new model using elu

.gitignore

@@ -1 +1,3 @@
 /a.out
+/checkpoint.h5
+/.ipynb_checkpoints/

README.md

@@ -37,19 +37,14 @@ Predict Free Chlorine & Bromine ppm from pH and ORP sensors.
 - Data source:  https://jenfitch.com/free-chlorine-orpmv-vs-ph-new/
 
 
-Build the model: [build_model.py](https://github.com/mzakharo/tubby/blob/main/build_model.py).
+Build the model: [fc_model.ipynb](https://github.com/mzakharo/tubby/blob/main/fc_model.ipynb).
 
-Prebuilt models:
+Prebuilt model:
 - [model_fc.h5](https://github.com/mzakharo/tubby/blob/main/model_fc.h5): model for predicting FC from  ORP/pH
 
-- [model_orp.h5](https://github.com/mzakharo/tubby/blob/main/model_orp.h5): model for predicting ORP from FC/PH. This one can be used to calibrate your ORP sensor by getting your water lab tested
-
-Basic [script](https://github.com/mzakharo/tubby/blob/main/predict.py) to load and run these models. usage:
+Basic [script](https://github.com/mzakharo/tubby/blob/main/predict.py) to load and run the model. usage:
 
 ```bash
-python3 ./predict.py  model_fc.h5 666 7.6
--> output:  0.83490694
-./predict.py model_orp.h5 7.6 0.85
--> output:  667.04565
+python3 predict.py  model_fc.h5 ORP_VALUE PH_VALUE
 ```
 

build_model.py

@@ -1,86 +1 @@
-import numpy as np
-import pandas as pd
-import sys
-import matplotlib.pyplot as plt
-import seaborn as sns
-import tensorflow as tf
-from tensorflow import keras
-from tensorflow.keras import layers
-
-TARGET = sys.argv[1] if len(sys.argv) == 2 else 'fc'
-
-print(tf.__version__)
-np.set_printoptions(precision=3, suppress=True)
-
-#Data from https://jenfitch.com/free-chlorine-orpmv-vs-ph-new/
-df = pd.read_csv('https://raw.githubusercontent.com/mzakharo/micropython/tubby/ports/esp32/orp.csv')
-print(df)
-labels = df.pop('ppm CL')
-
-def plot_loss(history):
-  plt.plot(history.history['loss'], label='loss')
-  plt.plot(history.history['val_loss'], label='val_loss')
-  #plt.ylim([0, 10])
-  plt.xlabel('Epoch')
-  plt.ylabel('Error')
-  plt.legend()
-  plt.grid(True)
-
-vs = []
-for column in df:
-    vals= df[column].to_numpy()
-    c = float(column)
-    for i, v in enumerate(vals):
-        vs.append((v, c, labels[i]))
-dataset = pd.DataFrame(vs, columns= ('orp', 'ph', 'fc'))
-
-train_dataset = dataset.sample(frac=0.9, random_state=0)
-test_dataset = dataset.drop(train_dataset.index)
-
-train_features = train_dataset.copy()
-train_labels = train_features.pop(TARGET)
-
-test_features = test_dataset.copy()
-test_labels = test_features.pop(TARGET)
-
-print(train_dataset.describe().transpose()[['mean', 'std']])
-#sns.pairplot(train_dataset, diag_kind='kde')
-
-inorm = tf.keras.layers.Normalization(axis=-1, input_shape=[2, ])
-inorm.adapt(np.array(train_features))
-onorm = tf.keras.layers.Normalization(axis=-1, invert=True)
-onorm.adapt(train_labels)
-
-def build_and_compile_model(inorm, onorm):
-  model = keras.Sequential([
-      inorm,
-      layers.Dense(10, activation='relu'),
-      layers.Dense(10, activation='relu'),
-      layers.Dense(1), 
-      onorm,
-  ])
-  model.compile(loss='mean_squared_error', optimizer='adam')
-  return model
-model = build_and_compile_model(inorm, onorm)
-model.summary()
-history = model.fit(
-    train_features,
-    train_labels,
-    validation_split=0.1,
-    verbose=0, epochs=200)
-print('mse:', model.evaluate(test_features, test_labels, verbose=0))
-plot_loss(history)
-y = model.predict(test_features, verbose=0)
-y = pd.DataFrame(y)[0]
-
-plt.figure(figsize=(5,5))
-sns.scatterplot(x=test_labels, y=y.to_numpy(), alpha=0.7)
-plt.grid()
-plt.show()
-
-model.save(f'model_{TARGET}.h5')
-converter = tf.lite.TFLiteConverter.from_keras_model(model)
-tflite_model = converter.convert()
-with open(f'model_{TARGET}.tflite', "wb") as f:
-  f.write(tflite_model)
-print('done')
+moved to fc_model.ipynb