Question

我对keras / tensorflow非常陌生，在使用训练有素的模型进行预测时遇到错误，我也不知道为什么。我的目的是预测产品的需求（请忽略我称之为weather_data的情况）。下面是我使用的代码，收到的错误消息以及最后的产品每周需求数据（以矢量形式，为期151周）：

max_len <- 6 # the number of previous examples we'll look at
batch_size <- 32 # number of sequences to look at at one time during training
total_epochs <- 15 # how many times we'll look @ the whole dataset while training our model

# set a random seed for reproducability
set.seed(123)

# Cut the text in overlapping sample sequences of max_len characters
  ## get a list of start indexes for our (overlapping) chunks
  start_indexes <- seq(1, length(Order_volume_vector) - (max_len + 1), by = 3)
  ## create an empty matrix to store our data in
  weather_matrix <- matrix(nrow = length(start_indexes), ncol = max_len + 1)
  ## fill our matrix with the overlapping slices of our dataset
  for (i in 1:length(start_indexes)){
    weather_matrix[i,] <- Order_volume_vector[start_indexes[i]:(start_indexes[i] + max_len)]
  }

# split our data into the day we're predict (y), and the 
# sequence of days leading up to it (X)
X <- weather_matrix[,-ncol(weather_matrix)]
y <- weather_matrix[,ncol(weather_matrix)]

# create an index to split data into testing & training sets
training_index <- createDataPartition(y, p = .9, 
                                      list = FALSE, 
                                      times = 1)
# training data
X_train <- array(X[training_index,], dim = c(length(training_index), max_len, 1))
y_train <- y[training_index]

# testing data
X_test <- array(X[-training_index,], dim = c(length(y) - length(training_index), max_len, 1))
y_test <- y[-training_index]

# initialize model
model <- keras_model_sequential()
model %>%
  layer_dense(input_shape = dim(X_train)[2:3], units = max_len)
model %>% 
  layer_simple_rnn(units = 6, activation = 'linear')
model %>%
  layer_dense(units = 1) # output
summary(model)
model %>% compile(loss = 'mse', 
                  optimizer = 'adam', 
                  metrics = 'mae')

##Train model
trained_model <- model %>% fit(
  x = X_train, # sequence we're using for prediction 
  y = y_train, # sequence we're predicting
  batch_size = batch_size, # how many samples to pass to our model at a time
  epochs = total_epochs, # how many times we'll look @ the whole dataset
  validation_split = 0.1) #

当我现在尝试使用模型进行预测时，出现以下错误：

Error in py_call_impl(callable, dots$args, dots$keywords) : 
  TypeError: Input 'b' of 'MatMul' Op has type float32 that does not match type int32 of argument 'a'.

Detailed traceback: 
  File "C:\Users\JENNIF~1\AppData\Local\CONTIN~1\ANACON~1\envs\R-RETI~1\lib\site-packages\tensorflow_core\python\keras\engine\training.py", line 909, in predict
    use_multiprocessing=use_multiprocessing)
  File "C:\Users\JENNIF~1\AppData\Local\CONTIN~1\ANACON~1\envs\R-RETI~1\lib\site-packages\tensorflow_core\python\keras\engine\training_v2.py", line 462, in predict
    steps=steps, callbacks=callbacks, **kwargs)
  File "C:\Users\JENNIF~1\AppData\Local\CONTIN~1\ANACON~1\envs\R-RETI~1\lib\site-packages\tensorflow_core\python\keras\engine\training_v2.py", line 444, in _model_iteration
    total_epochs=1)
  File "C:\Users\JENNIF~1\AppData\Local\CONTIN~1\ANACON~1\envs\R-RETI~1\lib\site-packages\tensorflow_core\python\keras\engine\training_v2.py", line 123, in run_one_epoch
    batch_outs = execution_function(iterator)
  File "C:\Users\JENNIF~1\AppData\Local\

我尝试使用不同的tensorflow软件包等重新安装软件包，但似乎无济于事。在网上我只能找到python的解决方案，我不知道该如何在R中应用。这是原始数据：

R，keras，TypeError：“ MatMul”操作的输入“ b”的类型为float32，与参数“ a”的类型int32不匹配

0 个答案: