It’s clear that source code should be included, but what about the followings?\n* Base models’ packages with weights pretrained on imagenet (e.g. https://github.com/qubvel/efficientnet)\n* Saved weights that will be loaded into the model to generate predictions\n* Ensemble strategy
", "messages": [ { "id": 647019, "postDate": "2019-10-12T01:25:44.917Z", "content": "It’s clear that source code should be included, but what about the followings?\n* Base models’ packages with weights pretrained on imagenet (e.g. https://github.com/qubvel/efficientnet)\n* Saved weights that will be loaded into the model to generate predictions\n* Ensemble strategy
", "rawMarkdown": "It’s clear that source code should be included, but what about the followings?\n* Base models’ packages with weights pretrained on imagenet (e.g. https://github.com/qubvel/efficientnet)\n* Saved weights that will be loaded into the model to generate predictions\n* Ensemble strategy", "votes": 10 }, { "id": 648839, "postDate": "2019-10-14T17:09:32.363Z", "content": "If you use pre-trained models, you should include which models were used. Output weight files do not need to be included, just the code that was used to generate them. If you ensemble multiple models, include all of them in a single zip.
", "rawMarkdown": "If you use pre-trained models, you should include which models were used. Output weight files do not need to be included, just the code that was used to generate them. If you ensemble multiple models, include all of them in a single zip.", "votes": 2, "replies": [ { "id": 649601, "postDate": "2019-10-15T15:05:29.763Z", "content": "Thanks for the reply!
", "rawMarkdown": "Thanks for the reply!", "votes": 1 }, { "id": 649864, "postDate": "2019-10-15T20:56:57.087Z", "content": "@juliaelliott Another quick question. I use git version control and modify the model on the fly. Some of my models have changed parameters over time but it's traceable with git. Is that fine?
", "rawMarkdown": "@juliaelliott Another quick question. I use git version control and modify the model on the fly. Some of my models have changed parameters over time but it's traceable with git. Is that fine?", "votes": 1 }, { "id": 649870, "postDate": "2019-10-15T21:06:57.747Z", "content": "@realethanzou the model upload is still required. You may reference your git repo in any associated README documentation, but the physical code is still required to be uploaded to kaggle by the end of stage 1.
", "rawMarkdown": "@realethanzou the model upload is still required. You may reference your git repo in any associated README documentation, but the physical code is still required to be uploaded to kaggle by the end of stage 1.", "votes": 2 }, { "id": 663046, "postDate": "2019-11-01T11:41:16.833Z", "content": "it means one py or ipython file is ok and during the stage 2, i can't change angthing in the py file?
", "rawMarkdown": "it means one py or ipython file is ok and during the stage 2, i can't change angthing in the py file?", "votes": 1 }, { "id": 663101, "postDate": "2019-11-01T12:58:55.137Z", "content": "@juliaelliott Don't need the code to run the models? Just a README of how it is to be used? In the case of ensemble, my zip will have:\n- model1.h5\n- model2.h5\n- model3.h5\n- README
\n\nThe Readme must contain what was used for image processing (in case I used a dataset that is hosted on Kaggle) could you send his link and the code used to generate the dataset?
\n\nAnother question, should I put my code used to train my models? How many epochs and etc?
", "rawMarkdown": " @juliaelliott Don't need the code to run the models? Just a README of how it is to be used? In the case of ensemble, my zip will have:\n- model1.h5\n- model2.h5\n- model3.h5\n- README\n\nThe Readme must contain what was used for image processing (in case I used a dataset that is hosted on Kaggle) could you send his link and the code used to generate the dataset?\n\nAnother question, should I put my code used to train my models? How many epochs and etc?\n\n" }, { "id": 663172, "postDate": "2019-11-01T14:37:45.160Z", "content": "@custodiogabriel Yes, the code is required, except for any pretrained model(s) you are using. For pretrained models, you can reference them in a README. Otherwise the full code for your models that are used to generate your submission should be included in your upload.
", "rawMarkdown": "@custodiogabriel Yes, the code is required, except for any pretrained model(s) you are using. For pretrained models, you can reference them in a README. Otherwise the full code for your models that are used to generate your submission should be included in your upload.", "votes": 1 }, { "id": 663402, "postDate": "2019-11-02T00:39:19.580Z", "content": "If I have multiple ipython,weather I should merge these into one ipython. What's more,the local path \nis ok? or we should use the path like the public notebook
", "rawMarkdown": "If I have multiple ipython,weather I should merge these into one ipython. What's more,the local path \nis ok? or we should use the path like the public notebook" } ] } ], "comments": [ { "id": 648839, "author_name": "Julia Elliott", "author_url": "", "post_date": "2019-10-14T17:09:32.363000", "content": "If you use pre-trained models, you should include which models were used. Output weight files do not need to be included, just the code that was used to generate them. If you ensemble multiple models, include all of them in a single zip.
", "votes": 2, "replies": [ { "id": 649601, "author_name": "Yifeng (Ethan) Zou", "author_url": "", "post_date": "2019-10-15T15:05:29.763000", "content": "Thanks for the reply!
", "votes": 1, "replies": [] }, { "id": 649864, "author_name": "Yifeng (Ethan) Zou", "author_url": "", "post_date": "2019-10-15T20:56:57.087000", "content": "@juliaelliott Another quick question. I use git version control and modify the model on the fly. Some of my models have changed parameters over time but it's traceable with git. Is that fine?
", "votes": 1, "replies": [] }, { "id": 649870, "author_name": "Julia Elliott", "author_url": "", "post_date": "2019-10-15T21:06:57.747000", "content": "@realethanzou the model upload is still required. You may reference your git repo in any associated README documentation, but the physical code is still required to be uploaded to kaggle by the end of stage 1.
", "votes": 2, "replies": [] }, { "id": 663046, "author_name": "yhfu", "author_url": "", "post_date": "2019-11-01T11:41:16.833000", "content": "it means one py or ipython file is ok and during the stage 2, i can't change angthing in the py file?
", "votes": 1, "replies": [] }, { "id": 663101, "author_name": "Gabriel", "author_url": "", "post_date": "2019-11-01T12:58:55.137000", "content": "@juliaelliott Don't need the code to run the models? Just a README of how it is to be used? In the case of ensemble, my zip will have:\n- model1.h5\n- model2.h5\n- model3.h5\n- README
\n\nThe Readme must contain what was used for image processing (in case I used a dataset that is hosted on Kaggle) could you send his link and the code used to generate the dataset?
\n\nAnother question, should I put my code used to train my models? How many epochs and etc?
", "votes": 0, "replies": [] }, { "id": 663172, "author_name": "Julia Elliott", "author_url": "", "post_date": "2019-11-01T14:37:45.160000", "content": "@custodiogabriel Yes, the code is required, except for any pretrained model(s) you are using. For pretrained models, you can reference them in a README. Otherwise the full code for your models that are used to generate your submission should be included in your upload.
", "votes": 1, "replies": [] }, { "id": 663402, "author_name": "yhfu", "author_url": "", "post_date": "2019-11-02T00:39:19.580000", "content": "If I have multiple ipython,weather I should merge these into one ipython. What's more,the local path \nis ok? or we should use the path like the public notebook
", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "647019": "It’s clear that source code should be included, but what about the followings?\n* Base models’ packages with weights pretrained on imagenet (e.g. https://github.com/qubvel/efficientnet)\n* Saved weights that will be loaded into the model to generate predictions\n* Ensemble strategy", "648839": "If you use pre-trained models, you should include which models were used. Output weight files do not need to be included, just the code that was used to generate them. If you ensemble multiple models, include all of them in a single zip." } }