\n
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F10705026%2F8669db08eb82db3ffaea871d74c28809%2Ferror1.png?generation=1678957934082686&alt=media\")
\nbut it can run successful.
How do I submit a tflite model, and why do I fail to submit it.
\n
\nbut it can run successful.
How do I submit a tflite model, and why do I fail to submit it.
\n
\nbut it can run successful.
Your notebook is expected to produce a submission.zip file, which should contain a saved tflite model. That model will be used to evaluate the hidden test set. Personally, I prepare a submission.zip file locally on my laptop and upload it to kaggle datasets. The inference notebook just takes this submission and saves it to the output folder. That's it.
\nHope, that helps you. If not, please provide more insight about your inference notebook
", "rawMarkdown": "Your notebook is expected to produce a submission.zip file, which should contain a saved tflite model. That model will be used to evaluate the hidden test set. Personally, I prepare a submission.zip file locally on my laptop and upload it to kaggle datasets. The inference notebook just takes this submission and saves it to the output folder. That's it.\n\nHope, that helps you. If not, please provide more insight about your inference notebook", "votes": null }, { "id": "2184406", "postDate": "03/16/2023 11:07:00", "content": "most likely tflite runtime is giving error, etc.
\nIf my model input shape is not (543,3), Should I cut the input data to the shape that my modle need? And write the preprocess function in my model structure?
", "rawMarkdown": "If my model input shape is not (543,3), Should I cut the input data to the shape that my modle need? And write the preprocess function in my model structure?", "votes": null }, { "id": "2191858", "postDate": "03/22/2023 08:43:10", "content": "Exactly. During the evaluation, your model will be given a (None, 543, 3) sample. If you use only hands, you should slice the input tensors inside your model. That may be tricky sometimes, especially, if you are going to exploit some advanced data preprocessing - all the data processing should be converted to the TfLite model.
", "rawMarkdown": "Exactly. During the evaluation, your model will be given a (None, 543, 3) sample. If you use only hands, you should slice the input tensors inside your model. That may be tricky sometimes, especially, if you are going to exploit some advanced data preprocessing - all the data processing should be converted to the TfLite model.", "votes": null }, { "id": "2191876", "postDate": "03/22/2023 09:02:11", "content": "ok，I got it. Thanks very much.
", "rawMarkdown": "ok，I got it. Thanks very much.", "votes": null } ], "comments": [ { "id": 2184312, "author_name": "meowmeowmeowmeowmeow", "author_url": "", "post_date": "03/16/2023 09:49:34", "content": "Your notebook is expected to produce a submission.zip file, which should contain a saved tflite model. That model will be used to evaluate the hidden test set. Personally, I prepare a submission.zip file locally on my laptop and upload it to kaggle datasets. The inference notebook just takes this submission and saves it to the output folder. That's it.
\nHope, that helps you. If not, please provide more insight about your inference notebook
", "votes": null, "replies": [] }, { "id": 2184406, "author_name": "hengck23", "author_url": "", "post_date": "03/16/2023 11:07:00", "content": "most likely tflite runtime is giving error, etc.
\nIf my model input shape is not (543,3), Should I cut the input data to the shape that my modle need? And write the preprocess function in my model structure?
", "votes": null, "replies": [ { "id": 2191858, "author_name": "meowmeowmeowmeowmeow", "author_url": "", "post_date": "03/22/2023 08:43:10", "content": "Exactly. During the evaluation, your model will be given a (None, 543, 3) sample. If you use only hands, you should slice the input tensors inside your model. That may be tricky sometimes, especially, if you are going to exploit some advanced data preprocessing - all the data processing should be converted to the TfLite model.
", "votes": null, "replies": [ { "id": 2191876, "author_name": "gavin113", "author_url": "", "post_date": "03/22/2023 09:02:11", "content": "ok，I got it. Thanks very much.
", "votes": null, "replies": [] } ] } ] } ] } ], "raw_markdown_by_id": { "2184276": "How do I submit a tflite model, and why do I fail to submit it.\n\nbut it can run successful.", "2184312": "Your notebook is expected to produce a submission.zip file, which should contain a saved tflite model. That model will be used to evaluate the hidden test set. Personally, I prepare a submission.zip file locally on my laptop and upload it to kaggle datasets. The inference notebook just takes this submission and saves it to the output folder. That's it.\n\nHope, that helps you. If not, please provide more insight about your inference notebook", "2184406": "most likely tflite runtime is giving error, etc.\n1. the input size is wrong (e.g. dynamic axis)\n2. some of the ops you used are not supported by tflite runtime\n (e.g. flexi ops, tf ops)", "2188830": "1. check your input shape is (543, 3)\n2. check the name of the input layer is \"inputs\"\ne.g. tf.keras.Input(shape=(543,3)\n3. check your output layer name is \"outputs\"\n4. try running your model with the code of the evaluation [page](https://www.kaggle.com/competitions/asl-signs/overview/evaluation)", "2191837": "If my model input shape is not (543,3), Should I cut the input data to the shape that my modle need? And write the preprocess function in my model structure?", "2191858": "Exactly. During the evaluation, your model will be given a (None, 543, 3) sample. If you use only hands, you should slice the input tensors inside your model. That may be tricky sometimes, especially, if you are going to exploit some advanced data preprocessing - all the data processing should be converted to the TfLite model.", "2191876": "ok，I got it. Thanks very much." }, "source": "meta" }