[SPARKNLP-856] Adding CamemBertForZeroShotClassification to ResourceD…

…ownloader
JohnSnowLabs · Aug 7, 2024 · 2197ad8 · 2197ad8
1 parent 79643e9
commit 2197ad8
Show file tree

Hide file tree

Showing 4 changed files with 7 additions and 29 deletions.
diff --git a/python/sparknlp/annotator/classifier_dl/camembert_for_zero_shot_classification.py b/python/sparknlp/annotator/classifier_dl/camembert_for_zero_shot_classification.py
@@ -34,7 +34,7 @@ class CamemBertForZeroShotClassification(AnnotatorModel,
     >>> sequenceClassifier = CamemBertForZeroShotClassification.pretrained() \\
     ...     .setInputCols(["token", "document"]) \\
     ...     .setOutputCol("label")
-    The default model is ``"deberta_base_zero_shot_classifier_mnli_anli_v3"``, if no name is
+    The default model is ``"camembert_zero_shot_classifier_xnli_onnx"``, if no name is
     provided.
     For available pretrained models please see the `Models Hub
     <https://sparknlp.orgtask=Text+Classification>`__.
@@ -179,14 +179,14 @@ def loadSavedModel(folder, spark_session):
         return CamemBertForZeroShotClassification(java_model=jModel)
 
     @staticmethod
-    def pretrained(name="camembert-base-xnli", lang="fr", remote_loc=None):
+    def pretrained(name="camembert_zero_shot_classifier_xnli_onnx", lang="fr", remote_loc=None):
         """Downloads and loads a pretrained model.
 
         Parameters
         ----------
         name : str, optional
             Name of the pretrained model, by default
-            "camembert_base_sequence_classifier_allocine"
+            "camembert_zero_shot_classifier_xnli_onnx"
         lang : str, optional
             Language of the pretrained model, by default "fr"
         remote_loc : str, optional

diff --git a/...la/com/johnsnowlabs/nlp/annotators/classifier/dl/CamemBertForZeroShotClassification.scala b/...la/com/johnsnowlabs/nlp/annotators/classifier/dl/CamemBertForZeroShotClassification.scala
@@ -284,10 +284,8 @@ class CamemBertForZeroShotClassification(override val uid: String)
 trait ReadPretrainedCamemBertForZeroShotClassification
     extends ParamsAndFeaturesReadable[CamemBertForZeroShotClassification]
     with HasPretrained[CamemBertForZeroShotClassification] {
-  override val defaultModelName: Some[String] = Some(
-    "camembert-zero-shot-classifier-xnli-onnx"
-  )
-  override val defaultLang: String = "en"
+  override val defaultModelName: Some[String] = Some("camembert_zero_shot_classifier_xnli_onnx")
+  override val defaultLang: String = "fr"
 
   override def pretrained(): CamemBertForZeroShotClassification = super.pretrained()
 

diff --git a/src/main/scala/com/johnsnowlabs/nlp/pretrained/ResourceDownloader.scala b/src/main/scala/com/johnsnowlabs/nlp/pretrained/ResourceDownloader.scala
@@ -690,7 +690,8 @@ object PythonResourceDownloader {
     "MPNetForQuestionAnswering" -> MPNetForQuestionAnswering,
     "LLAMA2Transformer" -> LLAMA2Transformer,
     "M2M100Transformer" -> M2M100Transformer,
-    "UAEEmbeddings" -> UAEEmbeddings)
+    "UAEEmbeddings" -> UAEEmbeddings,
+    "CamemBertForZeroShotClassification" -> CamemBertForZeroShotClassification)
 
   // List pairs of types such as the one with key type can load a pretrained model from the value type
   val typeMapper: Map[String, String] = Map("ZeroShotNerModel" -> "RoBertaForQuestionAnswering")

diff --git a/src/main/scala/com/johnsnowlabs/nlp/pretrained/ResourceMetadata.scala b/src/main/scala/com/johnsnowlabs/nlp/pretrained/ResourceMetadata.scala
@@ -109,27 +109,6 @@ object ResourceMetadata {
       candidates: List[ResourceMetadata],
       request: ResourceRequest): Option[ResourceMetadata] = {
 
-    val compatibleCandidatesName = candidates
-      .filter(item =>
-        item.readyToUse && item.libVersion.isDefined && item.sparkVersion.isDefined
-          && item.name == request.name)
-
-    val compatibleCandidatesLanguage = candidates
-      .filter(item =>
-        item.readyToUse && item.libVersion.isDefined && item.sparkVersion.isDefined
-          && item.name == request.name
-          && (request.language.isEmpty || item.language.isEmpty || request.language.get == item.language.get)
-      )
-
-    val compatibleCandidatesSparkNLPVersion =
-      candidates
-        .filter(item =>
-          item.readyToUse && item.libVersion.isDefined && item.sparkVersion.isDefined
-            && item.name == request.name
-            && (request.language.isEmpty || item.language.isEmpty || request.language.get == item.language.get)
-            && Version.isCompatible(request.libVersion, item.libVersion))
-
-    println("")
     val compatibleCandidates = candidates
       .filter(item =>
         item.readyToUse && item.libVersion.isDefined && item.sparkVersion.isDefined