opt() errors with "Error in as.numeric(tfe$tf_optimiser$iterations): cannot coerce type 'environment' to vector of type 'double'"

[njtierney]

library(greta)
#> 
#> Attaching package: 'greta'
#> The following objects are masked from 'package:stats':
#> 
#>     binomial, cov2cor, poisson
#> The following objects are masked from 'package:base':
#> 
#>     %*%, apply, backsolve, beta, chol2inv, colMeans, colSums, diag,
#>     eigen, forwardsolve, gamma, identity, rowMeans, rowSums, sweep,
#>     tapply
x <- rnorm(5, 2, 0.1)
z <- variable(dim = 5)
#> ℹ Initialising python and checking dependencies, this may take a moment.
#> ✔ Initialising python and checking dependencies ... done!
#> 
distribution(x) <- normal(z, 0.1)

m <- model(z)

opr_grad <- opt(m, optimiser = gradient_descent())
#> Error in as.numeric(tfe$tf_optimiser$iterations): cannot coerce type 'environment' to vector of type 'double'
opr_adadelta <- opt(m, optimiser = adadelta())
#> Error in as.numeric(tfe$tf_optimiser$iterations): cannot coerce type 'environment' to vector of type 'double'
opr_adagrad <- opt(m, optimiser = adagrad())
#> Error in as.numeric(tfe$tf_optimiser$iterations): cannot coerce type 'environment' to vector of type 'double'
opr_adam <- opt(m, optimiser = adam())
#> Error in as.numeric(tfe$tf_optimiser$iterations): cannot coerce type 'environment' to vector of type 'double'
opr_ftrl <- opt(m, optimiser = ftrl())
#> Error in as.numeric(tfe$tf_optimiser$iterations): cannot coerce type 'environment' to vector of type 'double'
opr_rms_prop <- opt(m, optimiser = rms_prop())
#> Error in as.numeric(tfe$tf_optimiser$iterations): cannot coerce type 'environment' to vector of type 'double'

^{Created on 2024-05-09 with reprex v2.1.0}

Session info

``` r sessioninfo::session_info() #> ─ Session info ─────────────────────────────────────────────────────────────── #> setting value #> version R version 4.4.0 (2024-04-24) #> os macOS Sonoma 14.3.1 #> system aarch64, darwin20 #> ui X11 #> language (EN) #> collate en_US.UTF-8 #> ctype en_US.UTF-8 #> tz Australia/Brisbane #> date 2024-05-09 #> pandoc 3.1.13 @ /opt/homebrew/bin/ (via rmarkdown) #> #> ─ Packages ─────────────────────────────────────────────────────────────────── #> package * version date (UTC) lib source #> abind 1.4-5 2016-07-21 [1] CRAN (R 4.4.0) #> backports 1.4.1 2021-12-13 [1] CRAN (R 4.4.0) #> base64enc 0.1-3 2015-07-28 [1] CRAN (R 4.4.0) #> callr 3.7.6 2024-03-25 [1] CRAN (R 4.4.0) #> cli 3.6.2 2023-12-11 [1] CRAN (R 4.4.0) #> coda 0.19-4.1 2024-01-31 [1] CRAN (R 4.4.0) #> codetools 0.2-20 2024-03-31 [2] CRAN (R 4.4.0) #> crayon 1.5.2 2022-09-29 [1] CRAN (R 4.4.0) #> digest 0.6.35 2024-03-11 [1] CRAN (R 4.4.0) #> evaluate 0.23 2023-11-01 [1] CRAN (R 4.4.0) #> fastmap 1.1.1 2023-02-24 [1] CRAN (R 4.4.0) #> fs 1.6.4 2024-04-25 [1] CRAN (R 4.4.0) #> future 1.33.2 2024-03-26 [1] CRAN (R 4.4.0) #> globals 0.16.3 2024-03-08 [1] CRAN (R 4.4.0) #> glue 1.7.0 2024-01-09 [1] CRAN (R 4.4.0) #> greta * 0.4.5.9000 2024-05-09 [1] local #> hms 1.1.3 2023-03-21 [1] CRAN (R 4.4.0) #> htmltools 0.5.8.1 2024-04-04 [1] CRAN (R 4.4.0) #> jsonlite 1.8.8 2023-12-04 [1] CRAN (R 4.4.0) #> knitr 1.46 2024-04-06 [1] CRAN (R 4.4.0) #> lattice 0.22-6 2024-03-20 [2] CRAN (R 4.4.0) #> lifecycle 1.0.4 2023-11-07 [1] CRAN (R 4.4.0) #> listenv 0.9.1 2024-01-29 [1] CRAN (R 4.4.0) #> magrittr 2.0.3 2022-03-30 [1] CRAN (R 4.4.0) #> Matrix 1.7-0 2024-03-22 [2] CRAN (R 4.4.0) #> parallelly 1.37.1 2024-02-29 [1] CRAN (R 4.4.0) #> pkgconfig 2.0.3 2019-09-22 [1] CRAN (R 4.4.0) #> png 0.1-8 2022-11-29 [1] CRAN (R 4.4.0) #> prettyunits 1.2.0 2023-09-24 [1] CRAN (R 4.4.0) #> processx 3.8.4 2024-03-16 [1] CRAN (R 4.4.0) #> progress 1.2.3 2023-12-06 [1] CRAN (R 4.4.0) #> ps 1.7.6 2024-01-18 [1] CRAN (R 4.4.0) #> purrr 1.0.2 2023-08-10 [1] CRAN (R 4.4.0) #> R.cache 0.16.0 2022-07-21 [1] CRAN (R 4.4.0) #> R.methodsS3 1.8.2 2022-06-13 [1] CRAN (R 4.4.0) #> R.oo 1.26.0 2024-01-24 [1] CRAN (R 4.4.0) #> R.utils 2.12.3 2023-11-18 [1] CRAN (R 4.4.0) #> R6 2.5.1 2021-08-19 [1] CRAN (R 4.4.0) #> Rcpp 1.0.12 2024-01-09 [1] CRAN (R 4.4.0) #> reprex 2.1.0 2024-01-11 [1] CRAN (R 4.4.0) #> reticulate 1.36.1 2024-04-22 [1] CRAN (R 4.4.0) #> rlang 1.1.3 2024-01-10 [1] CRAN (R 4.4.0) #> rmarkdown 2.26 2024-03-05 [1] CRAN (R 4.4.0) #> rstudioapi 0.16.0 2024-03-24 [1] CRAN (R 4.4.0) #> sessioninfo 1.2.2 2021-12-06 [1] CRAN (R 4.4.0) #> styler 1.10.3 2024-04-07 [1] CRAN (R 4.4.0) #> tensorflow 2.16.0 2024-04-15 [1] CRAN (R 4.4.0) #> tfautograph 0.3.2 2021-09-17 [1] CRAN (R 4.4.0) #> tfruns 1.5.3 2024-04-19 [1] CRAN (R 4.4.0) #> vctrs 0.6.5 2023-12-01 [1] CRAN (R 4.4.0) #> whisker 0.4.1 2022-12-05 [1] CRAN (R 4.4.0) #> withr 3.0.0 2024-01-16 [1] CRAN (R 4.4.0) #> xfun 0.43 2024-03-25 [1] CRAN (R 4.4.0) #> yaml 2.3.8 2023-12-11 [1] CRAN (R 4.4.0) #> #> [1] /Users/nick/Library/R/arm64/4.4/library #> [2] /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library #> #> ─ Python configuration ─────────────────────────────────────────────────────── #> python: /Users/nick/Library/r-miniconda-arm64/envs/greta-env-tf2/bin/python #> libpython: /Users/nick/Library/r-miniconda-arm64/envs/greta-env-tf2/lib/libpython3.10.dylib #> pythonhome: /Users/nick/Library/r-miniconda-arm64/envs/greta-env-tf2:/Users/nick/Library/r-miniconda-arm64/envs/greta-env-tf2 #> version: 3.10.14 | packaged by conda-forge | (main, Mar 20 2024, 12:51:49) [Clang 16.0.6 ] #> numpy: /Users/nick/Library/r-miniconda-arm64/envs/greta-env-tf2/lib/python3.10/site-packages/numpy #> numpy_version: 1.26.4 #> tensorflow: /Users/nick/Library/r-miniconda-arm64/envs/greta-env-tf2/lib/python3.10/site-packages/tensorflow #> #> NOTE: Python version was forced by use_python() function #> #> ────────────────────────────────────────────────────────────────────────────── ```

Just noticed this change in tf2 branxh as I was migrating to R 4.4.0 and using TF 2.16.1 TFP 0.24.0

[njtierney]

Issue is to do with using keras version 3 by default, as it uses a new interface to optimisation.

See https://keras.io/api/optimizers/

import keras
from keras import layers

model = keras.Sequential()
model.add(layers.Dense(64, kernel_initializer='uniform', input_shape=(10,)))
model.add(layers.Activation('softmax'))

opt = keras.optimizers.Adam(learning_rate=0.01)
model.compile(loss='categorical_crossentropy', optimizer=opt)

vs

https://keras.io/2.16/api/optimizers/sgd/

>>> opt = tf.keras.optimizers.SGD(learning_rate=0.1)
>>> var = tf.Variable(1.0)
>>> loss = lambda: (var ** 2)/2.0         # d(loss)/d(var1) = var1
>>> opt.minimize(loss, [var])
>>> # Step is `- learning_rate * grad`
>>> var.numpy()
0.9

This can be fixed by using keras version 2, which requires specific versions of TF and TFP to work. These are:

• TFP 0.23.0
• TF 2.15
• Keras 2.15

This means that we will also need to lock in these versions and update the installation instructions for greta. Which should be a separate issue.

In addition, we need to use the legacy optimizers, a la tf$keras$optimizers$legacy$SGD in order for this to work...which are in TF version 2.15 but not 2.16.1

All of this indicates that we will at some point need to update to keras version 3 in the near future.

Also just as a note for Keras 3.0

# doesn't work in K3
self$it <- as.numeric(tfe$tf_optimiser$iterations) + 1

# works in K3
# self$it <- tfe$tf_optimiser$iterations$numpy() + 1