Using tensorflow through Keras - ValueError: Only input tensors may be passed as positional arguments

I am trying to get keras_model_sequential to work in R:

library(tidyverse)
library(keras)

model <- keras_model_sequential(input_shape = c(8)) 
model %>% 
  layer_dense(units = 32) %>% 
  layer_activation('softmax') 

Defining model goes just fine. Running the last line, however, gives the following error.

Error in py_call_impl(callable, call_args$unnamed, call_args$named) : 
  ValueError: Only input tensors may be passed as positional arguments. The following argument value should be passed as a keyword argument: <Sequential name=sequential_5, built=False> (of type <class 'keras.src.models.sequential.Sequential'>)
Run `reticulate::py_last_error()` for details.

Following the advice in the last line I get

── R Traceback ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
    ▆
 1. ├─model %>% layer_dense(units = 32) %>% ...
 2. ├─keras::layer_activation(., "softmax")
 3. │ └─keras::create_layer(...)
 4. └─keras::layer_dense(., units = 32)
 5.   └─keras::create_layer(...)
 6.     ├─keras:::compose_layer(object, layer)
 7.     └─keras:::compose_layer.default(object, layer)
 8.       └─reticulate (local) layer(object, ...)
 9.         └─reticulate:::py_call_impl(callable, call_args$unnamed, call_args$named)
See `reticulate::py_last_error()$r_trace$full_call` for more details.

I do not know what to make of this. I have tried to look online what the error means, but I am not getting problems that look similar, mostly a bunch of Python code.

In my efforts to solve this, I have updated my RStudio to 2024.09.0, Keras to 2.15.0 and R to 4.4.1 However, I am still getting exactly the same error.

How can I solve this error and successfully run keras_model_sequential?

The error seems to be a bug with current release of keras package. Please remove it and try keras3 instead as suggested here