Hello everyone!\nI'm relatively a beginner here, but I've been learning as fast as I can, this place is filled with more learning tutors and experience than I can imagine.
\n\nI have a few fairly performing kernels with a good LB score, an example is this . However, since the kaggle maintenance update, I've been having troubles rerunning previously well performing kernels. The LB scores have dwindled from 0.95+ to 0.49-0.6. \nI've somehow poinpointed the error to the Image Data Generator *flow_from_directory* (or dataframe) function from the keras package, since manual generators are performing well. \nSo I'm just asking if anyone has noticed a bug, or is it just me?
\n\nYou can check my code here if you have the time.\nI'm not sure if anyone have noticed or had the same problem,or perhpas it is not even a problem at all.
", "messages": [ { "id": 486819, "postDate": "2019-03-09T13:38:23.857Z", "content": "Hello everyone!\nI'm relatively a beginner here, but I've been learning as fast as I can, this place is filled with more learning tutors and experience than I can imagine.
\n\nI have a few fairly performing kernels with a good LB score, an example is this . However, since the kaggle maintenance update, I've been having troubles rerunning previously well performing kernels. The LB scores have dwindled from 0.95+ to 0.49-0.6. \nI've somehow poinpointed the error to the Image Data Generator *flow_from_directory* (or dataframe) function from the keras package, since manual generators are performing well. \nSo I'm just asking if anyone has noticed a bug, or is it just me?
\n\nYou can check my code here if you have the time.\nI'm not sure if anyone have noticed or had the same problem,or perhpas it is not even a problem at all.
", "rawMarkdown": "Hello everyone!\nI'm relatively a beginner here, but I've been learning as fast as I can, this place is filled with more learning tutors and experience than I can imagine.\n\nI have a few fairly performing kernels with a good LB score, an example is [this](https://www.kaggle.com/phekima/covnet-review-beginner-lb-0-94) . However, since the kaggle maintenance update, I've been having troubles rerunning previously well performing kernels. The LB scores have [dwindled from 0.95+ to 0.49-0.6.](https://www.kaggle.com/phekima/image-data-gen-test-2 ) \nI've somehow poinpointed the error to the Image Data Generator *flow_from_directory* (or dataframe) function from the keras package, since manual generators are performing well. \n**So I'm just asking if anyone has noticed a bug, or is it just me?**\n\nYou can check my code [here](https://www.kaggle.com/phekima/image-data-gen-test-2 ) if you have the time.\nI'm not sure if anyone have noticed or had the same problem,or perhpas it is not even a problem at all." } ], "comments": [], "raw_markdown_by_id": { "486819": "Hello everyone!\nI'm relatively a beginner here, but I've been learning as fast as I can, this place is filled with more learning tutors and experience than I can imagine.\n\nI have a few fairly performing kernels with a good LB score, an example is [this](https://www.kaggle.com/phekima/covnet-review-beginner-lb-0-94) . However, since the kaggle maintenance update, I've been having troubles rerunning previously well performing kernels. The LB scores have [dwindled from 0.95+ to 0.49-0.6.](https://www.kaggle.com/phekima/image-data-gen-test-2 ) \nI've somehow poinpointed the error to the Image Data Generator *flow_from_directory* (or dataframe) function from the keras package, since manual generators are performing well. \n**So I'm just asking if anyone has noticed a bug, or is it just me?**\n\nYou can check my code [here](https://www.kaggle.com/phekima/image-data-gen-test-2 ) if you have the time.\nI'm not sure if anyone have noticed or had the same problem,or perhpas it is not even a problem at all." } }