Error when checking input: expected input_1 to have 4 dimensions, but got array with shape (1505280, 1)

[jeroen]

I can no longer reproduce this rmd post by @thomasp85 about lime package. The explain() function I errors with: Error when checking input: expected input_1 to have 4 dimensions, but got array with shape (1505280, 1). Tried updating lime, keras, tensorflow, python, but no luck. Which part of the pipeline could be causing this problem?

I'm using the homebrew version of tensorflow, and keras was installed via keras::install_keras().

> sessionInfo()
R version 3.5.1 (2018-07-02)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS  10.14

Matrix products: default
BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.5/Resources/lib/libRlapack.dylib

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] magick_2.0      lime_0.4.0.9999 keras_2.2.0    

loaded via a namespace (and not attached):
 [1] Rcpp_0.12.19       later_0.7.5        gower_0.1.2        pillar_1.3.0       compiler_3.5.1    
 [6] plyr_1.8.4         bindr_0.1.1        base64enc_0.1-3    iterators_1.0.10   tools_3.5.1       
[11] digest_0.6.17      zeallot_0.1.0      jsonlite_1.5       tibble_1.4.2       gtable_0.2.0      
[16] lattice_0.20-35    pkgconfig_2.0.2    rlang_0.2.2        Matrix_1.2-14      foreach_1.4.4     
[21] shiny_1.1.0        rstudioapi_0.8     parallel_3.5.1     bindrcpp_0.2.2     dplyr_0.7.6       
[26] knitr_1.20         htmlwidgets_1.3    grid_3.5.1         glmnet_2.0-16      tidyselect_0.2.4  
[31] reticulate_1.10    glue_1.3.0         R6_2.3.0           ggplot2_3.0.0      purrr_0.2.5       
[36] magrittr_1.5       whisker_0.3-2      shinythemes_1.1.1  promises_1.0.1     htmltools_0.3.6   
[41] codetools_0.2-15   scales_1.0.0       tfruns_1.4         stringdist_0.9.5.1 assertthat_0.2.0  
[46] xtable_1.8-3       mime_0.6           colorspace_1.3-2   httpuv_1.4.5       tensorflow_1.9    
[51] stringi_1.2.4      lazyeval_0.2.1     munsell_0.5.0      crayon_1.3.4

[skeydan]

This code (copy-pasted from the .Rmd you linked (thx!)) runs fine for me:

library(keras)
library(lime)
library(magick)

model <- application_vgg16(
  weights = "imagenet",
  include_top = TRUE
)
model

img <- image_read('https://www.data-imaginist.com/assets/images/kitten.jpg')
img_path <- file.path(tempdir(), 'kitten.jpg')
image_write(img, img_path)
plot(as.raster(img))

image_prep <- function(x) {
  arrays <- lapply(x, function(path) {
    img <- image_load(path, target_size = c(224,224))
    x <- image_to_array(img)
    x <- array_reshape(x, c(1, dim(x)))
    x <- imagenet_preprocess_input(x)
  })
  do.call(abind::abind, c(arrays, list(along = 1)))
}

model_labels <- readRDS(system.file('extdata', 'imagenet_labels.rds', package = 'lime'))
explainer <- lime(img_path, as_classifier(model, model_labels), image_prep)

explanation <- explain(img_path, explainer, n_labels = 2, n_features = 20)
plot_image_explanation(explanation)

My sessionInfo, for comparison:

R version 3.5.1 (2018-07-02)
Platform: x86_64-redhat-linux-gnu (64-bit)
Running under: Fedora 28 (Workstation Edition)

Matrix products: default
BLAS/LAPACK: /usr/lib64/R/lib/libRblas.so

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C               LC_TIME=en_US.UTF-8       
 [4] LC_COLLATE=en_US.UTF-8     LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                  LC_ADDRESS=C              
[10] LC_TELEPHONE=C             LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] magick_2.0       lime_0.4.0       keras_2.2.0.9001

loaded via a namespace (and not attached):
 [1] Rcpp_0.12.19           later_0.7.5            gower_0.1.2           
 [4] pillar_1.3.0           compiler_3.5.1         plyr_1.8.4            
 [7] bindr_0.1.1            base64enc_0.1-3        iterators_1.0.10      
[10] tools_3.5.1            digest_0.6.18          zeallot_0.1.0         
[13] jsonlite_1.5           tibble_1.4.2           gtable_0.2.0          
[16] lattice_0.20-35        pkgconfig_2.0.2        rlang_0.2.2           
[19] Matrix_1.2-14          foreach_1.4.4          shiny_1.1.0           
[22] rstudioapi_0.8         curl_3.2               parallel_3.5.1        
[25] bindrcpp_0.2.2         dplyr_0.7.6            htmlwidgets_1.3       
[28] grid_3.5.1             glmnet_2.0-16          tidyselect_0.2.4      
[31] reticulate_1.10.0.9002 glue_1.3.0             R6_2.3.0              
[34] ggplot2_3.0.0.9000     purrr_0.2.5            magrittr_1.5          
[37] whisker_0.3-2          shinythemes_1.1.1      promises_1.0.1        
[40] htmltools_0.3.6        codetools_0.2-15       scales_1.0.0          
[43] tfruns_1.4             abind_1.4-5            stringdist_0.9.5.1    
[46] assertthat_0.2.0       xtable_1.8-3           mime_0.6              
[49] colorspace_1.3-2       httpuv_1.4.5           labeling_0.3          
[52] tensorflow_1.9         stringi_1.2.4          lazyeval_0.2.1        
[55] munsell_0.5.0          crayon_1.3.4          

Does this alone (prediction, not involving lime) run for you?

res <- predict(model, image_prep(img_path))

[jeroen]

Yes, predict() works fine:

[skeydan]

Okay, so I can reproduce with the current lime version from github.
With lime from CRAN, it works.

So I guess I'd ask you to open an issue there :-)

[jeroen]

Hmm OK you might be right. I downgraded lime and now it no longer shows that error, but instead if freezes at explain (or at least I've been waiting for 30min...).

[skeydan]

It takes long for me too (5-10 mins?) but it eventually returns (and I can plot it)

[jeroen]

OK thanks! Will move the discussion to --> thomasp85/lime#127