Closed philippmuench closed 6 years ago
skeydan
commented
6 years ago Hi, we added a gist here
https://gist.github.com/jjallaire/92740db8588ce5e62bc8863e487c2134
that shows a way to do it (without making any changes to keras).
Let us know how this works for you.
philippmuench
commented
4 years ago Let us know how this works for you.
@skeydan It works, thanks! However, I am not sure how I can add this to a sequential model? Can you provide a short example?
What I have tried:
model1 <- keras::keras_model_sequential()
model1 %>% layer_multiplicative_lstm(40, input_shape = c(80, 6), return_sequences = TRUE)which leads to
Error in py_call_impl(callable, dots$args, dots$keywords) :
TypeError: The added layer must be an instance of class Layer. Found: <multiplicative_lstm.MultiplicativeLSTM object at 0x18ed50450> and
keras::use_implementation("tensorflow")
model2 <- keras::keras_model_sequential()
model2 %>% layer_multiplicative_lstm(40, input_shape = c(80, 6), return_sequences = TRUE)leads to a different error
Error in py_call_impl(callable, dots$args, dots$keywords) :
ValueError: Layer multiplicative_lstm_8 was called with an input that isn't a symbolic tensor. Received type: <class 'tensorflow.python.keras.engine.sequential.Sequential'>. Full input: [<tensorflow.python.keras.engine.sequential.Sequential object at 0x18ed50490>]. All inputs to the layer should be tensors. However, I can reproduce the example from here
model <- keras_model_sequential()
model %>% layer_dense(units = 256, input_shape = c(784))
model %>% layer_antirectifier()> tensorflow::tf$`__version__`
[1] "2.0.0"
> sessionInfo()
R version 3.6.1 (2019-07-05)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS Catalina 10.15.1
Matrix products: default
BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.6/Resources/lib/libRlapack.dylib
Random number generation:
RNG: Mersenne-Twister
Normal: Inversion
Sample: Rounding
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] deepG_0.2.0
loaded via a namespace (and not attached):
[1] pkgload_1.0.2 bit64_0.9-7 jsonlite_1.6 hdf5r_1.3.0
[5] assertthat_0.2.1 stats4_3.6.1 remotes_2.1.0 slam_0.1-46
[9] sessioninfo_1.1.1 pillar_1.4.2 backports_1.1.5 lattice_0.20-38
[13] glue_1.3.1 reticulate_1.13 digest_0.6.22 XVector_0.26.0
[17] colorspace_1.4-1 Matrix_1.2-17 plyr_1.8.4 tm_0.7-7
[21] pkgconfig_2.0.3 devtools_2.2.1 zlibbioc_1.32.0 purrr_0.3.3
[25] scales_1.0.0 processx_3.4.1 whisker_0.4 Rtsne_0.15
[29] tibble_2.1.3 generics_0.0.2 IRanges_2.20.0 ggplot2_3.2.1
[33] usethis_1.5.1 ellipsis_0.3.0 withr_2.1.2 keras_2.2.5.0
[37] lazyeval_0.2.2 BiocGenerics_0.32.0 NLP_0.2-0 cli_1.1.0
[41] magrittr_1.5 crayon_1.3.4 memoise_1.1.0 ps_1.3.0
[45] tokenizers_0.2.1 fs_1.3.1 SnowballC_0.6.0 xml2_1.2.2
[49] pkgbuild_1.0.6 tools_3.6.1 prettyunits_1.0.2 hms_0.5.2
[53] ArgumentCheck_0.10.2 stringr_1.4.0 S4Vectors_0.24.0 munsell_0.5.0
[57] callr_3.3.2 Biostrings_2.54.0 compiler_3.6.1 rlang_0.4.1
[61] grid_3.6.1 rstudioapi_0.10 base64enc_0.1-3 testthat_2.3.0
[65] gtable_0.3.0 abind_1.4-5 roxygen2_6.1.1 R6_2.4.0
[69] tfruns_1.4 dplyr_0.8.3 tensorflow_2.0.0 bit_1.1-14
[73] zeallot_0.1.0 commonmark_1.7 rprojroot_1.3-2 readr_1.3.1
[77] desc_1.2.0 stringi_1.4.3 parallel_3.6.1 Rcpp_1.0.2
[81] vctrs_0.2.0 tidyselect_0.2.5 xfun_0.10
>
I would like to replace the
layer_lstm()with a multiplicative LSTM implementation (e.g. from here). I'm not sure which files I have to change, it looks like theR/layers-recurrent.Rshould be changed and I have to put themultiplicative_lstm.pysomewhere to the keras folder but I have trouble finding the right locations (on Ubuntu using virtualenv). Maybe someone can point me to the file paths I should change? Or is there a other way to allow such custom modules, e.g. usingreticulate? Thanks!