tensorflow 值错误：形状不兼容

Question

My model's x is an array of float arrays (each sample is an array containing 40 elements). My model's y is also an array of float arrays (each sample is an array containing 80 elements). Here's the code reproducing my issue:

import tensorflow as tf
from tensorflow.keras import models, layers
import numpy as np

x = []
for i in range(100):
  array_of_random_floats = np.random.random_sample((40))
  x.append(array_of_random_floats)
x = np.asarray(x)

y = []
for i in range(100):
  array_of_random_floats = np.random.random_sample((80))
  y.append(array_of_random_floats)
y = np.asarray(y)

print(f"x has {len(x)} elements. Each element has {len(x[0])} elements")
# x has 100 elements. Each element has 40 elements

print(f"y has {len(y)} elements. Each element has {len(y[0])} elements")
# y has 100 elements. Each element has 80 elements


model = models.Sequential([
  layers.Input(shape=(40,)),
  layers.Dense(units=40),
])

model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
print(model.summary())

history = model.fit(x=x,
                    y=y,
                    epochs=100)

And this is the error produced.

ValueError: Shapes (None, 80) and (None, 40) are incompatible

What is going wrong?

Answer 1

In order to measure the loss, the dimensions need to match. You're trying to compare an output of (100, 40) with a target array of (100, 80) .