[FIXED] Incorrect input shape after loading Keras model

Issue

I'm successfully loading a model from a file, and when I try to predict output on any input with shape (480,) I'm getting ValueError as follows:

ValueError: Input 0 of layer sequential_1 is incompatible with the layer: expected axis -1 of input shape to have value 480 but received input with shape (32, 1)

The code that I'm using

import numpy as np
from keras.models import load_model

model = load_model("my_model.h5")
sample = np.zeros((480,))
model.predict(sample)

Do you have any idea why is this happening?

My model summary after loading:

Training of this model looks reasonable, but if I'm trying to predict values just after invoking model.compile() the same error happens.

    model = keras.models.Sequential([
        keras.layers.Input(shape=(480,)),
        keras.layers.Dense(1024, activation="relu", name="layer1"),
        keras.layers.Dense(1024, activation="relu", name="layer2"),
        keras.layers.Dense(1024, activation="relu", name="layer3"),
        keras.layers.Dense(1024, activation="relu", name="layer4"),
        keras.layers.Dense(38, activation="softmax", name="output")
    ])
    model.compile(
        optimizer=keras.optimizers.Adam(lr=0.0003), 
        loss='categorical_crossentropy',
        metrics=['categorical_accuracy']
   )
   
   sample = np.zeros((480,))
   model.predict(sample)  # ValueError
   model.fit(...)  # works fine

Solution

This works for me instead; I added "1" to the shape of the input sample:

import numpy as np

model = keras.models.Sequential([
    keras.layers.Input(shape=(480,)),
    keras.layers.Dense(1024, activation="relu", name="layer1"),
    keras.layers.Dense(1024, activation="relu", name="layer2"),
    keras.layers.Dense(1024, activation="relu", name="layer3"),
    keras.layers.Dense(1024, activation="relu", name="layer4"),
    keras.layers.Dense(38, activation="softmax", name="output")
])
model.compile(
    optimizer=keras.optimizers.Adam(lr=0.0003),
    loss='categorical_crossentropy',
    metrics=['categorical_accuracy']
)

sample = np.zeros((1, 480,))
model.predict(sample)  # No ValueError

Answered By - J369