Thank you for your participation in this competition! We will finalize the leaderboard in the next few days, and in the meantime we are opening up submissions for the TF2.0 Prizes as defined on the TensorFlow 2.0 Prizes page. If you are eligible, please submit this form, by January 31, 2020.
\n\nThe specifics (also on the form):\n- If you are part of a team, you need only submit one form for the team.\n- If you are a prospective winner, you will be required to provide the following:\n - Full TF2.0 solution training code & documentation for review and validation of eligibility by the host\n - Open sourcing of your solution on the competition forum\n - Saved Model to publish in TF-Hub\n - Winners' call with the host\n- If you are not willing to provide the above, do not submit for these prizes, as you will not be eligible.
\n\nIf your submission performed in the top 3 of eligible submissions, based on private leaderboard score, the host will verify your model and we will announce the winners upon verification.
", "messages": [ { "id": "726592", "postDate": "01/23/2020 05:39:09", "content": "Thank you for your participation in this competition! We will finalize the leaderboard in the next few days, and in the meantime we are opening up submissions for the TF2.0 Prizes as defined on the TensorFlow 2.0 Prizes page. If you are eligible, please submit this form, by January 31, 2020.
\n\nThe specifics (also on the form):\n- If you are part of a team, you need only submit one form for the team.\n- If you are a prospective winner, you will be required to provide the following:\n - Full TF2.0 solution training code & documentation for review and validation of eligibility by the host\n - Open sourcing of your solution on the competition forum\n - Saved Model to publish in TF-Hub\n - Winners' call with the host\n- If you are not willing to provide the above, do not submit for these prizes, as you will not be eligible.
\n\nIf your submission performed in the top 3 of eligible submissions, based on private leaderboard score, the host will verify your model and we will announce the winners upon verification.
", "rawMarkdown": "Thank you for your participation in this competition! We will finalize the leaderboard in the next few days, and in the meantime we are opening up submissions for the TF2.0 Prizes as defined on the [TensorFlow 2.0 Prizes](https://www.kaggle.com/c/tensorflow2-question-answering/overview/prizes) page. If you are eligible, please [submit this form](https://www.kaggle.com/TF20_Prize_Questionnaire), by **January 31, 2020**. \n\nThe specifics (also on the form):\n- If you are part of a team, you need only submit one form for the team.\n- If you are a prospective winner, you will be required to provide the following:\n - Full TF2.0 solution training code & documentation for review and validation of eligibility by the host\n - Open sourcing of your solution on the competition forum\n - Saved Model to publish in TF-Hub\n - Winners' call with the host\n- If you are not willing to provide the above, do not submit for these prizes, as you will not be eligible.\n\nIf your submission performed in the top 3 of eligible submissions, based on private leaderboard score, the host will verify your model and we will announce the winners upon verification.", "votes": null }, { "id": "726659", "postDate": "01/23/2020 06:36:17", "content": "@juliaelliott can we submit our best private LB score even if we didn't choose it as our final submission. My best TF 2.0 solution has a score of 0.64 but I didn't choose that one for my final submission.
", "rawMarkdown": "juliaelliott can we submit our best private LB score even if we didn't choose it as our final submission. My best TF 2.0 solution has a score of 0.64 but I didn't choose that one for my final submission.", "votes": null }, { "id": "726930", "postDate": "01/23/2020 10:18:18", "content": "Ah, that's unlucky. It must have cost you a lot of places.
", "rawMarkdown": "Ah, that's unlucky. It must have cost you a lot of places.", "votes": null }, { "id": "726938", "postDate": "01/23/2020 10:21:54", "content": "@juliaelliott my inference code is TF 2.1 but, at the time this comp started, TF2 had not yet implemented TPU support. So, in order to use the quota provided, I did training in TF1.15 and inference in TF2 in a kaggle kernel. Does that mean I am not eligible, or can I still translate my training code to TF2?
", "rawMarkdown": "juliaelliott my inference code is TF 2.1 but, at the time this comp started, TF2 had not yet implemented TPU support. So, in order to use the quota provided, I did training in TF1.15 and inference in TF2 in a kaggle kernel. Does that mean I am not eligible, or can I still translate my training code to TF2?", "votes": null }, { "id": "727036", "postDate": "01/23/2020 11:58:24", "content": "yes, it did😓
", "rawMarkdown": "yes, it did😓", "votes": null }, { "id": "727066", "postDate": "01/23/2020 12:35:55", "content": "@juliaelliott is it obliged to be a submitted (and selected) TF 2.0 Notebook? Or I can still refactor one of my Notebooks to TF 2.0?
", "rawMarkdown": "juliaelliott is it obliged to be a submitted (and selected) TF 2.0 Notebook? Or I can still refactor one of my Notebooks to TF 2.0?", "votes": null }, { "id": "727116", "postDate": "01/23/2020 13:23:49", "content": "interesting question
", "rawMarkdown": "interesting question", "votes": null }, { "id": "727127", "postDate": "01/23/2020 13:29:31", "content": "Well, common sense says no. Otherwise, one can adapt a publicly shared solution and claim that it's \"almost the same as my linear baseline\"
", "rawMarkdown": "Well, common sense says no. Otherwise, one can adapt a publicly shared solution and claim that it's \"almost the same as my linear baseline\"", "votes": null }, { "id": "727356", "postDate": "01/23/2020 16:43:45", "content": "I agree. I think both our suggestions will be against the rules, but it is worth checking with the referee to see where the boundary is.
", "rawMarkdown": "I agree. I think both our suggestions will be against the rules, but it is worth checking with the referee to see where the boundary is.", "votes": null }, { "id": "727458", "postDate": "01/23/2020 18:25:43", "content": "I think submitted before deadline is necessary. Maybe it is not necessary to be a selected one? 😄
", "rawMarkdown": "I think submitted before deadline is necessary. Maybe it is not necessary to be a selected one? 😄", "votes": null }, { "id": "727509", "postDate": "01/23/2020 18:59:08", "content": "Thanks for the clarifying questions. To respond:\n1. Your submitted code must have been written in TF2.0/TF2.1 at time of submission, including your training code. So, not it's not permitted to retroactively refactor 1.x code.\n2. Non-selected submissions (i.e. submissions that were not one of your 2 \"final submissions\") are eligible, as we have access to confirm your claimed private LB score.\n3. Yes, only submissions prior to the deadline are permitted.
", "rawMarkdown": "Thanks for the clarifying questions. To respond:\n1. Your submitted code must have been written in TF2.0/TF2.1 at time of submission, including your training code. So, not it's not permitted to retroactively refactor 1.x code.\n2. Non-selected submissions (i.e. submissions that were not one of your 2 \"final submissions\") **are eligible**, as we have access to confirm your claimed private LB score.\n3. Yes, only submissions prior to the deadline are permitted.", "votes": null }, { "id": "727816", "postDate": "01/24/2020 04:53:11", "content": "Thanks for the reply
", "rawMarkdown": "Thanks for the reply", "votes": null }, { "id": "727998", "postDate": "01/24/2020 10:11:27", "content": "@juliaelliott Could you provide some details on Saved Model to publish in TF-Hub. I didn't use TF-Hub yet. What format should the model have? How should it be packaged? I found this information so far: https://www.tensorflow.org/hub/publish. I think the tutorials only show how to use the models.
Edit: I found some more information regarding the model format: https://www.tensorflow.org/hub/tf2_saved_model
", "rawMarkdown": "juliaelliott Could you provide some details on `Saved Model to publish in TF-Hub`. I didn't use TF-Hub yet. What format should the model have? How should it be packaged? I found this information so far: https://www.tensorflow.org/hub/publish. I think the tutorials only show how to use the models.\n\nEdit: I found some more information regarding the model format: https://www.tensorflow.org/hub/tf2_saved_model", "votes": null }, { "id": "731511", "postDate": "01/28/2020 18:17:49", "content": "@seesee Great question - this is the detailed documentation on how to publish to TF Hub once you have produced your SavedModel.
", "rawMarkdown": "seesee Great question - [this is the detailed documentation](https://github.com/tensorflow/hub/tree/master/tfhub_dev) on how to publish to TF Hub once you have produced your SavedModel.", "votes": null }, { "id": "732946", "postDate": "01/30/2020 13:31:16", "content": "A reminder that tomorrow (1/31) is the deadline to enter your eligible submission for the TF2.0 prizes. Thanks to all who have already submitted. We’ll be in touch at the end of the day to collect materials for those in the top 3.
", "rawMarkdown": "A reminder that tomorrow (1/31) is the deadline to enter your eligible submission for the TF2.0 prizes. Thanks to all who have already submitted. We’ll be in touch at the end of the day to collect materials for those in the top 3.", "votes": null }, { "id": "733266", "postDate": "01/30/2020 22:53:12", "content": "@juliaelliott Thanks for the reminder. My code runs under tf 2.1 and mostly tf 2.1 but I found 'tf.compat.dimension_value' in my inference code and some 'tf.compat.v1' in code logic that i didn't use. Also I don't remember warnings while running my code. Can I still apply ?
", "rawMarkdown": "juliaelliott Thanks for the reminder. My code runs under tf 2.1 and mostly tf 2.1 but I found 'tf.compat.dimension_value' in my inference code and some 'tf.compat.v1' in code logic that i didn't use. Also I don't remember warnings while running my code. Can I still apply ?", "votes": null }, { "id": "733888", "postDate": "01/31/2020 16:38:52", "content": "@user189546 If you feel your solution meets the spirit of the description of eligibility in the prizes, I'd recommend you submit. The host will ultimately request & evaluate the code to judge whether it qualifies, should your private LB score be in the top 3 of eligible submissions.
", "rawMarkdown": "user189546 If you feel your solution meets the spirit of the description of eligibility in the prizes, I'd recommend you submit. The host will ultimately request & evaluate the code to judge whether it qualifies, should your private LB score be in the top 3 of eligible submissions.", "votes": null }, { "id": "733917", "postDate": "01/31/2020 17:19:12", "content": "Thank you! I applied.
", "rawMarkdown": "Thank you! I applied.", "votes": null } ], "comments": [ { "id": 726659, "author_name": "axel81", "author_url": "", "post_date": "01/23/2020 06:36:17", "content": "@juliaelliott can we submit our best private LB score even if we didn't choose it as our final submission. My best TF 2.0 solution has a score of 0.64 but I didn't choose that one for my final submission.
", "votes": null, "replies": [ { "id": 726930, "author_name": "kenkrige", "author_url": "", "post_date": "01/23/2020 10:18:18", "content": "Ah, that's unlucky. It must have cost you a lot of places.
", "votes": null, "replies": [] }, { "id": 727036, "author_name": "axel81", "author_url": "", "post_date": "01/23/2020 11:58:24", "content": "yes, it did😓
", "votes": null, "replies": [] } ] }, { "id": 726938, "author_name": "kenkrige", "author_url": "", "post_date": "01/23/2020 10:21:54", "content": "@juliaelliott my inference code is TF 2.1 but, at the time this comp started, TF2 had not yet implemented TPU support. So, in order to use the quota provided, I did training in TF1.15 and inference in TF2 in a kaggle kernel. Does that mean I am not eligible, or can I still translate my training code to TF2?
", "votes": null, "replies": [] }, { "id": 727066, "author_name": "kashnitsky", "author_url": "", "post_date": "01/23/2020 12:35:55", "content": "@juliaelliott is it obliged to be a submitted (and selected) TF 2.0 Notebook? Or I can still refactor one of my Notebooks to TF 2.0?
", "votes": null, "replies": [ { "id": 727116, "author_name": "christofhenkel", "author_url": "", "post_date": "01/23/2020 13:23:49", "content": "interesting question
", "votes": null, "replies": [] }, { "id": 727127, "author_name": "kashnitsky", "author_url": "", "post_date": "01/23/2020 13:29:31", "content": "Well, common sense says no. Otherwise, one can adapt a publicly shared solution and claim that it's \"almost the same as my linear baseline\"
", "votes": null, "replies": [] }, { "id": 727356, "author_name": "kenkrige", "author_url": "", "post_date": "01/23/2020 16:43:45", "content": "I agree. I think both our suggestions will be against the rules, but it is worth checking with the referee to see where the boundary is.
", "votes": null, "replies": [] }, { "id": 727458, "author_name": "jiweiliu", "author_url": "", "post_date": "01/23/2020 18:25:43", "content": "I think submitted before deadline is necessary. Maybe it is not necessary to be a selected one? 😄
", "votes": null, "replies": [] } ] }, { "id": 727509, "author_name": "juliaelliott", "author_url": "", "post_date": "01/23/2020 18:59:08", "content": "Thanks for the clarifying questions. To respond:\n1. Your submitted code must have been written in TF2.0/TF2.1 at time of submission, including your training code. So, not it's not permitted to retroactively refactor 1.x code.\n2. Non-selected submissions (i.e. submissions that were not one of your 2 \"final submissions\") are eligible, as we have access to confirm your claimed private LB score.\n3. Yes, only submissions prior to the deadline are permitted.
", "votes": null, "replies": [ { "id": 727816, "author_name": "axel81", "author_url": "", "post_date": "01/24/2020 04:53:11", "content": "Thanks for the reply
", "votes": null, "replies": [] } ] }, { "id": 727998, "author_name": "seesee", "author_url": "", "post_date": "01/24/2020 10:11:27", "content": "@juliaelliott Could you provide some details on Saved Model to publish in TF-Hub. I didn't use TF-Hub yet. What format should the model have? How should it be packaged? I found this information so far: https://www.tensorflow.org/hub/publish. I think the tutorials only show how to use the models.
Edit: I found some more information regarding the model format: https://www.tensorflow.org/hub/tf2_saved_model
", "votes": null, "replies": [ { "id": 731511, "author_name": "juliaelliott", "author_url": "", "post_date": "01/28/2020 18:17:49", "content": "@seesee Great question - this is the detailed documentation on how to publish to TF Hub once you have produced your SavedModel.
", "votes": null, "replies": [] } ] }, { "id": 732946, "author_name": "juliaelliott", "author_url": "", "post_date": "01/30/2020 13:31:16", "content": "A reminder that tomorrow (1/31) is the deadline to enter your eligible submission for the TF2.0 prizes. Thanks to all who have already submitted. We’ll be in touch at the end of the day to collect materials for those in the top 3.
", "votes": null, "replies": [ { "id": 733266, "author_name": "user189546", "author_url": "", "post_date": "01/30/2020 22:53:12", "content": "@juliaelliott Thanks for the reminder. My code runs under tf 2.1 and mostly tf 2.1 but I found 'tf.compat.dimension_value' in my inference code and some 'tf.compat.v1' in code logic that i didn't use. Also I don't remember warnings while running my code. Can I still apply ?
", "votes": null, "replies": [] }, { "id": 733888, "author_name": "juliaelliott", "author_url": "", "post_date": "01/31/2020 16:38:52", "content": "@user189546 If you feel your solution meets the spirit of the description of eligibility in the prizes, I'd recommend you submit. The host will ultimately request & evaluate the code to judge whether it qualifies, should your private LB score be in the top 3 of eligible submissions.
", "votes": null, "replies": [] }, { "id": 733917, "author_name": "user189546", "author_url": "", "post_date": "01/31/2020 17:19:12", "content": "Thank you! I applied.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "726592": "Thank you for your participation in this competition! We will finalize the leaderboard in the next few days, and in the meantime we are opening up submissions for the TF2.0 Prizes as defined on the [TensorFlow 2.0 Prizes](https://www.kaggle.com/c/tensorflow2-question-answering/overview/prizes) page. If you are eligible, please [submit this form](https://www.kaggle.com/TF20_Prize_Questionnaire), by **January 31, 2020**. \n\nThe specifics (also on the form):\n- If you are part of a team, you need only submit one form for the team.\n- If you are a prospective winner, you will be required to provide the following:\n - Full TF2.0 solution training code & documentation for review and validation of eligibility by the host\n - Open sourcing of your solution on the competition forum\n - Saved Model to publish in TF-Hub\n - Winners' call with the host\n- If you are not willing to provide the above, do not submit for these prizes, as you will not be eligible.\n\nIf your submission performed in the top 3 of eligible submissions, based on private leaderboard score, the host will verify your model and we will announce the winners upon verification.", "726659": "juliaelliott can we submit our best private LB score even if we didn't choose it as our final submission. My best TF 2.0 solution has a score of 0.64 but I didn't choose that one for my final submission.", "726930": "Ah, that's unlucky. It must have cost you a lot of places.", "726938": "juliaelliott my inference code is TF 2.1 but, at the time this comp started, TF2 had not yet implemented TPU support. So, in order to use the quota provided, I did training in TF1.15 and inference in TF2 in a kaggle kernel. Does that mean I am not eligible, or can I still translate my training code to TF2?", "727036": "yes, it did😓", "727066": "juliaelliott is it obliged to be a submitted (and selected) TF 2.0 Notebook? Or I can still refactor one of my Notebooks to TF 2.0?", "727116": "interesting question", "727127": "Well, common sense says no. Otherwise, one can adapt a publicly shared solution and claim that it's \"almost the same as my linear baseline\"", "727356": "I agree. I think both our suggestions will be against the rules, but it is worth checking with the referee to see where the boundary is.", "727458": "I think submitted before deadline is necessary. Maybe it is not necessary to be a selected one? 😄", "727509": "Thanks for the clarifying questions. To respond:\n1. Your submitted code must have been written in TF2.0/TF2.1 at time of submission, including your training code. So, not it's not permitted to retroactively refactor 1.x code.\n2. Non-selected submissions (i.e. submissions that were not one of your 2 \"final submissions\") **are eligible**, as we have access to confirm your claimed private LB score.\n3. Yes, only submissions prior to the deadline are permitted.", "727816": "Thanks for the reply", "727998": "juliaelliott Could you provide some details on `Saved Model to publish in TF-Hub`. I didn't use TF-Hub yet. What format should the model have? How should it be packaged? I found this information so far: https://www.tensorflow.org/hub/publish. I think the tutorials only show how to use the models.\n\nEdit: I found some more information regarding the model format: https://www.tensorflow.org/hub/tf2_saved_model", "731511": "seesee Great question - [this is the detailed documentation](https://github.com/tensorflow/hub/tree/master/tfhub_dev) on how to publish to TF Hub once you have produced your SavedModel.", "732946": "A reminder that tomorrow (1/31) is the deadline to enter your eligible submission for the TF2.0 prizes. Thanks to all who have already submitted. We’ll be in touch at the end of the day to collect materials for those in the top 3.", "733266": "juliaelliott Thanks for the reminder. My code runs under tf 2.1 and mostly tf 2.1 but I found 'tf.compat.dimension_value' in my inference code and some 'tf.compat.v1' in code logic that i didn't use. Also I don't remember warnings while running my code. Can I still apply ?", "733888": "user189546 If you feel your solution meets the spirit of the description of eligibility in the prizes, I'd recommend you submit. The host will ultimately request & evaluate the code to judge whether it qualifies, should your private LB score be in the top 3 of eligible submissions.", "733917": "Thank you! I applied." }, "source": "meta" }