hours. On 13 JAN, 3:00 AM, the GPU timer soared to 22:00 hours. I was editing only one kernel at that time. How is this possible? Did anyone face such a problem?I used the innate Kaggle GPU to train my model for 9 epochs. I started the training on 12 Jan, 7:45 PM IST. The GPU timer started with 00:00. The training was to be done for approx. 8 hours. On 13 JAN, 3:00 AM , the GPU timer soared to 22:00 hours. I was editing only one kernel at that time. How is this possible? Did anyone face such a problem?
", "messages": [ { "id": "717432", "postDate": "01/13/2020 05:15:28", "content": "I used the innate Kaggle GPU to train my model for 9 epochs. I started the training on 12 Jan, 7:45 PM IST. The GPU timer started with 00:00. The training was to be done for approx. 8 hours. On 13 JAN, 3:00 AM, the GPU timer soared to 22:00 hours. I was editing only one kernel at that time. How is this possible? Did anyone face such a problem?I used the innate Kaggle GPU to train my model for 9 epochs. I started the training on 12 Jan, 7:45 PM IST. The GPU timer started with 00:00. The training was to be done for approx. 8 hours. On 13 JAN, 3:00 AM , the GPU timer soared to 22:00 hours. I was editing only one kernel at that time. How is this possible? Did anyone face such a problem?
", "rawMarkdown": "I used the innate Kaggle GPU to train my model for 9 epochs. I started the training on 12 Jan, 7:45 PM IST. The GPU timer started with 00:00. The training was to be done for approx. 8 hours. On 13 JAN, 3:00 AM, the GPU timer soared to 22:00 hours. I was editing only one kernel at that time. How is this possible? Did anyone face such a problem?I used the innate Kaggle GPU to train my model for 9 epochs. I started the training on 12 Jan, 7:45 PM IST. The GPU timer started with 00:00. The training was to be done for approx. 8 hours. On 13 JAN, 3:00 AM , the GPU timer soared to 22:00 hours. I was editing only one kernel at that time. How is this possible? Did anyone face such a problem?", "votes": null }, { "id": "717614", "postDate": "01/13/2020 11:07:50", "content": "Yes Bertram Gilfoyle, AI Marília ,absolute Lack of Intelligence, is here to support you. Even if I didn't join that competition.
", "rawMarkdown": "Yes Bertram Gilfoyle, AI Marília ,absolute Lack of Intelligence, is here to support you. Even if I didn't join that competition.", "votes": null }, { "id": "718041", "postDate": "01/13/2020 23:19:29", "content": "Hi @venky2506 - There are some important points to keep in mind when calculating your GPU quota usage. First, interactive sessions use GPU quota until they are stopped via the Run -> Power Off menu. Even if you start a Commit, the interactive session will continue to run, consuming your quota. Second, a Commit will continue to run until it's finished or you stop it using the controls at www.kaggle.com/your_user_name/account/gpu. Additionally, since you are permitted to run multiple Commits at a time, be careful not to accidentally start multiple Commits, as each will run and consume quota independently. Hope that helps.
", "rawMarkdown": "Hi @venky2506 - There are some important points to keep in mind when calculating your GPU quota usage. First, interactive sessions use GPU quota until they are stopped via the Run -> Power Off menu. Even if you start a Commit, the interactive session will continue to run, consuming your quota. Second, a Commit will continue to run until it's finished or you stop it using the controls at www.kaggle.com/your_user_name/account/gpu. Additionally, since you are permitted to run multiple Commits at a time, be careful not to accidentally start multiple Commits, as each will run and consume quota independently. Hope that helps.", "votes": null }, { "id": "718053", "postDate": "01/13/2020 23:34:54", "content": "@jplotts Do you mind taking a look at my quota? In just a couple of hours I created 29h of usage. I actually used the \"GPU toggle button\" to turn the GPU off. Are you saying that this button does not turn off the GPU in interactive mode?
\n\n
Edit: After committing I stopped the interactive GPU via \"Return to editor\" -> toggle \"GPU Off\"
", "rawMarkdown": "jplotts Do you mind taking a look at my quota? In just a couple of hours I created 29h of usage. I actually used the \"GPU toggle button\" to turn the GPU off. Are you saying that this button does not turn off the GPU in interactive mode?\n\n\n\nEdit: After committing I stopped the interactive GPU via \"Return to editor\" -> toggle \"GPU Off\"", "votes": null }, { "id": "718161", "postDate": "01/14/2020 05:33:52", "content": "It does clear up things. Thanks for the help
", "rawMarkdown": "It does clear up things. Thanks for the help", "votes": null }, { "id": "718162", "postDate": "01/14/2020 05:34:43", "content": "😂
", "rawMarkdown": "😂", "votes": null }, { "id": "718546", "postDate": "01/14/2020 14:15:26", "content": "Hi @seesee - That slider does indeed turn on/off GPU quota for the interactive session and all subsequent (but not currently running) Commits. I will respond to your other post called \"Notebook Errors\" regarding your quota usage.
", "rawMarkdown": "Hi @seesee - That slider does indeed turn on/off GPU quota for the interactive session and all subsequent (but not currently running) Commits. I will respond to your other post called \"Notebook Errors\" regarding your quota usage.", "votes": null } ], "comments": [ { "id": 717614, "author_name": "mpwolke", "author_url": "", "post_date": "01/13/2020 11:07:50", "content": "Yes Bertram Gilfoyle, AI Marília ,absolute Lack of Intelligence, is here to support you. Even if I didn't join that competition.
", "votes": null, "replies": [ { "id": 718162, "author_name": "venky2506", "author_url": "", "post_date": "01/14/2020 05:34:43", "content": "😂
", "votes": null, "replies": [] } ] }, { "id": 718041, "author_name": "jplotts", "author_url": "", "post_date": "01/13/2020 23:19:29", "content": "Hi @venky2506 - There are some important points to keep in mind when calculating your GPU quota usage. First, interactive sessions use GPU quota until they are stopped via the Run -> Power Off menu. Even if you start a Commit, the interactive session will continue to run, consuming your quota. Second, a Commit will continue to run until it's finished or you stop it using the controls at www.kaggle.com/your_user_name/account/gpu. Additionally, since you are permitted to run multiple Commits at a time, be careful not to accidentally start multiple Commits, as each will run and consume quota independently. Hope that helps.
", "votes": null, "replies": [ { "id": 718053, "author_name": "seesee", "author_url": "", "post_date": "01/13/2020 23:34:54", "content": "@jplotts Do you mind taking a look at my quota? In just a couple of hours I created 29h of usage. I actually used the \"GPU toggle button\" to turn the GPU off. Are you saying that this button does not turn off the GPU in interactive mode?
\n\n
Edit: After committing I stopped the interactive GPU via \"Return to editor\" -> toggle \"GPU Off\"
", "votes": null, "replies": [] }, { "id": 718161, "author_name": "venky2506", "author_url": "", "post_date": "01/14/2020 05:33:52", "content": "It does clear up things. Thanks for the help
", "votes": null, "replies": [] }, { "id": 718546, "author_name": "jplotts", "author_url": "", "post_date": "01/14/2020 14:15:26", "content": "Hi @seesee - That slider does indeed turn on/off GPU quota for the interactive session and all subsequent (but not currently running) Commits. I will respond to your other post called \"Notebook Errors\" regarding your quota usage.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "717432": "I used the innate Kaggle GPU to train my model for 9 epochs. I started the training on 12 Jan, 7:45 PM IST. The GPU timer started with 00:00. The training was to be done for approx. 8 hours. On 13 JAN, 3:00 AM, the GPU timer soared to 22:00 hours. I was editing only one kernel at that time. How is this possible? Did anyone face such a problem?I used the innate Kaggle GPU to train my model for 9 epochs. I started the training on 12 Jan, 7:45 PM IST. The GPU timer started with 00:00. The training was to be done for approx. 8 hours. On 13 JAN, 3:00 AM , the GPU timer soared to 22:00 hours. I was editing only one kernel at that time. How is this possible? Did anyone face such a problem?", "717614": "Yes Bertram Gilfoyle, AI Marília ,absolute Lack of Intelligence, is here to support you. Even if I didn't join that competition.", "718041": "Hi @venky2506 - There are some important points to keep in mind when calculating your GPU quota usage. First, interactive sessions use GPU quota until they are stopped via the Run -> Power Off menu. Even if you start a Commit, the interactive session will continue to run, consuming your quota. Second, a Commit will continue to run until it's finished or you stop it using the controls at www.kaggle.com/your_user_name/account/gpu. Additionally, since you are permitted to run multiple Commits at a time, be careful not to accidentally start multiple Commits, as each will run and consume quota independently. Hope that helps.", "718053": "jplotts Do you mind taking a look at my quota? In just a couple of hours I created 29h of usage. I actually used the \"GPU toggle button\" to turn the GPU off. Are you saying that this button does not turn off the GPU in interactive mode?\n\n\n\nEdit: After committing I stopped the interactive GPU via \"Return to editor\" -> toggle \"GPU Off\"", "718161": "It does clear up things. Thanks for the help", "718162": "😂", "718546": "Hi @seesee - That slider does indeed turn on/off GPU quota for the interactive session and all subsequent (but not currently running) Commits. I will respond to your other post called \"Notebook Errors\" regarding your quota usage." }, "source": "meta" }