Paste your server credentials
CSV format: timestamp,sensor1,sensor2,... — header row optional
If you have large datasets, train on a server (Keras) and convert to TFJS using tensorflowjs_converter. Upload model files to Firebase Storage and load in the app.
Server-side flow: 1) Train with Keras (Python) and save HDF5 or SavedModel 2) Use tensorflowjs_converter to convert to TFJS format 3) Upload generated model.json and shard files to Kumul Cloud Storage 4) Call tf.loadLayersModel() in the client