AttributeError: 'Context' object has no attribute '_rng'

[ChAoss0910]

Hi, I got an issue when running the training code below.

> wt_rnn(train_data = test_data,test_data = test_data,type="LSTM",catchment = "CM1",model_name = "LSTM")
*** Starting RNN computation for catchment CM1 ***
CM1 train data was sucessfully splitted!
number of missing days: 0 
CM1 validation data was sucessfully splitted!
number of missing days: 6 
CM1 test data was sucessfully splitted!
number of missing days: 6 
Mean and standard deviation used for feature scaling are saved under CM1/RNN/LSTM/LSTM/scaling_values.csv

Random hyperparameter sampling:
RNN type = LSTM, layers = 1, units = 200, dropout = 0.175, batch_size = 115, timesteps = 153, ensemble_runs = 1, Error in py_call_impl(callable, dots$args, dots$keywords) : 
  AttributeError: 'Context' object has no attribute '_rng'

I've double checked that there's no data format issue. Not sure if it's a bug with the source code.

Also tested with the original data given in the repo and got the same error:

> train_data <- feather::read_feather("test_catchment/train_data.feather")
> 
> test_data <- feather::read_feather("test_catchment/test_data.feather")
> 
> wt_fnn(train_data = train_data,test_data = test_data,catchment = "test_catchment",model_name = "fnn")
*** Starting FNN computation for catchment test_catchment ***
Mean and standard deviation used for feature scaling are saved under test_catchment/FNN/fnn/scaling_values.csv

Random hyperparameter sampling:
layers = 3, units = 53, dropout = 0.05, batch_size = 91, ensemble_runs = 1, Error in py_call_impl(callable, dots$args, dots$keywords) : 
  AttributeError: 'Context' object has no attribute '_rng'

Could you have a check if there's need of any upgrades with the newest version of dependent libraries? Looks like it has something wrong with the random number generator.

[mcvta]

Hi, I´m having the same problem as Hao Chen, while testing the original data. I don´t know what is "_rng". Can this be related with tensorflow version?
I´m using tensorflow 2.7.0. 9000
Thank you

data(test_catchment)
wt_preprocess(test_catchment)
train_data <- feather::read_feather("test_catchment/train_data.feather")
test_data <- feather::read_feather("test_catchment/test_data.feather")


wt_fnn(
  train_data,
  test_data = NULL,
  catchment = NULL,
  model_name = NULL,
  seed = NULL,
  n_iter = 40,
  n_random_initial_points = 20,
  epochs = 100,
  early_stopping_patience = 5,
  ensemble_runs = 5,
  bounds_layers = c(1, 5),
  bounds_units = c(5, 200),
  bounds_dropout = c(0, 0.2),
  bounds_batch_size = c(5, 150),
  initial_grid_from_model_scores = TRUE
)

wt_fnn(train_data, test_data, "test_catchment", "standard_FNN")

OUTPUT

wt_fnn(train_data, test_data, "test_catchment", "standard_FNN")
*** Starting FNN computation for catchment test_catchment ***
Mean and standard deviation used for feature scaling are saved under test_catchment/FNN/standard_FNN/scaling_values.csv

Random hyperparameter sampling:
layers = 3, units = 105, dropout = 0.025, batch_size = 27, ensemble_runs = 1,
Error in py_call_impl(callable, dots$args, dots$keywords) :
AttributeError: 'Context' object has no attribute '_rng'

[mcvta]

I know that the error is related with the random number generator (rng)

[ChAoss0910]

I know that the error is related with the random number generator (rng)

Hi just found out this could be resolved by adding a random seed number in the input parameters like this:
> wt_fnn(train_data = train_data,test_data = test_data,catchment = "CM1",seed = 42,model_name = "fnn")
Since the default seed number is set as NULL in the source code.

[mcvta]

Hi can you run this code?

library("wateRtemp")
library(tensorflow)
data(test_catchment)


wt_preprocess(test_catchment)
train_data <- feather::read_feather("test_catchment/train_data.feather")
test_data <- feather::read_feather("test_catchment/test_data.feather")


wt_fnn(
  train_data,
  test_data = NULL,
  catchment = NULL,
  model_name = NULL,
  seed = NULL,
  n_iter = 40,
  n_random_initial_points = 20,
  epochs = 100,
  early_stopping_patience = 5,
  ensemble_runs = 5,
  bounds_layers = c(1, 5),
  bounds_units = c(5, 200),
  bounds_dropout = c(0, 0.2),
  bounds_batch_size = c(5, 150),
  initial_grid_from_model_scores = TRUE
)

wt_fnn(train_data,test_data,catchment = "test_catchment",seed = 42,model_name = "fnn")

[mcvta]

Hi,
It´s running... great!

[mcvta]

Now I´m getting another error:

Error in py_call_impl(callable, dots$args, dots$keywords) :
TypeError: Exception encountered when calling layer "alpha_dropout_174" (type AlphaDropout).

'>' not supported between instances of 'dict' and 'float'

Call arguments received:
• inputs=tf.Tensor(shape=(None, 42), dtype=float32)
• training=None

[ChAoss0910]

Now I´m getting another error:

Error in py_call_impl(callable, dots$args, dots$keywords) : TypeError: Exception encountered when calling layer "alpha_dropout_174" (type AlphaDropout).

'>' not supported between instances of 'dict' and 'float'

Call arguments received: • inputs=tf.Tensor(shape=(None, 42), dtype=float32) • training=None

Got same error...looking into it now

[mcvta]

There may be missing values in the dataset?

[mcvta]

This are the parameters for the best model:
layers = 3
units = 200
max_epoc = 100
early_stopping_patience = 5
batch_size = 60
dropout = 2.22044604925031E-16
ensemble =1

Can this problem be related with the very small value of dropout =2.22044604925031E-16 ?