Keras Eval Stream¶
Overview¶
Compatible Pyspark only
Python node : launch with punchplatform-pyspark.sh or the punch UI job-editor
with environment pyspark
.
This node define and train a small demonstrative keras model to recognise digits from the mnist database. It publish the trained keras model to the output stream.
The model created is a feed-forward neural network with one hidden layer of 128 neurons. The training is done over 5 epoch with a batch size of 50.
Since it's mostly a specific demo node, it don't have any settings.
Example configuration :¶
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 | { type: keras_eval_stream component: my_keras subscribe: [ { component: data_input stream: x_test } { component: model_load stream: model } ] publish: [ { stream: predicts } ] } |
Example of complete pipeline :¶
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 | { type: pyspark runtime_id: simpleJob job: [ { type: keras_data_input component: data_input publish: [ { stream: x_test } ] } { type: keras_load_model component: model_load settings:{ path: my_model.h5 } publish: [ { stream: model } ] } { type: keras_eval_stream component: my_keras subscribe: [ { component: data_input stream: x_test } { component: model_load stream: model } ] publish: [ { stream: predicts } ] } { type: show component: show, subscribe: [ { component: my_keras stream: predicts } ] } ] } |