Hi, just wanted to check if I can submit a notebook for evaluation without actually training the entire model from scratch? Say I upload my trained model in the notebook using the 'Upload Dataset' option and then load this model in the notebook and infer on the test records to create submission.csv?
\nI like working on my local machine in scripts and am trying to avoid using the notebook structure as much as possible. Any thoughts on this would be appreciated!
", "messages": [ { "id": 2283241, "postDate": "2023-06-01T06:28:37.067Z", "content": "You just need the weights and inference code of your model to make a successful submission ! Also, notebook internet toggle must be off while submitting the notebook. Hope this helps.
", "rawMarkdown": "You just need the weights and inference code of your model to make a successful submission ! Also, notebook internet toggle must be off while submitting the notebook. Hope this helps.", "votes": 1, "replies": [ { "id": 2283249, "postDate": "2023-06-01T06:36:36.673Z", "content": "Thanks! That was helpful! One dumb question - is there a way to install a python package in the internet off mode? Is uploading the package wheels as data and then installing from it the only option? Thanks!
", "rawMarkdown": "Thanks! That was helpful! One dumb question - is there a way to install a python package in the internet off mode? Is uploading the package wheels as data and then installing from it the only option? Thanks!", "replies": [ { "id": 2283256, "postDate": "2023-06-01T06:43:54.180Z", "content": "Unfortunately, wheels is the only possible way right now. AFAIK.
", "rawMarkdown": "Unfortunately, wheels is the only possible way right now. AFAIK. ", "votes": 1 }, { "id": 2283781, "postDate": "2023-06-01T13:58:49.973Z", "content": "what also works is adding a Dataset to your notebook containing the repository of the desired library and loading it like described here: https://www.kaggle.com/datasets/kozodoi/timm-pytorch-image-models
", "rawMarkdown": "what also works is adding a Dataset to your notebook containing the repository of the desired library and loading it like described here: https://www.kaggle.com/datasets/kozodoi/timm-pytorch-image-models" } ] } ] }, { "id": 2283131, "postDate": "2023-06-01T04:47:37.677Z", "content": "Hi, just wanted to check if I can submit a notebook for evaluation without actually training the entire model from scratch? Say I upload my trained model in the notebook using the 'Upload Dataset' option and then load this model in the notebook and infer on the test records to create submission.csv?
\nI like working on my local machine in scripts and am trying to avoid using the notebook structure as much as possible. Any thoughts on this would be appreciated!
", "rawMarkdown": "Hi, just wanted to check if I can submit a notebook for evaluation without actually training the entire model from scratch? Say I upload my trained model in the notebook using the 'Upload Dataset' option and then load this model in the notebook and infer on the test records to create submission.csv?\n\nI like working on my local machine in scripts and am trying to avoid using the notebook structure as much as possible. Any thoughts on this would be appreciated!", "votes": 1 } ], "comments": [ { "id": 2283241, "author_name": "Kenni", "author_url": "", "post_date": "2023-06-01T06:28:37.067000", "content": "You just need the weights and inference code of your model to make a successful submission ! Also, notebook internet toggle must be off while submitting the notebook. Hope this helps.
", "votes": 1, "replies": [ { "id": 2283249, "author_name": "IlliniTango", "author_url": "", "post_date": "2023-06-01T06:36:36.673000", "content": "Thanks! That was helpful! One dumb question - is there a way to install a python package in the internet off mode? Is uploading the package wheels as data and then installing from it the only option? Thanks!
", "votes": 0, "replies": [ { "id": 2283256, "author_name": "Kenni", "author_url": "", "post_date": "2023-06-01T06:43:54.180000", "content": "Unfortunately, wheels is the only possible way right now. AFAIK.
", "votes": 1, "replies": [] }, { "id": 2283781, "author_name": "Fubusch", "author_url": "", "post_date": "2023-06-01T13:58:49.973000", "content": "what also works is adding a Dataset to your notebook containing the repository of the desired library and loading it like described here: https://www.kaggle.com/datasets/kozodoi/timm-pytorch-image-models
", "votes": 0, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "2283241": "You just need the weights and inference code of your model to make a successful submission ! Also, notebook internet toggle must be off while submitting the notebook. Hope this helps.", "2283131": "Hi, just wanted to check if I can submit a notebook for evaluation without actually training the entire model from scratch? Say I upload my trained model in the notebook using the 'Upload Dataset' option and then load this model in the notebook and infer on the test records to create submission.csv?\n\nI like working on my local machine in scripts and am trying to avoid using the notebook structure as much as possible. Any thoughts on this would be appreciated!" } }