Here is a thread which contains all TensorFlow and TPU related notebooks in one place . Most of the authors are TPU stars and write TF related notebooks frequently . So you can choose to follow them in Kaggle for regular updates of their work
\nGetting Started: TPUs + Cassava Leaf Disease(Official)
\nhttps://www.kaggle.com/jessemostipak/getting-started-tpus-cassava-leaf-disease
TF Records :
\n@dimitreoliveira Cassava Leaf Disease-Stratified TFRecords 256x256
\nhttps://www.kaggle.com/dimitreoliveira/cassava-leaf-disease-stratified-tfrecords-256x256
@spidermandance Resize TFRecords and JPEG
\nhttps://www.kaggle.com/spidermandance/resize-tfrecords-and-jpeg
Cassava Leaf Disease - TPU Tensorflow - Training
\nhttps://www.kaggle.com/dimitreoliveira/cassava-leaf-disease-tpu-tensorflow-training
Cassava Leaf Disease - TPU Tensorflow - Inference
\nhttps://www.kaggle.com/dimitreoliveira/cassava-leaf-disease-tpu-tensorflow-inference
@xhlulu Cassava: Train EfficientNet on TPU in 100 lines
\nhttps://www.kaggle.com/xhlulu/cassava-train-efficientnet-on-tpu-in-100-lines
@datafan07 Getting the Best out of TPUs | Efficientnet & EDA
\nhttps://www.kaggle.com/datafan07/getting-the-best-out-of-tpus-efficientnet-eda
@ludovick Baseline TPU TF EfficientNet KFOLD [GPU-Inference]
\nhttps://www.kaggle.com/ludovick/baseline-tpu-tf-efficientnet-kfold-gpu-inference
EfficientNet and CutMixUp with TPU Train Phase
\nhttps://www.kaggle.com/itsuki9180/efficientnet-and-cutmixup-with-tpu-train-phase
EfficientNet and CutMixUp with TPU Predict Phase
\nhttps://www.kaggle.com/itsuki9180/efficientnet-and-cutmixup-with-tpu-predict-phase
@ludovick Baseline TF TPU EfficientNet Kfold [Training]
\nhttps://www.kaggle.com/ludovick/baseline-tf-tpu-efficientnet-kfold-training
@frlemarchand EfficientNet+Aug (TF/Keras) for Cassava Diseases
\nhttps://www.kaggle.com/frlemarchand/efficientnet-aug-tf-keras-for-cassava-diseases
@adrianromano C(l)assava(y) Leaf Diseases with CNN Keras
\nhttps://www.kaggle.com/adrianromano/c-l-assava-y-leaf-diseases-with-cnn-keras
@tarunbisht11 Cassava Leaf Disease Starter Notebook TF
\nhttps://www.kaggle.com/tarunbisht11/cassava-leaf-disease-starter-notebook-tf
@maksymshkliarevskyi Cassava Leaf Disease: Keras CNN baseline
\nhttps://www.kaggle.com/maksymshkliarevskyi/cassava-leaf-disease-keras-cnn-baseline
@anubhav1302 Plant_Diease_TF
\nhttps://www.kaggle.com/anubhav1302/plant-diease-tf
@wuliaokaola
\nGetting Started: TPUs (New tfrecords)
\nhttps://www.kaggle.com/wuliaokaola/getting-started-tpus-new-tfrecords
Tensorflow Resnet50 (train with new tfrecords)
\nhttps://www.kaggle.com/wuliaokaola/tensorflow-resnet50-train-with-new-tfrecords
@bjoernjostein Cassava Leaf Disease Classification using TF
\nhttps://www.kaggle.com/bjoernjostein/cassava-leaf-disease-classification-using-tf
@awsaf49
\nEfficientNetB6 512 CutMixUpDropout TPU [Train]
\nhttps://www.kaggle.com/awsaf49/efficientnetb6-512-cutmixupdropout-tpu-train
EfficientNetB6 512 CutMixUpDropout TPU [Infer]
\nhttps://www.kaggle.com/awsaf49/efficientnetb6-512-cutmixupdropout-tpu-infer
@oldpeter Cassava Leaf Disease Classification - tf2
\nhttps://www.kaggle.com/oldpeter/cassava-leaf-disease-classification-tf2
@cdk292 Efficientnet with R and tf2
\nhttps://www.kaggle.com/cdk292/efficientnet-with-r-and-tf2
@sayedathar11 EffNet & Augs with TPU Blazing Fast π₯π₯(CV-0.84)
\nhttps://www.kaggle.com/sayedathar11/effnet-augs-with-tpu-blazing-fast-cv-0-84
If you found it useful . If you are writing a TensorFlow notebook , please mention it in the comment , so I can include it in the list
", "messages": [ { "id": 1089429, "postDate": "2020-11-24T13:41:45.197Z", "content": "Here is a thread which contains all TensorFlow and TPU related notebooks in one place . Most of the authors are TPU stars and write TF related notebooks frequently . So you can choose to follow them in Kaggle for regular updates of their work
\nGetting Started: TPUs + Cassava Leaf Disease(Official)
\nhttps://www.kaggle.com/jessemostipak/getting-started-tpus-cassava-leaf-disease
TF Records :
\n@dimitreoliveira Cassava Leaf Disease-Stratified TFRecords 256x256
\nhttps://www.kaggle.com/dimitreoliveira/cassava-leaf-disease-stratified-tfrecords-256x256
@spidermandance Resize TFRecords and JPEG
\nhttps://www.kaggle.com/spidermandance/resize-tfrecords-and-jpeg
Cassava Leaf Disease - TPU Tensorflow - Training
\nhttps://www.kaggle.com/dimitreoliveira/cassava-leaf-disease-tpu-tensorflow-training
Cassava Leaf Disease - TPU Tensorflow - Inference
\nhttps://www.kaggle.com/dimitreoliveira/cassava-leaf-disease-tpu-tensorflow-inference
@xhlulu Cassava: Train EfficientNet on TPU in 100 lines
\nhttps://www.kaggle.com/xhlulu/cassava-train-efficientnet-on-tpu-in-100-lines
@datafan07 Getting the Best out of TPUs | Efficientnet & EDA
\nhttps://www.kaggle.com/datafan07/getting-the-best-out-of-tpus-efficientnet-eda
@ludovick Baseline TPU TF EfficientNet KFOLD [GPU-Inference]
\nhttps://www.kaggle.com/ludovick/baseline-tpu-tf-efficientnet-kfold-gpu-inference
EfficientNet and CutMixUp with TPU Train Phase
\nhttps://www.kaggle.com/itsuki9180/efficientnet-and-cutmixup-with-tpu-train-phase
EfficientNet and CutMixUp with TPU Predict Phase
\nhttps://www.kaggle.com/itsuki9180/efficientnet-and-cutmixup-with-tpu-predict-phase
@ludovick Baseline TF TPU EfficientNet Kfold [Training]
\nhttps://www.kaggle.com/ludovick/baseline-tf-tpu-efficientnet-kfold-training
@frlemarchand EfficientNet+Aug (TF/Keras) for Cassava Diseases
\nhttps://www.kaggle.com/frlemarchand/efficientnet-aug-tf-keras-for-cassava-diseases
@adrianromano C(l)assava(y) Leaf Diseases with CNN Keras
\nhttps://www.kaggle.com/adrianromano/c-l-assava-y-leaf-diseases-with-cnn-keras
@tarunbisht11 Cassava Leaf Disease Starter Notebook TF
\nhttps://www.kaggle.com/tarunbisht11/cassava-leaf-disease-starter-notebook-tf
@maksymshkliarevskyi Cassava Leaf Disease: Keras CNN baseline
\nhttps://www.kaggle.com/maksymshkliarevskyi/cassava-leaf-disease-keras-cnn-baseline
@anubhav1302 Plant_Diease_TF
\nhttps://www.kaggle.com/anubhav1302/plant-diease-tf
@wuliaokaola
\nGetting Started: TPUs (New tfrecords)
\nhttps://www.kaggle.com/wuliaokaola/getting-started-tpus-new-tfrecords
Tensorflow Resnet50 (train with new tfrecords)
\nhttps://www.kaggle.com/wuliaokaola/tensorflow-resnet50-train-with-new-tfrecords
@bjoernjostein Cassava Leaf Disease Classification using TF
\nhttps://www.kaggle.com/bjoernjostein/cassava-leaf-disease-classification-using-tf
@awsaf49
\nEfficientNetB6 512 CutMixUpDropout TPU [Train]
\nhttps://www.kaggle.com/awsaf49/efficientnetb6-512-cutmixupdropout-tpu-train
EfficientNetB6 512 CutMixUpDropout TPU [Infer]
\nhttps://www.kaggle.com/awsaf49/efficientnetb6-512-cutmixupdropout-tpu-infer
@oldpeter Cassava Leaf Disease Classification - tf2
\nhttps://www.kaggle.com/oldpeter/cassava-leaf-disease-classification-tf2
@cdk292 Efficientnet with R and tf2
\nhttps://www.kaggle.com/cdk292/efficientnet-with-r-and-tf2
@sayedathar11 EffNet & Augs with TPU Blazing Fast π₯π₯(CV-0.84)
\nhttps://www.kaggle.com/sayedathar11/effnet-augs-with-tpu-blazing-fast-cv-0-84
If you found it useful . If you are writing a TensorFlow notebook , please mention it in the comment , so I can include it in the list
", "rawMarkdown": "Here is a thread which contains all TensorFlow and TPU related notebooks in one place . Most of the authors are TPU stars and write TF related notebooks frequently . So you can choose to follow them in Kaggle for regular updates of their work \n\nGetting Started: TPUs + Cassava Leaf Disease(Official)\nhttps://www.kaggle.com/jessemostipak/getting-started-tpus-cassava-leaf-disease\n\n\nTF Records :\n\n@dimitreoliveira Cassava Leaf Disease-Stratified TFRecords 256x256\nhttps://www.kaggle.com/dimitreoliveira/cassava-leaf-disease-stratified-tfrecords-256x256\n\n@spidermandance Resize TFRecords and JPEG\nhttps://www.kaggle.com/spidermandance/resize-tfrecords-and-jpeg\n\n@dimitreoliveira\n\nCassava Leaf Disease - TPU Tensorflow - Training\nhttps://www.kaggle.com/dimitreoliveira/cassava-leaf-disease-tpu-tensorflow-training\n\nCassava Leaf Disease - TPU Tensorflow - Inference\nhttps://www.kaggle.com/dimitreoliveira/cassava-leaf-disease-tpu-tensorflow-inference\n\n@xhlulu Cassava: Train EfficientNet on TPU in 100 lines\nhttps://www.kaggle.com/xhlulu/cassava-train-efficientnet-on-tpu-in-100-lines\n\n@datafan07 Getting the Best out of TPUs | Efficientnet & EDA\nhttps://www.kaggle.com/datafan07/getting-the-best-out-of-tpus-efficientnet-eda\n\n@ludovick Baseline TPU TF EfficientNet KFOLD [GPU-Inference]\nhttps://www.kaggle.com/ludovick/baseline-tpu-tf-efficientnet-kfold-gpu-inference\n\n@itsuki9180 \n\nEfficientNet and CutMixUp with TPU Train Phase\nhttps://www.kaggle.com/itsuki9180/efficientnet-and-cutmixup-with-tpu-train-phase\n\nEfficientNet and CutMixUp with TPU Predict Phase\nhttps://www.kaggle.com/itsuki9180/efficientnet-and-cutmixup-with-tpu-predict-phase\n\n@ludovick Baseline TF TPU EfficientNet Kfold [Training]\nhttps://www.kaggle.com/ludovick/baseline-tf-tpu-efficientnet-kfold-training\n\n@frlemarchand EfficientNet+Aug (TF/Keras) for Cassava Diseases\nhttps://www.kaggle.com/frlemarchand/efficientnet-aug-tf-keras-for-cassava-diseases\n\n@adrianromano C(l)assava(y) Leaf Diseases with CNN Keras\nhttps://www.kaggle.com/adrianromano/c-l-assava-y-leaf-diseases-with-cnn-keras\n\n@tarunbisht11 Cassava Leaf Disease Starter Notebook TF\nhttps://www.kaggle.com/tarunbisht11/cassava-leaf-disease-starter-notebook-tf\n\n\n@maksymshkliarevskyi Cassava Leaf Disease: Keras CNN baseline\nhttps://www.kaggle.com/maksymshkliarevskyi/cassava-leaf-disease-keras-cnn-baseline\n\n@anubhav1302 Plant_Diease_TF\nhttps://www.kaggle.com/anubhav1302/plant-diease-tf\n\n@wuliaokaola \nGetting Started: TPUs (New tfrecords)\nhttps://www.kaggle.com/wuliaokaola/getting-started-tpus-new-tfrecords\n\nTensorflow Resnet50 (train with new tfrecords)\nhttps://www.kaggle.com/wuliaokaola/tensorflow-resnet50-train-with-new-tfrecords\n\n@bjoernjostein Cassava Leaf Disease Classification using TF\nhttps://www.kaggle.com/bjoernjostein/cassava-leaf-disease-classification-using-tf\n\n@awsaf49 \nEfficientNetB6 512 CutMixUpDropout TPU [Train]\nhttps://www.kaggle.com/awsaf49/efficientnetb6-512-cutmixupdropout-tpu-train\n\nEfficientNetB6 512 CutMixUpDropout TPU [Infer]\nhttps://www.kaggle.com/awsaf49/efficientnetb6-512-cutmixupdropout-tpu-infer\n\n@oldpeter Cassava Leaf Disease Classification - tf2\nhttps://www.kaggle.com/oldpeter/cassava-leaf-disease-classification-tf2\n\n@cdk292 Efficientnet with R and tf2\nhttps://www.kaggle.com/cdk292/efficientnet-with-r-and-tf2\n\n@sayedathar11 EffNet & Augs with TPU Blazing Fast π₯π₯(CV-0.84)\nhttps://www.kaggle.com/sayedathar11/effnet-augs-with-tpu-blazing-fast-cv-0-84\n\nIf you found it useful . If you are writing a TensorFlow notebook , please mention it in the comment , so I can include it in the list ", "votes": 80 }, { "id": 2185603, "postDate": "2023-03-17T07:23:06.260Z", "content": "How to get started with TPU? @usharengaraju can anyone share some notebooks for beginners to get started
", "rawMarkdown": "How to get started with TPU? @usharengaraju can anyone share some notebooks for beginners to get started", "votes": 2 }, { "id": 1094574, "postDate": "2020-11-28T18:26:33.987Z", "content": "Great post!
\nThis is what I was looking for.
Thanks for the compilation @usharengaraju , and for mentioning my work.
\nThere is a lot of notebooks using Tensorflow and TPUs, but I feel that some are making things more complicated than it is, one advice I give is to be very careful with the tf.data functions that build the data pipeline, it is very easy to mess things there and go unnoticed.
@dimitreoliveira Thank you for sharing this very important advice . I totally second this . I have also made a similar suggestion in TF community forum in Kaggle as well .
", "rawMarkdown": "@dimitreoliveira Thank you for sharing this very important advice . I totally second this . I have also made a similar suggestion in TF community forum in Kaggle as well .", "votes": 1 } ] }, { "id": 1136191, "postDate": "2021-01-02T20:09:37.123Z", "content": "Thanks a lot for sharing all the notebooks in one place @usharengaraju
", "rawMarkdown": "Thanks a lot for sharing all the notebooks in one place @usharengaraju ", "votes": 1 }, { "id": 1135552, "postDate": "2021-01-02T10:40:06.460Z", "content": "https://www.kaggle.com/shanmukh05/cassave-leaf-diseases-tpu . My LB score is 0.879 ,also thanks for sharing.
", "rawMarkdown": "https://www.kaggle.com/shanmukh05/cassave-leaf-diseases-tpu . My LB score is 0.879 ,also thanks for sharing.", "votes": 1 }, { "id": 1107402, "postDate": "2020-12-09T17:05:35.490Z", "content": "Thanks for sharing,
\nI will definitely try to refer them in my upcoming projects!
I have written a Tensorflow / Keras TPU Notebook here , The Cv result is 0.84 , it was the first time I got to work with TPU , learnt alot Thanks to Chris Deotte's Kernel on TPU Petals to Metals Competition and a Kernel at a Youtube Live Session. I am planning to make and publish a Kernel on Vision Transformer using TPU with better visualizations of results and accuracy !
", "rawMarkdown": "I have written a Tensorflow / Keras TPU Notebook [here ](https://www.kaggle.com/sayedathar11/effnet-augs-with-tpu-blazing-fast-cv-0-84) , The Cv result is 0.84 , it was the first time I got to work with TPU , learnt alot Thanks to Chris Deotte's Kernel on TPU Petals to Metals Competition and a Kernel at a Youtube Live Session. I am planning to make and publish a Kernel on Vision Transformer using TPU with better visualizations of results and accuracy !", "votes": 1, "replies": [ { "id": 1104922, "postDate": "2020-12-07T11:18:38.943Z", "content": "@sayedathar11 Great Kernel Athar ..Thanks for sharing it
", "rawMarkdown": "@sayedathar11 Great Kernel Athar ..Thanks for sharing it " }, { "id": 1104936, "postDate": "2020-12-07T11:37:06.117Z", "content": "@usharengaraju pleasure βΊοΈ is mine , Happy that you found it great .
", "rawMarkdown": "@usharengaraju pleasure βΊοΈ is mine , Happy that you found it great .", "votes": 1 } ] }, { "id": 1097040, "postDate": "2020-12-01T00:09:35.727Z", "content": "https://www.kaggle.com/venkat555/cassava-classification-with-densenet-and-cutmix?scriptVersionId=48146729 Thanks for the collection , learnt a lot from other notebooks here
", "rawMarkdown": "https://www.kaggle.com/venkat555/cassava-classification-with-densenet-and-cutmix?scriptVersionId=48146729 Thanks for the collection , learnt a lot from other notebooks here ", "votes": 1 }, { "id": 1096519, "postDate": "2020-11-30T15:08:44.453Z", "content": "Great source to have the bests in one post
", "rawMarkdown": "Great source to have the bests in one post", "votes": 1 }, { "id": 1094981, "postDate": "2020-11-29T06:42:04.850Z", "content": "which is best to work with image databases keras or tensorflow
", "rawMarkdown": "which is best to work with image databases keras or tensorflow", "votes": 1, "replies": [ { "id": 1095289, "postDate": "2020-11-29T12:53:50.320Z", "content": "Keras is a high level API with all functionalities possible as TF. I would suggest to go with Keras as its syntax is very much like sklearn (pythonic)
", "rawMarkdown": "Keras is a high level API with all functionalities possible as TF. I would suggest to go with Keras as its syntax is very much like sklearn (pythonic)", "votes": 1 } ] }, { "id": 1092809, "postDate": "2020-11-27T07:28:07.177Z", "content": "Great Work!!
", "rawMarkdown": "Great Work!!", "votes": 1 }, { "id": 1092790, "postDate": "2020-11-27T07:07:24.077Z", "content": "Why do most people prefer to train on TPUs instead of GPUs?
", "rawMarkdown": "Why do most people prefer to train on TPUs instead of GPUs? ", "votes": 1, "replies": [ { "id": 1092822, "postDate": "2020-11-27T07:37:04.773Z", "content": "@rashidali Compared to a GPU, TPU designed for a high volume of low precision computation with more input/output. TPU is highly-optimised for large batches and CNNs and has the highest training throughput. In simple words, TPU has a higher speed of training models.
", "rawMarkdown": "@rashidali Compared to a GPU, TPU designed for a high volume of low precision computation with more input/output. TPU is highly-optimised for large batches and CNNs and has the highest training throughput. In simple words, TPU has a higher speed of training models.", "votes": 3 } ] }, { "id": 1091287, "postDate": "2020-11-25T23:07:28.340Z", "content": "As someone with little exposure to TPUs, this is a very helpful collection of notebooks. Thanks for putting it together @usharengaraju π.
", "rawMarkdown": "As someone with little exposure to TPUs, this is a very helpful collection of notebooks. Thanks for putting it together @usharengaraju π.", "votes": 1 }, { "id": 1090555, "postDate": "2020-11-25T12:21:22.647Z", "content": "Thank you for collecting! Although I just wrote a simple notebook.
\nIn my notebook, I've released some useful tools including GeM, MixUp/CutMix, SWA, Focalloss.
@oldpeter Thanks for sharing very good Baseline notebook Peter . I went through it now . Was wondering why mentioned all the concepts as comments in your notebook
", "rawMarkdown": "@oldpeter Thanks for sharing very good Baseline notebook Peter . I went through it now . Was wondering why mentioned all the concepts as comments in your notebook ", "votes": -1 }, { "id": 1090569, "postDate": "2020-11-25T12:41:38.777Z", "content": "Actually, I usually use TF2 and Keras for rapid prototyping, and then use Pytorch to design better model.
", "rawMarkdown": "Actually, I usually use TF2 and Keras for rapid prototyping, and then use Pytorch to design better model.", "votes": 1 } ] }, { "id": 1089561, "postDate": "2020-11-24T15:52:58.683Z", "content": "Nice compilation, also thank you for sharing my work here!
", "rawMarkdown": "Nice compilation, also thank you for sharing my work here!", "votes": 2, "replies": [ { "id": 1090325, "postDate": "2020-11-25T08:58:02.687Z", "content": "@datafan07 Thank you so much . I am a keen follower of your notebook . Keep producing great work !!!
", "rawMarkdown": "@datafan07 Thank you so much . I am a keen follower of your notebook . Keep producing great work !!!", "votes": 1 } ] }, { "id": 1089472, "postDate": "2020-11-24T14:29:05.370Z", "content": "Nice work! A very useful list for adepts of TensorFlow π
\nAnd thanks for mentioning my notebook.
@maksymshkliarevskyi Thank you so much Maksym
", "rawMarkdown": "@maksymshkliarevskyi Thank you so much Maksym " } ] }, { "id": 1101670, "postDate": "2020-12-04T06:02:42.750Z", "rawMarkdown": "", "votes": 1, "isDeleted": true }, { "id": 1095288, "postDate": "2020-11-29T12:53:28.943Z", "rawMarkdown": "", "isDeleted": true }, { "id": 1118836, "postDate": "2020-12-19T12:44:17.063Z", "content": "Thanks for sharing ! @usharengaraju
", "rawMarkdown": "Thanks for sharing ! @usharengaraju ", "votes": 3 }, { "id": 1132672, "postDate": "2020-12-30T15:46:00.780Z", "content": "Thanks for sharing @usharengaraju
", "rawMarkdown": "Thanks for sharing @usharengaraju ", "votes": 1 }, { "id": 1103716, "postDate": "2020-12-06T07:06:13.787Z", "content": "Thanks for sharing!
", "rawMarkdown": "Thanks for sharing!", "votes": 1 }, { "id": 1098019, "postDate": "2020-12-01T11:58:39.727Z", "content": "Thanks for your great works!!
", "rawMarkdown": "Thanks for your great works!!", "votes": 1 }, { "id": 1097408, "postDate": "2020-12-01T03:40:32.637Z", "content": "Thanks for sharing!
", "rawMarkdown": "Thanks for sharing!", "votes": 1 }, { "id": 1093833, "postDate": "2020-11-28T04:19:48.977Z", "content": "Thanks for sharing the notebooks..
", "rawMarkdown": "Thanks for sharing the notebooks.. ", "votes": 1 }, { "id": 1089692, "postDate": "2020-11-24T17:37:54.140Z", "content": "Thanks for sharing my work. π
", "rawMarkdown": "Thanks for sharing my work. π", "votes": 1 }, { "id": 1089574, "postDate": "2020-11-24T16:01:27.570Z", "content": "Thank you for sharingπ
", "rawMarkdown": "Thank you for sharingπ", "votes": 1 }, { "id": 1096172, "postDate": "2020-11-30T09:26:50.903Z", "content": "Thanks for sharing πwill help a lotπ
", "rawMarkdown": "Thanks for sharing πwill help a lotπ", "votes": 2 }, { "id": 1147239, "postDate": "2021-01-10T11:23:59.480Z", "content": "Thanks for the compilation.
", "rawMarkdown": "Thanks for the compilation." }, { "id": 1137066, "postDate": "2021-01-03T16:40:00.487Z", "content": "Thanks for sharing!)π
", "rawMarkdown": "Thanks for sharing!)π" }, { "id": 1136616, "postDate": "2021-01-03T08:54:30.827Z", "content": "Thanks for sharing this! πππ
", "rawMarkdown": "Thanks for sharing this! πππ" } ], "comments": [ { "id": 2185603, "author_name": "Yeakub Sadlil", "author_url": "", "post_date": "2023-03-17T07:23:06.260000", "content": "How to get started with TPU? @usharengaraju can anyone share some notebooks for beginners to get started
", "votes": 2, "replies": [] }, { "id": 1094574, "author_name": "Abhinav Padmawar", "author_url": "", "post_date": "2020-11-28T18:26:33.987000", "content": "Great post!
\nThis is what I was looking for.
Thanks for the compilation @usharengaraju , and for mentioning my work.
\nThere is a lot of notebooks using Tensorflow and TPUs, but I feel that some are making things more complicated than it is, one advice I give is to be very careful with the tf.data functions that build the data pipeline, it is very easy to mess things there and go unnoticed.
@dimitreoliveira Thank you for sharing this very important advice . I totally second this . I have also made a similar suggestion in TF community forum in Kaggle as well .
", "votes": 1, "replies": [] } ] }, { "id": 1136191, "author_name": "Salman Ibne Eunus", "author_url": "", "post_date": "2021-01-02T20:09:37.123000", "content": "Thanks a lot for sharing all the notebooks in one place @usharengaraju
", "votes": 1, "replies": [] }, { "id": 1135552, "author_name": "Shanmukh", "author_url": "", "post_date": "2021-01-02T10:40:06.460000", "content": "https://www.kaggle.com/shanmukh05/cassave-leaf-diseases-tpu . My LB score is 0.879 ,also thanks for sharing.
", "votes": 1, "replies": [] }, { "id": 1107402, "author_name": "Abhinav Padmawar", "author_url": "", "post_date": "2020-12-09T17:05:35.490000", "content": "Thanks for sharing,
\nI will definitely try to refer them in my upcoming projects!
I have written a Tensorflow / Keras TPU Notebook here , The Cv result is 0.84 , it was the first time I got to work with TPU , learnt alot Thanks to Chris Deotte's Kernel on TPU Petals to Metals Competition and a Kernel at a Youtube Live Session. I am planning to make and publish a Kernel on Vision Transformer using TPU with better visualizations of results and accuracy !
", "votes": 1, "replies": [ { "id": 1104922, "author_name": "Tensor Girl", "author_url": "", "post_date": "2020-12-07T11:18:38.943000", "content": "@sayedathar11 Great Kernel Athar ..Thanks for sharing it
", "votes": 0, "replies": [] }, { "id": 1104936, "author_name": "Athar Sayed", "author_url": "", "post_date": "2020-12-07T11:37:06.117000", "content": "@usharengaraju pleasure βΊοΈ is mine , Happy that you found it great .
", "votes": 1, "replies": [] } ] }, { "id": 1097040, "author_name": "venkata yerubandi", "author_url": "", "post_date": "2020-12-01T00:09:35.727000", "content": "https://www.kaggle.com/venkat555/cassava-classification-with-densenet-and-cutmix?scriptVersionId=48146729 Thanks for the collection , learnt a lot from other notebooks here
", "votes": 1, "replies": [] }, { "id": 1096519, "author_name": "Jahed Naghipoor", "author_url": "", "post_date": "2020-11-30T15:08:44.453000", "content": "Great source to have the bests in one post
", "votes": 1, "replies": [] }, { "id": 1094981, "author_name": "vedasri bodavula", "author_url": "", "post_date": "2020-11-29T06:42:04.850000", "content": "which is best to work with image databases keras or tensorflow
", "votes": 1, "replies": [ { "id": 1095289, "author_name": "Kevin Joseph Scaria", "author_url": "", "post_date": "2020-11-29T12:53:50.320000", "content": "Keras is a high level API with all functionalities possible as TF. I would suggest to go with Keras as its syntax is very much like sklearn (pythonic)
", "votes": 1, "replies": [] } ] }, { "id": 1092809, "author_name": "MAYANK MONANI", "author_url": "", "post_date": "2020-11-27T07:28:07.177000", "content": "Great Work!!
", "votes": 1, "replies": [] }, { "id": 1092790, "author_name": "Rashid Ali", "author_url": "", "post_date": "2020-11-27T07:07:24.077000", "content": "Why do most people prefer to train on TPUs instead of GPUs?
", "votes": 1, "replies": [ { "id": 1092822, "author_name": "Maksym Shkliarevskyi", "author_url": "", "post_date": "2020-11-27T07:37:04.773000", "content": "@rashidali Compared to a GPU, TPU designed for a high volume of low precision computation with more input/output. TPU is highly-optimised for large batches and CNNs and has the highest training throughput. In simple words, TPU has a higher speed of training models.
", "votes": 3, "replies": [] } ] }, { "id": 1091287, "author_name": "Sean Burgess", "author_url": "", "post_date": "2020-11-25T23:07:28.340000", "content": "As someone with little exposure to TPUs, this is a very helpful collection of notebooks. Thanks for putting it together @usharengaraju π.
", "votes": 1, "replies": [] }, { "id": 1090555, "author_name": "Peter", "author_url": "", "post_date": "2020-11-25T12:21:22.647000", "content": "Thank you for collecting! Although I just wrote a simple notebook.
\nIn my notebook, I've released some useful tools including GeM, MixUp/CutMix, SWA, Focalloss.
@oldpeter Thanks for sharing very good Baseline notebook Peter . I went through it now . Was wondering why mentioned all the concepts as comments in your notebook
", "votes": -1, "replies": [] }, { "id": 1090569, "author_name": "Peter", "author_url": "", "post_date": "2020-11-25T12:41:38.777000", "content": "Actually, I usually use TF2 and Keras for rapid prototyping, and then use Pytorch to design better model.
", "votes": 1, "replies": [] } ] }, { "id": 1089561, "author_name": "ErtuΔrul Demir", "author_url": "", "post_date": "2020-11-24T15:52:58.683000", "content": "Nice compilation, also thank you for sharing my work here!
", "votes": 2, "replies": [ { "id": 1090325, "author_name": "Tensor Girl", "author_url": "", "post_date": "2020-11-25T08:58:02.687000", "content": "@datafan07 Thank you so much . I am a keen follower of your notebook . Keep producing great work !!!
", "votes": 1, "replies": [] } ] }, { "id": 1089472, "author_name": "Maksym Shkliarevskyi", "author_url": "", "post_date": "2020-11-24T14:29:05.370000", "content": "Nice work! A very useful list for adepts of TensorFlow π
\nAnd thanks for mentioning my notebook.
@maksymshkliarevskyi Thank you so much Maksym
", "votes": 0, "replies": [] } ] }, { "id": 1101670, "author_name": "", "author_url": "", "post_date": "2020-12-04T06:02:42.750000", "content": "", "votes": 1, "replies": [] }, { "id": 1095288, "author_name": "", "author_url": "", "post_date": "2020-11-29T12:53:28.943000", "content": "", "votes": 0, "replies": [] }, { "id": 1118836, "author_name": "Saurabh Shahane", "author_url": "", "post_date": "2020-12-19T12:44:17.063000", "content": "Thanks for sharing ! @usharengaraju
", "votes": 3, "replies": [] }, { "id": 1132672, "author_name": "Muslum Polat", "author_url": "", "post_date": "2020-12-30T15:46:00.780000", "content": "Thanks for sharing @usharengaraju
", "votes": 1, "replies": [] }, { "id": 1103716, "author_name": "Denny", "author_url": "", "post_date": "2020-12-06T07:06:13.787000", "content": "Thanks for sharing!
", "votes": 1, "replies": [] }, { "id": 1098019, "author_name": "Taewoo Jeong", "author_url": "", "post_date": "2020-12-01T11:58:39.727000", "content": "Thanks for your great works!!
", "votes": 1, "replies": [] }, { "id": 1097408, "author_name": "Theo Ginting", "author_url": "", "post_date": "2020-12-01T03:40:32.637000", "content": "Thanks for sharing!
", "votes": 1, "replies": [] }, { "id": 1093833, "author_name": "Ritacheta Das", "author_url": "", "post_date": "2020-11-28T04:19:48.977000", "content": "Thanks for sharing the notebooks..
", "votes": 1, "replies": [] }, { "id": 1089692, "author_name": "Awsaf", "author_url": "", "post_date": "2020-11-24T17:37:54.140000", "content": "Thanks for sharing my work. π
", "votes": 1, "replies": [] }, { "id": 1089574, "author_name": "Takayoshi Makabe", "author_url": "", "post_date": "2020-11-24T16:01:27.570000", "content": "Thank you for sharingπ
", "votes": 1, "replies": [] }, { "id": 1096172, "author_name": "datajameson", "author_url": "", "post_date": "2020-11-30T09:26:50.903000", "content": "Thanks for sharing πwill help a lotπ
", "votes": 2, "replies": [] }, { "id": 1147239, "author_name": "Ashutosh Parmar", "author_url": "", "post_date": "2021-01-10T11:23:59.480000", "content": "Thanks for the compilation.
", "votes": 0, "replies": [] }, { "id": 1137066, "author_name": "Rinat Din", "author_url": "", "post_date": "2021-01-03T16:40:00.487000", "content": "Thanks for sharing!)π
", "votes": 0, "replies": [] }, { "id": 1136616, "author_name": "", "author_url": "", "post_date": "2021-01-03T08:54:30.827000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1089429": "Here is a thread which contains all TensorFlow and TPU related notebooks in one place . Most of the authors are TPU stars and write TF related notebooks frequently . So you can choose to follow them in Kaggle for regular updates of their work \n\nGetting Started: TPUs + Cassava Leaf Disease(Official)\nhttps://www.kaggle.com/jessemostipak/getting-started-tpus-cassava-leaf-disease\n\n\nTF Records :\n\n@dimitreoliveira Cassava Leaf Disease-Stratified TFRecords 256x256\nhttps://www.kaggle.com/dimitreoliveira/cassava-leaf-disease-stratified-tfrecords-256x256\n\n@spidermandance Resize TFRecords and JPEG\nhttps://www.kaggle.com/spidermandance/resize-tfrecords-and-jpeg\n\n@dimitreoliveira\n\nCassava Leaf Disease - TPU Tensorflow - Training\nhttps://www.kaggle.com/dimitreoliveira/cassava-leaf-disease-tpu-tensorflow-training\n\nCassava Leaf Disease - TPU Tensorflow - Inference\nhttps://www.kaggle.com/dimitreoliveira/cassava-leaf-disease-tpu-tensorflow-inference\n\n@xhlulu Cassava: Train EfficientNet on TPU in 100 lines\nhttps://www.kaggle.com/xhlulu/cassava-train-efficientnet-on-tpu-in-100-lines\n\n@datafan07 Getting the Best out of TPUs | Efficientnet & EDA\nhttps://www.kaggle.com/datafan07/getting-the-best-out-of-tpus-efficientnet-eda\n\n@ludovick Baseline TPU TF EfficientNet KFOLD [GPU-Inference]\nhttps://www.kaggle.com/ludovick/baseline-tpu-tf-efficientnet-kfold-gpu-inference\n\n@itsuki9180 \n\nEfficientNet and CutMixUp with TPU Train Phase\nhttps://www.kaggle.com/itsuki9180/efficientnet-and-cutmixup-with-tpu-train-phase\n\nEfficientNet and CutMixUp with TPU Predict Phase\nhttps://www.kaggle.com/itsuki9180/efficientnet-and-cutmixup-with-tpu-predict-phase\n\n@ludovick Baseline TF TPU EfficientNet Kfold [Training]\nhttps://www.kaggle.com/ludovick/baseline-tf-tpu-efficientnet-kfold-training\n\n@frlemarchand EfficientNet+Aug (TF/Keras) for Cassava Diseases\nhttps://www.kaggle.com/frlemarchand/efficientnet-aug-tf-keras-for-cassava-diseases\n\n@adrianromano C(l)assava(y) Leaf Diseases with CNN Keras\nhttps://www.kaggle.com/adrianromano/c-l-assava-y-leaf-diseases-with-cnn-keras\n\n@tarunbisht11 Cassava Leaf Disease Starter Notebook TF\nhttps://www.kaggle.com/tarunbisht11/cassava-leaf-disease-starter-notebook-tf\n\n\n@maksymshkliarevskyi Cassava Leaf Disease: Keras CNN baseline\nhttps://www.kaggle.com/maksymshkliarevskyi/cassava-leaf-disease-keras-cnn-baseline\n\n@anubhav1302 Plant_Diease_TF\nhttps://www.kaggle.com/anubhav1302/plant-diease-tf\n\n@wuliaokaola \nGetting Started: TPUs (New tfrecords)\nhttps://www.kaggle.com/wuliaokaola/getting-started-tpus-new-tfrecords\n\nTensorflow Resnet50 (train with new tfrecords)\nhttps://www.kaggle.com/wuliaokaola/tensorflow-resnet50-train-with-new-tfrecords\n\n@bjoernjostein Cassava Leaf Disease Classification using TF\nhttps://www.kaggle.com/bjoernjostein/cassava-leaf-disease-classification-using-tf\n\n@awsaf49 \nEfficientNetB6 512 CutMixUpDropout TPU [Train]\nhttps://www.kaggle.com/awsaf49/efficientnetb6-512-cutmixupdropout-tpu-train\n\nEfficientNetB6 512 CutMixUpDropout TPU [Infer]\nhttps://www.kaggle.com/awsaf49/efficientnetb6-512-cutmixupdropout-tpu-infer\n\n@oldpeter Cassava Leaf Disease Classification - tf2\nhttps://www.kaggle.com/oldpeter/cassava-leaf-disease-classification-tf2\n\n@cdk292 Efficientnet with R and tf2\nhttps://www.kaggle.com/cdk292/efficientnet-with-r-and-tf2\n\n@sayedathar11 EffNet & Augs with TPU Blazing Fast π₯π₯(CV-0.84)\nhttps://www.kaggle.com/sayedathar11/effnet-augs-with-tpu-blazing-fast-cv-0-84\n\nIf you found it useful . If you are writing a TensorFlow notebook , please mention it in the comment , so I can include it in the list ", "2185603": "How to get started with TPU? @usharengaraju can anyone share some notebooks for beginners to get started", "1094574": "Great post!\nThis is what I was looking for.", "1089439": "Thanks for the compilation @usharengaraju , and for mentioning my work.\nThere is a lot of notebooks using Tensorflow and TPUs, but I feel that some are making things more complicated than it is, one advice I give is to be very careful with the `tf.data` functions that build the data pipeline, it is very easy to mess things there and go unnoticed.", "1136191": "Thanks a lot for sharing all the notebooks in one place @usharengaraju ", "1135552": "https://www.kaggle.com/shanmukh05/cassave-leaf-diseases-tpu . My LB score is 0.879 ,also thanks for sharing.", "1107402": "Thanks for sharing,\nI will definitely try to refer them in my upcoming projects!", "1104887": "I have written a Tensorflow / Keras TPU Notebook [here ](https://www.kaggle.com/sayedathar11/effnet-augs-with-tpu-blazing-fast-cv-0-84) , The Cv result is 0.84 , it was the first time I got to work with TPU , learnt alot Thanks to Chris Deotte's Kernel on TPU Petals to Metals Competition and a Kernel at a Youtube Live Session. I am planning to make and publish a Kernel on Vision Transformer using TPU with better visualizations of results and accuracy !", "1097040": "https://www.kaggle.com/venkat555/cassava-classification-with-densenet-and-cutmix?scriptVersionId=48146729 Thanks for the collection , learnt a lot from other notebooks here ", "1096519": "Great source to have the bests in one post", "1094981": "which is best to work with image databases keras or tensorflow", "1092809": "Great Work!!", "1092790": "Why do most people prefer to train on TPUs instead of GPUs? ", "1091287": "As someone with little exposure to TPUs, this is a very helpful collection of notebooks. Thanks for putting it together @usharengaraju π.", "1090555": "Thank you for collecting! Although I just wrote a simple notebook.\nIn my notebook, I've released some useful tools including GeM, MixUp/CutMix, SWA, Focalloss.", "1089561": "Nice compilation, also thank you for sharing my work here!", "1089472": "Nice work! A very useful list for adepts of TensorFlow π\nAnd thanks for mentioning my notebook.", "1101670": "", "1095288": "", "1118836": "Thanks for sharing ! @usharengaraju ", "1132672": "Thanks for sharing @usharengaraju ", "1103716": "Thanks for sharing!", "1098019": "Thanks for your great works!!", "1097408": "Thanks for sharing!", "1093833": "Thanks for sharing the notebooks.. ", "1089692": "Thanks for sharing my work. π", "1089574": "Thank you for sharingπ", "1096172": "Thanks for sharing πwill help a lotπ", "1147239": "Thanks for the compilation.", "1137066": "Thanks for sharing!)π", "1136616": "Thanks for sharing this! πππ" } }