Draft 0.2.1.9002

DESCRIPTION

@@ -1,7 +1,7 @@
 Type: Package
 Package: aifeducation
 Title: Artificial Intelligence for Education
-Version: 0.2.1.9001
+Version: 0.2.1.9002
 Authors@R: c(
     person("Berding", "Florian", , "florian.berding@uni-hamburg.de", role = c("aut", "cre"),
            comment = c(ORCID = "0000-0002-3593-1695")),

NAMESPACE

@@ -3,6 +3,7 @@
 export(EmbeddedText)
 export(TextEmbeddingClassifierNeuralNet)
 export(TextEmbeddingModel)
+export(aifeducation_config)
 export(array_to_matrix)
 export(bow_pp_create_basic_text_rep)
 export(bow_pp_create_vocab_draft)

NEWS.md

@@ -1,4 +1,4 @@
-# aifeducation (Development Version)
+# aifeducation 0.2.1.9002
 
 -   Added DeBERTa and Funnel-Transformer support
 -   Fixed issues in training transformer models

R/install_and_config.R

@@ -5,6 +5,9 @@
 #'@param envname \code{string} Name of the environment where the packages should
 #'be installed.
 #'@param tf_version \code{string} determining the desired version of 'tensorflow'.
+#'@param remove_first \code{bool} If \code{TRUE} removes the environment completely before
+#'recreating the environment and installing the packages. If \code{FALSE} the packages
+#'are installed in the existing environment without any prior changes.
 #'@return Returns no values or objects. Function is used for installing the
 #'necessary python libraries in a conda environment.
 #'@importFrom reticulate conda_create
@@ -13,24 +16,26 @@
 #'@family Installation and Configuration
 #'@export
 install_py_modules<-function(envname="aifeducation",
-                             tf_version="<=2.14"){
+                             tf_version="<=2.14",
+                             remove_first=FALSE){
   relevant_modules<-c("transformers",
                       "tokenizers",
                       "datasets",
                       "codecarbon",
                       "accelerate")
 
-  conda_environments<-reticulate::conda_list()
+  if(remove_first==TRUE){
+    conda_environments<-reticulate::conda_list()
+      if((envname %in% conda_environments$name)==TRUE){
+      reticulate::conda_remove(envname = envname)
+    }
 
-  if((envname %in% conda_environments$name)==TRUE){
-    reticulate::conda_remove(envname = envname)
+    reticulate::conda_create(
+      envname = envname,
+      channel=c("conda-forge")
+    )
   }
 
-  reticulate::conda_create(
-    envname = envname,
-    channel=c("conda-forge")
-  )
-
   reticulate::conda_install(
     packages = c(
       paste0("tensorflow",tf_version),
@@ -180,8 +185,6 @@ set_config_os_environ_logger<-function(level="ERROR"){
 #'at least version 3. If you have an older version 'tensorflow' is used.
 #'
 #'@family Installation and Configuration
-#'@keywords internal
-#'
 AifeducationConfiguration<-R6::R6Class(
   classname = "aifeducationConfiguration",
   private = list(
@@ -210,23 +213,27 @@ AifeducationConfiguration<-R6::R6Class(
         stop("backend must be 'tensorflow' or 'pytorch'.")
       }
 
+
+      py_package_list<-reticulate::py_list_packages()
+      keras_version<-as.character(py_package_list[which(py_package_list$package=="keras"),"version"])
+
       if(private$TextEmbeddingFramework=="not_specified"){
-        if(utils::compareVersion(keras["__version__"],"2.4.0")>=0 &
-           utils::compareVersion(keras["__version__"],"3.0.0")<0){
+        if(utils::compareVersion(keras_version,"2.4.0")>=0 &
+           utils::compareVersion(keras_version,"3.0.0")<0){
           private$TextEmbeddingFramework=backend
           private$ClassifierFramework="tensorflow"
           os$environ$setdefault("KERAS_BACKEND","tensorflow")
 
-          cat("keras Version:",keras["__version__"],"\n")
+          cat("keras Version:",keras_version,"\n")
           cat("Backend for TextEmbeddingModels:",private$TextEmbeddingFramework,"\n")
           cat("Backend for Classifiers:",private$ClassifierFramework,"\n")
 
-        } else if(utils::compareVersion(keras["__version__"],"3.0.0")>=0){
+        } else if(utils::compareVersion(keras_version,"3.0.0")>=0){
           private$TextEmbeddingFramework=backend
           private$ClassifierFramework=backend
           os$environ$setdefault("KERAS_BACKEND",backend)
 
-          cat("keras Version:",keras["__version__"],"\n")
+          cat("keras Version:",keras_version,"\n")
           cat("Backend for TextEmbeddingModels:",private$TextEmbeddingFramework,"\n")
           cat("Backend for Classifiers:",private$ClassifierFramework,"\n")
 
@@ -251,3 +258,11 @@ AifeducationConfiguration<-R6::R6Class(
     }
   )
 )
+
+#' R6 object of class AifeducationConfiguration
+#'
+#' Object for managing setting the machine learning framework of a session.
+#'
+#'@family Installation and Configuration
+#'@export
+aifeducation_config<-NULL

R/onLoad.R

@@ -7,6 +7,7 @@ codecarbon<-NULL
 torch<-NULL
 os<-NULL
 keras<-NULL
+
 aifeducation_config<-NULL
 
 

R/te_classifier_neuralnet_model.R

@@ -1717,6 +1717,19 @@ TextEmbeddingClassifierNeuralNet<-R6::R6Class(
       if(sustain_track==TRUE){
         sustainability_tracker$stop()
         private$sustainability<-summarize_tracked_sustainability(sustainability_tracker)
+      } else {
+        private$sustainability=list(
+          sustainability_tracked=FALSE,
+          date=NA,
+          sustainability_data=list(
+            duration_sec=NA,
+            co2eq_kg=NA,
+            cpu_energy_kwh=NA,
+            gpu_energy_kwh=NA,
+            ram_energy_kwh=NA,
+            total_energy_kwh=NA
+          )
+        )
       }
 
       if(trace==TRUE){

R/text_embedding_model.R

@@ -126,7 +126,7 @@ TextEmbeddingModel<-R6::R6Class(
     #'@param model_dir \code{string} path to the directory where the
     #'BERT model is stored.
     #'@param bow_basic_text_rep object of class \code{basic_text_rep} created via
-    #'the function \link{bow_pp_create_basic_text_rep}. Only relevant for \code{method="glove"}
+    #'the function \link{bow_pp_create_basic_text_rep}. Only relevant for \code{method="glove_cluster"}
     #'and \code{method="lda"}.
     #'@param bow_n_dim \code{int} Number of dimensions of the GlobalVector or
     #'number of topics for LDA.
@@ -170,7 +170,7 @@ TextEmbeddingModel<-R6::R6Class(
                         model_version=NULL,
                         model_language=NULL,
                         method=NULL,
-                        ml_framework="tensorflow",
+                        ml_framework=aifeducation_config$get_framework()$TextEmbeddingFramework,
                         max_length=0,
                         chunks=1,
                         overlap=0,
@@ -229,6 +229,18 @@ TextEmbeddingModel<-R6::R6Class(
       private$r_package_versions$aifeducation<-packageVersion("aifeducation")
       private$r_package_versions$reticulate<-packageVersion("reticulate")
 
+      #Load Sustainability Data-----------------------------------------------
+      if(!(method %in% c("lda","glove_cluster"))==TRUE){
+        sustainability_datalog_path=paste0(model_dir,"/","sustainability.csv")
+        if(file.exists(sustainability_datalog_path)){
+          tmp_sustainability_data<-read.csv(sustainability_datalog_path)
+          private$sustainability$sustainability_tracked=TRUE
+          private$sustainability$track_log=tmp_sustainability_data
+        } else {
+          private$sustainability$sustainability_tracked=FALSE
+          private$sustainability$track_log=NA
+        }
+      }
 
       #basic_components-------------------------------------------------------
       private$basic_components$method=method

README.Rmd

@@ -44,14 +44,16 @@ educational and social sciences.
     social and educational sciences.
 -   Supporting both 'PyTorch' and 'Tensorflow' as machine learning
     frameworks.
--   Supporting pre-trained language models from Hugging Face.
 -   Supporting the usage of trained models on both frameworks providing
-    a high level of flexibility (supported models are BERT, RoBERTa,
-    DeBERTa, Longformer, and Funnel Transformer).
+    a high level of flexibility
+-   Supporting pre-trained language models from Hugging Face.
+-   Supporting BERT, RoBERTa, DeBERTa, Longformer, and Funnel Transformer for
+    creating context sensitive text embedding. 
 -   Making sharing pre-trained models very easy.
 -   Integrating sustainability tracking.
 -   Integrating special statistical techniques for dealing with data
     structures common in the social and educational sciences.
+-   Supporting the classification of long textual documents.
 
 Currently, the package focuses on classification tasks which can either
 be used to diagnose characteristics of learners from written material or
@@ -70,11 +72,18 @@ You can install the development version of aifeducation from
 [GitHub](https://github.com/) with:
 
 ``` r
-# install.packages("devtools")
-devtools::install_github("FBerding/aifeducation",
-                          dependencies = TRUE)
+install.packages("devtools")
+devtools::install_github(repo="FBerding/aifeducation",
+                         ref="master",
+                         dependencies = TRUE)
 ```
 
+> **Note:** There is currently a bug in reticulate 1.32.0 which does not allow
+to load python packages such as tensorflow, keras or torch. The error has been
+fixed in the developer version of reticulate. You can install this version with
+`install.packages("pak")` and `pak::pak("rstudio/reticulate")`.
+With the release of reticulate 1.33.0 you can ignore this note.
+
 ## Sustainability
 
 Training AI models consumes time and energy. To help researchers
@@ -90,7 +99,7 @@ computer is located.
 This package allows all supported models based either on 'PyTorch' or
 'tensorflow', providing a high level of flexibility. Even pre-trained
 models can be used with both frameworks. In general, users only have to
-decide which framework they would like to use when loading a model.
+decide which framework they would like to use when starting a new session.
 
 At the moment, using 'PyTorch' for classifier objects requires 'keras-core'
 or 'keras3'. If your system does not support these python packages you can
@@ -226,13 +235,12 @@ AI/Models](https://fberding.github.io/aifeducation/articles/sharing_and_publishi
 you can find a detailed guide on how to document and share your models.
 
 ## Tutorial and Guides
-
-A guide on how to install and configure this package can be found via
-[Get
-started](https://fberding.github.io/aifeducation/articles/aifeducation.html).
-A short introduction into the package with examples for classification
-tasks can be found in vignette [02 classification
-tasks](https://fberding.github.io/aifeducation/articles/classification_tasks.html).
+- Installation and configuration of the package: 
+  [01 get started](https://fberding.github.io/aifeducation/articles/aifeducation.html).
+- A short introduction into the package with examples for classification
+  tasks:[02 classification tasks](https://fberding.github.io/aifeducation/articles/classification_tasks.html).
+- A description for sharing models: 
+  [03 sharing and using trained AI/models](https://fberding.github.io/aifeducation/articles/sharing_and_publishing.html)
 
 ## References
 

README.md

@@ -29,14 +29,16 @@ educational and social sciences.
   educational sciences.
 - Supporting both ‘PyTorch’ and ‘Tensorflow’ as machine learning
   frameworks.
-- Supporting pre-trained language models from Hugging Face.
 - Supporting the usage of trained models on both frameworks providing a
-  high level of flexibility (supported models are BERT, RoBERTa,
-  DeBERTa, Longformer, and Funnel Transformer).
+  high level of flexibility
+- Supporting pre-trained language models from Hugging Face.
+- Supporting BERT, RoBERTa, DeBERTa, Longformer, and Funnel Transformer
+  for creating context sensitive text embedding.
 - Making sharing pre-trained models very easy.
 - Integrating sustainability tracking.
 - Integrating special statistical techniques for dealing with data
   structures common in the social and educational sciences.
+- Supporting the classification of long textual documents.
 
 Currently, the package focuses on classification tasks which can either
 be used to diagnose characteristics of learners from written material or
@@ -55,11 +57,19 @@ You can install the development version of aifeducation from
 [GitHub](https://github.com/) with:
 
 ``` r
-# install.packages("devtools")
-devtools::install_github("FBerding/aifeducation",
-                          dependencies = TRUE)
+install.packages("devtools")
+devtools::install_github(repo="FBerding/aifeducation",
+                         ref="master",
+                         dependencies = TRUE)
 ```
 
+> **Note:** There is currently a bug in reticulate 1.32.0 which does not
+> allow to load python packages such as tensorflow, keras or torch. The
+> error has been fixed in the developer version of reticulate. You can
+> install this version with `install.packages("pak")` and
+> `pak::pak("rstudio/reticulate")`. With the release of reticulate
+> 1.33.0 you can ignore this note.
+
 ## Sustainability
 
 Training AI models consumes time and energy. To help researchers
@@ -75,7 +85,8 @@ computer is located.
 This package allows all supported models based either on ‘PyTorch’ or
 ‘tensorflow’, providing a high level of flexibility. Even pre-trained
 models can be used with both frameworks. In general, users only have to
-decide which framework they would like to use when loading a model.
+decide which framework they would like to use when starting a new
+session.
 
 At the moment, using ‘PyTorch’ for classifier objects requires
 ‘keras-core’ or ‘keras3’. If your system does not support these python
@@ -210,12 +221,13 @@ you can find a detailed guide on how to document and share your models.
 
 ## Tutorial and Guides
 
-A guide on how to install and configure this package can be found via
-[Get
-started](https://fberding.github.io/aifeducation/articles/aifeducation.html).
-A short introduction into the package with examples for classification
-tasks can be found in vignette [02 classification
-tasks](https://fberding.github.io/aifeducation/articles/classification_tasks.html).
+- Installation and configuration of the package: [01 get
+  started](https://fberding.github.io/aifeducation/articles/aifeducation.html).
+- A short introduction into the package with examples for classification
+  tasks:[02 classification
+  tasks](https://fberding.github.io/aifeducation/articles/classification_tasks.html).
+- A description for sharing models: [03 sharing and using trained
+  AI/models](https://fberding.github.io/aifeducation/articles/sharing_and_publishing.html)
 
 ## References