spark-nlp 6.0.5__py2.py3-none-any.whl → 6.1.0__py2.py3-none-any.whl

{spark_nlp-6.0.5.dist-info → spark_nlp-6.1.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: spark-nlp
-Version: 6.0.5
+Version: 6.1.0
 Summary: John Snow Labs Spark NLP is a natural language processing library built on top of Apache Spark ML. It provides simple, performant & accurate NLP annotations for machine learning pipelines, that scale easily in a distributed environment.
 Home-page: https://github.com/JohnSnowLabs/spark-nlp
 Author: John Snow Labs
@@ -102,7 +102,7 @@ $ java -version
 $ conda create -n sparknlp python=3.7 -y
 $ conda activate sparknlp
 # spark-nlp by default is based on pyspark 3.x
-$ pip install spark-nlp==6.0.5 pyspark==3.3.1
+$ pip install spark-nlp==6.1.0 pyspark==3.3.1
 ```
 
 In Python console or Jupyter `Python3` kernel:
@@ -168,7 +168,7 @@ For a quick example of using pipelines and models take a look at our official [d
 
 ### Apache Spark Support
 
-Spark NLP *6.0.5* has been built on top of Apache Spark 3.4 while fully supports Apache Spark 3.0.x, 3.1.x, 3.2.x, 3.3.x, 3.4.x, and 3.5.x
+Spark NLP *6.1.0* has been built on top of Apache Spark 3.4 while fully supports Apache Spark 3.0.x, 3.1.x, 3.2.x, 3.3.x, 3.4.x, and 3.5.x
 
 | Spark NLP | Apache Spark 3.5.x | Apache Spark 3.4.x | Apache Spark 3.3.x | Apache Spark 3.2.x | Apache Spark 3.1.x | Apache Spark 3.0.x | Apache Spark 2.4.x | Apache Spark 2.3.x |
 |-----------|--------------------|--------------------|--------------------|--------------------|--------------------|--------------------|--------------------|--------------------|
@@ -198,7 +198,7 @@ Find out more about 4.x `SparkNLP` versions in our official [documentation](http
 
 ### Databricks Support
 
-Spark NLP 6.0.5 has been tested and is compatible with the following runtimes:
+Spark NLP 6.1.0 has been tested and is compatible with the following runtimes:
 
 | **CPU**            | **GPU**            |
 |--------------------|--------------------|
@@ -215,7 +215,7 @@ We are compatible with older runtimes. For a full list check databricks support
 
 ### EMR Support
 
-Spark NLP 6.0.5 has been tested and is compatible with the following EMR releases:
+Spark NLP 6.1.0 has been tested and is compatible with the following EMR releases:
 
 | **EMR Release**    |
 |--------------------|

{spark_nlp-6.0.5.dist-info → spark_nlp-6.1.0.dist-info}/RECORD

@@ -3,7 +3,7 @@ com/johnsnowlabs/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,
 com/johnsnowlabs/ml/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 com/johnsnowlabs/ml/ai/__init__.py,sha256=YQiK2M7U4d8y5irPy_HB8ae0mSpqS9583MH44pnKJXc,295
 com/johnsnowlabs/nlp/__init__.py,sha256=DPIVXtONO5xXyOk-HB0-sNiHAcco17NN13zPS_6Uw8c,294
-sparknlp/__init__.py,sha256=peVwWLyO7M5yWlQdRGzOBql6cvaWcdQyy5pNhGR8-sg,13814
+sparknlp/__init__.py,sha256=wxPbTrab8A3tELe8XRaGCfuZ-T8Dc8szbOHXH9ZgLIU,13814
 sparknlp/annotation.py,sha256=I5zOxG5vV2RfPZfqN9enT1i4mo6oBcn3Lrzs37QiOiA,5635
 sparknlp/annotation_audio.py,sha256=iRV_InSVhgvAwSRe9NTbUH9v6OGvTM-FPCpSAKVu0mE,1917
 sparknlp/annotation_image.py,sha256=xhCe8Ko-77XqWVuuYHFrjKqF6zPd8Z-RY_rmZXNwCXU,2547
@@ -167,8 +167,8 @@ sparknlp/annotator/sentence/sentence_detector_dl.py,sha256=-Osj9Bm9KyZRTAWkOsK9c
 sparknlp/annotator/sentiment/__init__.py,sha256=Lq3vKaZS1YATLMg0VNXSVtkWL5q5G9taGBvdrvSwnfg,766
 sparknlp/annotator/sentiment/sentiment_detector.py,sha256=m545NGU0Xzg_PO6_qIfpli1uZj7JQcyFgqe9R6wAPFI,8154
 sparknlp/annotator/sentiment/vivekn_sentiment.py,sha256=4rpXWDgzU6ddnbrSCp9VdLb2epCc9oZ3c6XcqxEw8nk,9655
-sparknlp/annotator/seq2seq/__init__.py,sha256=4h6taLL4pgzs_pR2uFx9AcLrCqtu8bx3hXNeK1_u_EE,1723
-sparknlp/annotator/seq2seq/auto_gguf_model.py,sha256=PYHfljxtSSTL0lBm1-6ZBEHxkBLZvID63HSuxWv7XS4,11554
+sparknlp/annotator/seq2seq/__init__.py,sha256=Aj43G1MuQE0mW7LakCWPjiTkIGl7iHPAnKIwT_DfdIM,1781
+sparknlp/annotator/seq2seq/auto_gguf_model.py,sha256=Oah_RvOy9YrvfnnMRMKOGJHnAMYxo0SeczBZsndM3kY,11638
 sparknlp/annotator/seq2seq/auto_gguf_vision_model.py,sha256=EYrm8EW7AMq3AoIKPe7Gp6ayBlFpWeg76AsAr4nanqU,15346
 sparknlp/annotator/seq2seq/bart_transformer.py,sha256=I1flM4yeCzEAKOdQllBC30XuedxVJ7ferkFhZ6gwEbE,18481
 sparknlp/annotator/seq2seq/cohere_transformer.py,sha256=43LZBVazZMgJRCsN7HaYjVYfJ5hRMV95QZyxMtXq-m4,13496
@@ -183,6 +183,7 @@ sparknlp/annotator/seq2seq/nllb_transformer.py,sha256=1ys01yaC0nVzXETy8oD2wZHyom
 sparknlp/annotator/seq2seq/olmo_transformer.py,sha256=B_zhYkAfYycw5uBq1tVNPmaKuYtpJOxRC6PArit7XiE,13634
 sparknlp/annotator/seq2seq/phi2_transformer.py,sha256=WwKCUOH8qGFv62YF63HjuT7bMVldh06gHvaZH3tbSDk,13787
 sparknlp/annotator/seq2seq/phi3_transformer.py,sha256=arIcw5NDMv3ubBwWz3KYRdLMsspTiEI8vk4s00lyq1c,14293
+sparknlp/annotator/seq2seq/phi4_transformer.py,sha256=iVbsqIzKS2MG-LmA3tljjsjeCUzBqATw1btDBOnFviM,14324
 sparknlp/annotator/seq2seq/qwen_transformer.py,sha256=IYxveoHGWWoiwzJ_VMLMgUBe6jr1JSHKSY0PApnTCOI,14640
 sparknlp/annotator/seq2seq/starcoder_transformer.py,sha256=BTXbSMRpXnDvrfh-6iFS5k6g6EcPV9zBl4U-SSC19wA,14293
 sparknlp/annotator/seq2seq/t5_transformer.py,sha256=wDVxNLluIU1HGZFqaKKc4YTt4l-elPlAtQ7EEa0f5tg,17308
@@ -222,12 +223,12 @@ sparknlp/common/annotator_properties.py,sha256=7B1os7pBUfHo6b7IPQAXQ-nir0u3tQLzD
 sparknlp/common/annotator_type.py,sha256=ash2Ip1IOOiJamPVyy_XQj8Ja_DRHm0b9Vj4Ni75oKM,1225
 sparknlp/common/coverage_result.py,sha256=No4PSh1HSs3PyRI1zC47x65tWgfirqPI290icHQoXEI,823
 sparknlp/common/match_strategy.py,sha256=kt1MUPqU1wCwk5qCdYk6jubHbU-5yfAYxb9jjAOrdnY,1678
-sparknlp/common/properties.py,sha256=v8PUB0YqeiZRzP8mX3kXSFoQVMZOg_ips0Y5M54hUIc,51493
+sparknlp/common/properties.py,sha256=4jDyxr2IGWEuNlGtOoPzqdCF7oLAKGy1z6MtqxUVMug,52704
 sparknlp/common/read_as.py,sha256=imxPGwV7jr4Li_acbo0OAHHRGCBbYv-akzEGaBWEfcY,1226
 sparknlp/common/recursive_annotator_approach.py,sha256=vqugBw22cE3Ff7PIpRlnYFuOlchgL0nM26D8j-NdpqU,1449
 sparknlp/common/storage.py,sha256=D91H3p8EIjNspjqAYu6ephRpCUtdcAir4_PrAbkIQWE,4842
 sparknlp/common/utils.py,sha256=Yne6yYcwKxhOZC-U4qfYoDhWUP_6BIaAjI5X_P_df1E,1306
-sparknlp/internal/__init__.py,sha256=wvC7ovDfII5GiYSwNpA1HHttnlXjbFgpYAGV68NsiQo,40446
+sparknlp/internal/__init__.py,sha256=f5kxLz6bEX8FCMo6lek30MXbn_lDC-dBTI9s_fARblI,40748
 sparknlp/internal/annotator_java_ml.py,sha256=UGPoThG0rGXUOXGSQnDzEDW81Mu1s5RPF29v7DFyE3c,1187
 sparknlp/internal/annotator_transformer.py,sha256=fXmc2IWXGybqZpbEU9obmbdBYPc798y42zvSB4tqV9U,1448
 sparknlp/internal/extended_java_wrapper.py,sha256=hwP0133-hDiDf5sBF-P3MtUsuuDj1PpQbtGZQIRwzfk,2240
@@ -246,7 +247,8 @@ sparknlp/pretrained/utils.py,sha256=T1MrvW_DaWk_jcOjVLOea0NMFE9w8fe0ZT_5urZ_nEY,
 sparknlp/reader/__init__.py,sha256=-Toj3AIBki-zXPpV8ezFTI2LX1yP_rK2bhpoa8nBkTw,685
 sparknlp/reader/enums.py,sha256=MNGug9oJ1BBLM1Pbske13kAabalDzHa2kucF5xzFpHs,770
 sparknlp/reader/pdf_to_text.py,sha256=eWw-cwjosmcSZ9eHso0F5QQoeGBBnwsOhzhCXXvMjZA,7169
-sparknlp/reader/sparknlp_reader.py,sha256=IG0_wYKT1cIIU3EibzOVBZ-GhvX50mC5meXYv0WsYKs,18524
+sparknlp/reader/reader2doc.py,sha256=xahxkEuNM21mb0-MHQoYLtDF1cbAYrMTRpN1-u5K3ec,6587
+sparknlp/reader/sparknlp_reader.py,sha256=MJs8v_ECYaV1SOabI1L_2MkVYEDVImtwgbYypO7DJSY,20623
 sparknlp/training/__init__.py,sha256=qREi9u-5Vc2VjpL6-XZsyvu5jSEIdIhowW7_kKaqMqo,852
 sparknlp/training/conll.py,sha256=wKBiSTrjc6mjsl7Nyt6B8f4yXsDJkZb-sn8iOjix9cE,6961
 sparknlp/training/conllu.py,sha256=8r3i-tmyrLsyk1DtZ9uo2mMDCWb1yw2Y5W6UsV13MkY,4953
@@ -277,7 +279,7 @@ sparknlp/training/_tf_graph_builders_1x/ner_dl/dataset_encoder.py,sha256=R4yHFN3
 sparknlp/training/_tf_graph_builders_1x/ner_dl/ner_model.py,sha256=EoCSdcIjqQ3wv13MAuuWrKV8wyVBP0SbOEW41omHlR0,23189
 sparknlp/training/_tf_graph_builders_1x/ner_dl/ner_model_saver.py,sha256=k5CQ7gKV6HZbZMB8cKLUJuZxoZWlP_DFWdZ--aIDwsc,2356
 sparknlp/training/_tf_graph_builders_1x/ner_dl/sentence_grouper.py,sha256=pAxjWhjazSX8Vg0MFqJiuRVw1IbnQNSs-8Xp26L4nko,870
-spark_nlp-6.0.5.dist-info/METADATA,sha256=BL1PeMYps-L3LAkmNpwxMkrGUw_KwO164VZ5AoqDZLg,19722
-spark_nlp-6.0.5.dist-info/WHEEL,sha256=JNWh1Fm1UdwIQV075glCn4MVuCRs0sotJIq-J6rbxCU,109
-spark_nlp-6.0.5.dist-info/top_level.txt,sha256=uuytur4pyMRw2H_txNY2ZkaucZHUs22QF8-R03ch_-E,13
-spark_nlp-6.0.5.dist-info/RECORD,,
+spark_nlp-6.1.0.dist-info/METADATA,sha256=MDLwobOveRxQL45CWF-NY26iHa3a7PijF9wntBXpeZE,19722
+spark_nlp-6.1.0.dist-info/WHEEL,sha256=JNWh1Fm1UdwIQV075glCn4MVuCRs0sotJIq-J6rbxCU,109
+spark_nlp-6.1.0.dist-info/top_level.txt,sha256=uuytur4pyMRw2H_txNY2ZkaucZHUs22QF8-R03ch_-E,13
+spark_nlp-6.1.0.dist-info/RECORD,,

sparknlp/__init__.py

@@ -66,7 +66,7 @@ sys.modules['com.johnsnowlabs.ml.ai'] = annotator
 annotators = annotator
 embeddings = annotator
 
-__version__ = "6.0.5"
+__version__ = "6.1.0"
 
 
 def start(gpu=False,

sparknlp/annotator/seq2seq/__init__.py

@@ -31,3 +31,4 @@ from sparknlp.annotator.seq2seq.starcoder_transformer import *
 from sparknlp.annotator.seq2seq.llama3_transformer import *
 from sparknlp.annotator.seq2seq.cohere_transformer import *
 from sparknlp.annotator.seq2seq.olmo_transformer import *
+from sparknlp.annotator.seq2seq.phi4_transformer import *

sparknlp/annotator/seq2seq/auto_gguf_model.py

@@ -253,7 +253,9 @@ class AutoGGUFModel(AnnotatorModel, HasBatchedAnnotate, HasLlamaCppProperties):
             nCtx=4096,
             nBatch=512,
             embedding=False,
-            nPredict=100
+            nPredict=100,
+            nGpuLayers=99,
+            systemPrompt="You are a helpful assistant."
         )
 
     @staticmethod

sparknlp/annotator/seq2seq/phi4_transformer.py

@@ -0,0 +1,387 @@
+#  Copyright 2017-2024 John Snow Labs
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+"""Contains classes for the Phi4Transformer."""
+
+from sparknlp.common import *
+
+class Phi4Transformer(AnnotatorModel, HasBatchedAnnotate, HasEngine):
+    """Phi-4: State-of-the-art open model by Microsoft Research
+    
+    phi-4 is a 14B parameter, dense decoder-only Transformer model trained on 9.8T tokens, designed for advanced reasoning, code, and general NLP tasks.
+    For more details, see: https://huggingface.co/microsoft/phi-4
+
+    Model Overview
+    --------------
+    - 14B parameters, dense decoder-only Transformer
+    - 16K context length
+    - Trained on 9.8T tokens (synthetic, public domain, academic, Q&A, code)
+    - Focus on high-quality, advanced reasoning, math, code, and general NLP
+    - Multilingual data: ~8% (primarily English)
+    - Released under MIT License
+
+    Intended Use
+    ------------
+    - General-purpose AI, research, and generative features
+    - Memory/compute constrained and latency-bound environments
+    - Reasoning, logic, and code generation
+
+    Benchmarks
+    ----------
+    - MMLU: 84.8 | HumanEval: 82.6 | GPQA: 56.1 | DROP: 75.5 | MATH: 80.6
+    - Outperforms or matches other 14B/70B models on many tasks
+
+    Safety & Limitations
+    -------------------
+    - Safety alignment via SFT and DPO, red-teamed by Microsoft AIRT
+    - Not intended for high-risk or consequential domains without further assessment
+    - Primarily English; other languages may have reduced performance
+    - May generate inaccurate, offensive, or biased content; use with care
+
+    Usage
+    -----
+    Pretrained models can be loaded with :meth:`.pretrained` of the companion object:
+
+    >>> phi4 = Phi4Transformer.pretrained() \
+    ...     .setInputCols(["document"]) \
+    ...     .setOutputCol("generation")
+
+    The default model is ``"phi-4"``, if no name is provided. For available pretrained models please see the `Models Hub <https://huggingface.co/microsoft/phi-4>`__.
+
+    Note
+    ----
+    This is a resource-intensive module, especially with larger models and sequences. Use of accelerators such as GPUs is strongly recommended.
+
+    References
+    ----------
+    - https://huggingface.co/microsoft/phi-4
+    - arXiv:2412.08905
+
+    ====================== ======================
+    Input Annotation types Output Annotation type
+    ====================== ======================
+    ``DOCUMENT``           ``DOCUMENT``
+    ====================== ======================
+    
+    Parameters
+    ----------
+    configProtoBytes
+        ConfigProto from tensorflow, serialized into byte array.
+    minOutputLength
+        Minimum length of the sequence to be generated, by default 0
+    maxOutputLength
+        Maximum length of output text, by default 60
+    doSample
+        Whether or not to use sampling; use greedy decoding otherwise, by default False
+    temperature
+        The value used to modulate the next token probabilities, by default 1.0
+    topK
+        The number of highest probability vocabulary tokens to keep for
+        top-k-filtering, by default 40
+    topP
+        Top cumulative probability for vocabulary tokens, by default 1.0
+
+        If set to float < 1, only the most probable tokens with probabilities
+        that add up to ``topP`` or higher are kept for generation.
+    repetitionPenalty
+        The parameter for repetition penalty, 1.0 means no penalty. , by default
+        1.0
+    noRepeatNgramSize
+        If set to int > 0, all ngrams of that size can only occur once, by
+        default 0
+    ignoreTokenIds
+        A list of token ids which are ignored in the decoder's output, by
+        default []
+
+    Notes
+    -----
+    This is a very computationally expensive module, especially on larger
+    sequences. The use of an accelerator such as GPU is recommended.
+
+    Examples
+    --------
+    >>> import sparknlp
+    >>> from sparknlp.base import *
+    >>> from sparknlp.annotator import *
+    >>> from pyspark.ml import Pipeline
+    >>> documentAssembler = DocumentAssembler() \
+    ...     .setInputCol("text") \
+    ...     .setOutputCol("documents")
+    >>> phi4 = Phi4Transformer.pretrained("phi-4") \
+    ...     .setInputCols(["documents"]) \
+    ...     .setMaxOutputLength(60) \
+    ...     .setOutputCol("generation")
+    >>> pipeline = Pipeline().setStages([documentAssembler, phi4])
+    >>> data = spark.createDataFrame([
+    ...     (
+    ...         1,
+    ...         "<|start_header_id|>system<|end_header_id|> \\n"+ \
+    ...         "You are a helpful assistant! \\n" + \
+    ...         "<|start_header_id|>user<|end_header_id|> \\n" + \
+    ...         "What is Phi-4? \\n" + \
+    ...         "<|start_header_id|>assistant<|end_header_id|> \\n"
+    ...         )
+    ... ]).toDF("id", "text")
+    >>> result = pipeline.fit(data).transform(data)
+    >>> result.select("generation.result").show(truncate=False)
+    +------------------------------------------------+
+    |result                                          |
+    +------------------------------------------------+
+    |[Phi-4 is a 14B parameter, dense decoder-only Transformer model developed by Microsoft Research for advanced reasoning, code, and general NLP tasks.]|
+    +------------------------------------------------+
+    """
+
+    name = "Phi4Transformer"
+
+    inputAnnotatorTypes = [AnnotatorType.DOCUMENT]
+
+    outputAnnotatorType = AnnotatorType.DOCUMENT
+
+    configProtoBytes = Param(Params._dummy(),
+                             "configProtoBytes",
+                             "ConfigProto from tensorflow, serialized into byte array. Get with config_proto.SerializeToString()",
+                             TypeConverters.toListInt)
+
+    minOutputLength = Param(Params._dummy(), "minOutputLength", "Minimum length of the sequence to be generated",
+                            typeConverter=TypeConverters.toInt)
+
+    maxOutputLength = Param(Params._dummy(), "maxOutputLength", "Maximum length of output text",
+                            typeConverter=TypeConverters.toInt)
+
+    doSample = Param(Params._dummy(), "doSample", "Whether or not to use sampling; use greedy decoding otherwise",
+                     typeConverter=TypeConverters.toBoolean)
+
+    temperature = Param(Params._dummy(), "temperature", "The value used to module the next token probabilities",
+                        typeConverter=TypeConverters.toFloat)
+
+    topK = Param(Params._dummy(), "topK",
+                 "The number of highest probability vocabulary tokens to keep for top-k-filtering",
+                 typeConverter=TypeConverters.toInt)
+
+    topP = Param(Params._dummy(), "topP",
+                 "If set to float < 1, only the most probable tokens with probabilities that add up to ``top_p`` or higher are kept for generation",
+                 typeConverter=TypeConverters.toFloat)
+
+    repetitionPenalty = Param(Params._dummy(), "repetitionPenalty",
+                              "The parameter for repetition penalty. 1.0 means no penalty. See `this paper <https://arxiv.org/pdf/1909.05858.pdf>`__ for more details",
+                              typeConverter=TypeConverters.toFloat)
+
+    noRepeatNgramSize = Param(Params._dummy(), "noRepeatNgramSize",
+                              "If set to int > 0, all ngrams of that size can only occur once",
+                              typeConverter=TypeConverters.toInt)
+
+    ignoreTokenIds = Param(Params._dummy(), "ignoreTokenIds",
+                           "A list of token ids which are ignored in the decoder's output",
+                           typeConverter=TypeConverters.toListInt)
+
+    beamSize = Param(Params._dummy(), "beamSize",
+                 "The number of beams to use for beam search",
+                 typeConverter=TypeConverters.toInt)
+
+    stopTokenIds = Param(Params._dummy(), "stopTokenIds",
+                         "A list of token ids which are considered as stop tokens in the decoder's output",
+                            typeConverter=TypeConverters.toListInt)
+
+    def setIgnoreTokenIds(self, value):
+        """A list of token ids which are ignored in the decoder's output.
+
+        Parameters
+        ----------
+        value : List[int]
+            The words to be filtered out
+        """
+        return self._set(ignoreTokenIds=value)
+
+    def setConfigProtoBytes(self, b):
+        """Sets configProto from tensorflow, serialized into byte array.
+
+        Parameters
+        ----------
+        b : List[int]
+            ConfigProto from tensorflow, serialized into byte array
+        """
+        return self._set(configProtoBytes=b)
+
+    def setMinOutputLength(self, value):
+        """Sets minimum length of the sequence to be generated.
+
+        Parameters
+        ----------
+        value : int
+            Minimum length of the sequence to be generated
+        """
+        return self._set(minOutputLength=value)
+
+    def setMaxOutputLength(self, value):
+        """Sets maximum length of output text.
+
+        Parameters
+        ----------
+        value : int
+            Maximum length of output text
+        """
+        return self._set(maxOutputLength=value)
+
+    def setDoSample(self, value):
+        """Sets whether or not to use sampling, use greedy decoding otherwise.
+
+        Parameters
+        ----------
+        value : bool
+            Whether or not to use sampling; use greedy decoding otherwise
+        """
+        return self._set(doSample=value)
+
+    def setTemperature(self, value):
+        """Sets the value used to module the next token probabilities.
+
+        Parameters
+        ----------
+        value : float
+            The value used to module the next token probabilities
+        """
+        return self._set(temperature=value)
+
+    def setTopK(self, value):
+        """Sets the number of highest probability vocabulary tokens to keep for
+        top-k-filtering.
+
+        Parameters
+        ----------
+        value : int
+            Number of highest probability vocabulary tokens to keep
+        """
+        return self._set(topK=value)
+
+    def setTopP(self, value):
+        """Sets the top cumulative probability for vocabulary tokens.
+
+        If set to float < 1, only the most probable tokens with probabilities
+        that add up to ``topP`` or higher are kept for generation.
+
+        Parameters
+        ----------
+        value : float
+            Cumulative probability for vocabulary tokens
+        """
+        return self._set(topP=value)
+
+    def setRepetitionPenalty(self, value):
+        """Sets the parameter for repetition penalty. 1.0 means no penalty.
+
+        Parameters
+        ----------
+        value : float
+            The repetition penalty
+
+        References
+        ----------
+        See `Ctrl: A Conditional Transformer Language Model For Controllable
+        Generation <https://arxiv.org/pdf/1909.05858.pdf>`__ for more details.
+        """
+        return self._set(repetitionPenalty=value)
+
+    def setNoRepeatNgramSize(self, value):
+        """Sets size of n-grams that can only occur once.
+
+        If set to int > 0, all ngrams of that size can only occur once.
+
+        Parameters
+        ----------
+        value : int
+            N-gram size can only occur once
+        """
+        return self._set(noRepeatNgramSize=value)
+
+    def setBeamSize(self, value):
+        """Sets the number of beams to use for beam search.
+
+        Parameters
+        ----------
+        value : int
+            The number of beams to use for beam search
+        """
+        return self._set(beamSize=value)
+
+    def setStopTokenIds(self, value):
+        """Sets a list of token ids which are considered as stop tokens in the decoder's output.
+
+        Parameters
+        ----------
+        value : List[int]
+            The words to be considered as stop tokens
+        """
+        return self._set(stopTokenIds=value)
+
+    @keyword_only
+    def __init__(self, classname="com.johnsnowlabs.nlp.annotators.seq2seq.Phi4Transformer", java_model=None):
+        super(Phi4Transformer, self).__init__(
+            classname=classname,
+            java_model=java_model
+        )
+        self._setDefault(
+            minOutputLength=0,
+            maxOutputLength=20,
+            doSample=False,
+            temperature=0.6,
+            topK=-1,
+            topP=0.9,
+            repetitionPenalty=1.0,
+            noRepeatNgramSize=3,
+            ignoreTokenIds=[],
+            batchSize=1,
+            beamSize=1,
+            stopTokenIds=[128001,]
+        )
+
+    @staticmethod
+    def loadSavedModel(folder, spark_session, use_openvino = False):
+        """Loads a locally saved model.
+
+        Parameters
+        ----------
+        folder : str
+            Folder of the saved model
+        spark_session : pyspark.sql.SparkSession
+            The current SparkSession
+
+        Returns
+        -------
+        Phi4Transformer
+            The restored model
+        """
+        from sparknlp.internal import _Phi4Loader
+        jModel = _Phi4Loader(folder, spark_session._jsparkSession, use_openvino)._java_obj
+        return Phi4Transformer(java_model=jModel)
+
+    @staticmethod
+    def pretrained(name="phi-4", lang="en", remote_loc=None):
+        """Downloads and loads a pretrained model.
+
+        Parameters
+        ----------
+        name : str, optional
+            Name of the pretrained model, by default "phi-4"
+        lang : str, optional
+            Language of the pretrained model, by default "en"
+        remote_loc : str, optional
+            Optional remote address of the resource, by default None. Will use
+            Spark NLPs repositories otherwise.
+
+        Returns
+        -------
+        Phi4Transformer
+            The restored model
+        """
+        from sparknlp.pretrained import ResourceDownloader
+        return ResourceDownloader.downloadModel(Phi4Transformer, name, lang, remote_loc) 

sparknlp/common/properties.py

@@ -765,14 +765,14 @@ class HasLlamaCppProperties:
     # -------- MODEl PARAMETERS --------
     nThreads = Param(Params._dummy(), "nThreads", "Set the number of threads to use during generation",
                      typeConverter=TypeConverters.toInt)
-    nThreadsDraft = Param(Params._dummy(), "nThreadsDraft", "Set the number of threads to use during draft generation",
-                          typeConverter=TypeConverters.toInt)
+    # nThreadsDraft = Param(Params._dummy(), "nThreadsDraft", "Set the number of threads to use during draft generation",
+    #                       typeConverter=TypeConverters.toInt)
     nThreadsBatch = Param(Params._dummy(), "nThreadsBatch",
                           "Set the number of threads to use during batch and prompt processing",
                           typeConverter=TypeConverters.toInt)
-    nThreadsBatchDraft = Param(Params._dummy(), "nThreadsBatchDraft",
-                               "Set the number of threads to use during batch and prompt processing",
-                               typeConverter=TypeConverters.toInt)
+    # nThreadsBatchDraft = Param(Params._dummy(), "nThreadsBatchDraft",
+    #                            "Set the number of threads to use during batch and prompt processing",
+    #                            typeConverter=TypeConverters.toInt)
     nCtx = Param(Params._dummy(), "nCtx", "Set the size of the prompt context", typeConverter=TypeConverters.toInt)
     nBatch = Param(Params._dummy(), "nBatch",
                    "Set the logical batch size for prompt processing (must be >=32 to use BLAS)",
@@ -782,12 +782,12 @@ class HasLlamaCppProperties:
                     typeConverter=TypeConverters.toInt)
     nDraft = Param(Params._dummy(), "nDraft", "Set the number of tokens to draft for speculative decoding",
                    typeConverter=TypeConverters.toInt)
-    nChunks = Param(Params._dummy(), "nChunks", "Set the maximal number of chunks to process",
-                    typeConverter=TypeConverters.toInt)
-    nSequences = Param(Params._dummy(), "nSequences", "Set the number of sequences to decode",
-                       typeConverter=TypeConverters.toInt)
-    pSplit = Param(Params._dummy(), "pSplit", "Set the speculative decoding split probability",
-                   typeConverter=TypeConverters.toFloat)
+    # nChunks = Param(Params._dummy(), "nChunks", "Set the maximal number of chunks to process",
+    #                 typeConverter=TypeConverters.toInt)
+    # nSequences = Param(Params._dummy(), "nSequences", "Set the number of sequences to decode",
+    #                    typeConverter=TypeConverters.toInt)
+    # pSplit = Param(Params._dummy(), "pSplit", "Set the speculative decoding split probability",
+    #                typeConverter=TypeConverters.toFloat)
     nGpuLayers = Param(Params._dummy(), "nGpuLayers", "Set the number of layers to store in VRAM (-1 - use default)",
                        typeConverter=TypeConverters.toInt)
     nGpuLayersDraft = Param(Params._dummy(), "nGpuLayersDraft",
@@ -802,10 +802,10 @@ class HasLlamaCppProperties:
                          typeConverter=TypeConverters.toString)
     mainGpu = Param(Params._dummy(), "mainGpu", "Set the main GPU that is used for scratch and small tensors.",
                     typeConverter=TypeConverters.toInt)
-    tensorSplit = Param(Params._dummy(), "tensorSplit", "Set how split tensors should be distributed across GPUs",
-                        typeConverter=TypeConverters.toListFloat)
-    grpAttnN = Param(Params._dummy(), "grpAttnN", "Set the group-attention factor", typeConverter=TypeConverters.toInt)
-    grpAttnW = Param(Params._dummy(), "grpAttnW", "Set the group-attention width", typeConverter=TypeConverters.toInt)
+    # tensorSplit = Param(Params._dummy(), "tensorSplit", "Set how split tensors should be distributed across GPUs",
+    #                     typeConverter=TypeConverters.toListFloat)
+    # grpAttnN = Param(Params._dummy(), "grpAttnN", "Set the group-attention factor", typeConverter=TypeConverters.toInt)
+    # grpAttnW = Param(Params._dummy(), "grpAttnW", "Set the group-attention width", typeConverter=TypeConverters.toInt)
     ropeFreqBase = Param(Params._dummy(), "ropeFreqBase", "Set the RoPE base frequency, used by NTK-aware scaling",
                          typeConverter=TypeConverters.toFloat)
     ropeFreqScale = Param(Params._dummy(), "ropeFreqScale",
@@ -837,7 +837,7 @@ class HasLlamaCppProperties:
                          typeConverter=TypeConverters.toString)
     # Set the RoPE frequency scaling method, defaults to linear unless specified by the model.
     #
-    #   - UNSPECIFIED: Don't use any scaling
+    #   - NONE: Don't use any scaling
     #   - LINEAR: Linear scaling
     #   - YARN: YaRN RoPE scaling
     ropeScalingType = Param(Params._dummy(), "ropeScalingType",
@@ -848,26 +848,28 @@ class HasLlamaCppProperties:
     #   - 0 NONE: Don't use any pooling
     #   - 1 MEAN: Mean Pooling
     #   - 2 CLS: CLS Pooling
+    #   - 3 LAST: Last token pooling
+    #   - 4 RANK: For reranked models
     poolingType = Param(Params._dummy(), "poolingType",
                         "Set the pooling type for embeddings, use model default if unspecified",
                         typeConverter=TypeConverters.toString)
     modelDraft = Param(Params._dummy(), "modelDraft", "Set the draft model for speculative decoding",
                        typeConverter=TypeConverters.toString)
     modelAlias = Param(Params._dummy(), "modelAlias", "Set a model alias", typeConverter=TypeConverters.toString)
-    lookupCacheStaticFilePath = Param(Params._dummy(), "lookupCacheStaticFilePath",
-                                      "Set path to static lookup cache to use for lookup decoding (not updated by generation)",
-                                      typeConverter=TypeConverters.toString)
-    lookupCacheDynamicFilePath = Param(Params._dummy(), "lookupCacheDynamicFilePath",
-                                       "Set path to dynamic lookup cache to use for lookup decoding (updated by generation)",
-                                       typeConverter=TypeConverters.toString)
+    # lookupCacheStaticFilePath = Param(Params._dummy(), "lookupCacheStaticFilePath",
+    #                                   "Set path to static lookup cache to use for lookup decoding (not updated by generation)",
+    #                                   typeConverter=TypeConverters.toString)
+    # lookupCacheDynamicFilePath = Param(Params._dummy(), "lookupCacheDynamicFilePath",
+    #                                    "Set path to dynamic lookup cache to use for lookup decoding (updated by generation)",
+    #                                    typeConverter=TypeConverters.toString)
     # loraAdapters = new StructFeature[Map[String, Float]](this, "loraAdapters")
     embedding = Param(Params._dummy(), "embedding", "Whether to load model with embedding support",
                       typeConverter=TypeConverters.toBoolean)
     flashAttention = Param(Params._dummy(), "flashAttention", "Whether to enable Flash Attention",
                            typeConverter=TypeConverters.toBoolean)
-    inputPrefixBos = Param(Params._dummy(), "inputPrefixBos",
-                           "Whether to add prefix BOS to user inputs, preceding the `--in-prefix` string",
-                           typeConverter=TypeConverters.toBoolean)
+    # inputPrefixBos = Param(Params._dummy(), "inputPrefixBos",
+    #                        "Whether to add prefix BOS to user inputs, preceding the `--in-prefix` string",
+    #                        typeConverter=TypeConverters.toBoolean)
     useMmap = Param(Params._dummy(), "useMmap",
                     "Whether to use memory-map model (faster load but may increase pageouts if not using mlock)",
                     typeConverter=TypeConverters.toBoolean)
@@ -948,17 +950,17 @@ class HasLlamaCppProperties:
         """Set the number of threads to use during generation"""
         return self._set(nThreads=nThreads)
 
-    def setNThreadsDraft(self, nThreadsDraft: int):
-        """Set the number of threads to use during draft generation"""
-        return self._set(nThreadsDraft=nThreadsDraft)
+    # def setNThreadsDraft(self, nThreadsDraft: int):
+    #     """Set the number of threads to use during draft generation"""
+    #     return self._set(nThreadsDraft=nThreadsDraft)
 
     def setNThreadsBatch(self, nThreadsBatch: int):
         """Set the number of threads to use during batch and prompt processing"""
         return self._set(nThreadsBatch=nThreadsBatch)
 
-    def setNThreadsBatchDraft(self, nThreadsBatchDraft: int):
-        """Set the number of threads to use during batch and prompt processing"""
-        return self._set(nThreadsBatchDraft=nThreadsBatchDraft)
+    # def setNThreadsBatchDraft(self, nThreadsBatchDraft: int):
+    #     """Set the number of threads to use during batch and prompt processing"""
+    #     return self._set(nThreadsBatchDraft=nThreadsBatchDraft)
 
     def setNCtx(self, nCtx: int):
         """Set the size of the prompt context"""
@@ -976,17 +978,17 @@ class HasLlamaCppProperties:
         """Set the number of tokens to draft for speculative decoding"""
         return self._set(nDraft=nDraft)
 
-    def setNChunks(self, nChunks: int):
-        """Set the maximal number of chunks to process"""
-        return self._set(nChunks=nChunks)
+    # def setNChunks(self, nChunks: int):
+    #     """Set the maximal number of chunks to process"""
+    #     return self._set(nChunks=nChunks)
 
-    def setNSequences(self, nSequences: int):
-        """Set the number of sequences to decode"""
-        return self._set(nSequences=nSequences)
+    # def setNSequences(self, nSequences: int):
+    #     """Set the number of sequences to decode"""
+    #     return self._set(nSequences=nSequences)
 
-    def setPSplit(self, pSplit: float):
-        """Set the speculative decoding split probability"""
-        return self._set(pSplit=pSplit)
+    # def setPSplit(self, pSplit: float):
+    #     """Set the speculative decoding split probability"""
+    #     return self._set(pSplit=pSplit)
 
     def setNGpuLayers(self, nGpuLayers: int):
         """Set the number of layers to store in VRAM (-1 - use default)"""
@@ -1004,17 +1006,17 @@ class HasLlamaCppProperties:
         """Set the main GPU that is used for scratch and small tensors."""
         return self._set(mainGpu=mainGpu)
 
-    def setTensorSplit(self, tensorSplit: List[float]):
-        """Set how split tensors should be distributed across GPUs"""
-        return self._set(tensorSplit=tensorSplit)
+    # def setTensorSplit(self, tensorSplit: List[float]):
+    #     """Set how split tensors should be distributed across GPUs"""
+    #     return self._set(tensorSplit=tensorSplit)
 
-    def setGrpAttnN(self, grpAttnN: int):
-        """Set the group-attention factor"""
-        return self._set(grpAttnN=grpAttnN)
+    # def setGrpAttnN(self, grpAttnN: int):
+    #     """Set the group-attention factor"""
+    #     return self._set(grpAttnN=grpAttnN)
 
-    def setGrpAttnW(self, grpAttnW: int):
-        """Set the group-attention width"""
-        return self._set(grpAttnW=grpAttnW)
+    # def setGrpAttnW(self, grpAttnW: int):
+    #     """Set the group-attention width"""
+    #     return self._set(grpAttnW=grpAttnW)
 
     def setRopeFreqBase(self, ropeFreqBase: float):
         """Set the RoPE base frequency, used by NTK-aware scaling"""
@@ -1049,7 +1051,16 @@ class HasLlamaCppProperties:
         return self._set(defragmentationThreshold=defragmentationThreshold)
 
     def setNumaStrategy(self, numaStrategy: str):
-        """Set optimization strategies that help on some NUMA systems (if available)"""
+        """Set optimization strategies that help on some NUMA systems (if available)
+
+        Possible values:
+
+        - DISABLED: No NUMA optimizations
+        - DISTRIBUTE: spread execution evenly over all
+        - ISOLATE: only spawn threads on CPUs on the node that execution started on
+        - NUMA_CTL: use the CPU map provided by numactl
+        - MIRROR: Mirrors the model across NUMA nodes
+        """
         numaUpper = numaStrategy.upper()
         numaStrategies = ["DISABLED", "DISTRIBUTE", "ISOLATE", "NUMA_CTL", "MIRROR"]
         if numaUpper not in numaStrategies:
@@ -1060,13 +1071,36 @@ class HasLlamaCppProperties:
         return self._set(numaStrategy=numaStrategy)
 
     def setRopeScalingType(self, ropeScalingType: str):
-        """Set the RoPE frequency scaling method, defaults to linear unless specified by the model"""
-        return self._set(ropeScalingType=ropeScalingType)
+        """Set the RoPE frequency scaling method, defaults to linear unless specified by the model.
+
+        Possible values:
+
+        - NONE: Don't use any scaling
+        - LINEAR: Linear scaling
+        - YARN: YaRN RoPE scaling
+        """
+        ropeScalingTypeUpper = ropeScalingType.upper()
+        ropeScalingTypes = ["NONE", "LINEAR", "YARN"]
+        if ropeScalingTypeUpper not in ropeScalingTypes:
+           raise ValueError(
+               f"Invalid RoPE scaling type: {ropeScalingType}. "
+               + f"Valid values are: {ropeScalingTypes}"
+           )
+        return self._set(ropeScalingType=ropeScalingTypeUpper)
 
     def setPoolingType(self, poolingType: str):
-        """Set the pooling type for embeddings, use model default if unspecified"""
+        """Set the pooling type for embeddings, use model default if unspecified
+
+        Possible values:
+
+        - 0 NONE: Don't use any pooling
+        - 1 MEAN: Mean Pooling
+        - 2 CLS: CLS Pooling
+        - 3 LAST: Last token pooling
+        - 4 RANK: For reranked models
+        """
         poolingTypeUpper = poolingType.upper()
-        poolingTypes = ["NONE", "MEAN", "CLS", "LAST"]
+        poolingTypes = ["NONE", "MEAN", "CLS", "LAST", "RANK"]
         if poolingTypeUpper not in poolingTypes:
             raise ValueError(
                 f"Invalid pooling type: {poolingType}. "
@@ -1082,13 +1116,13 @@ class HasLlamaCppProperties:
         """Set a model alias"""
         return self._set(modelAlias=modelAlias)
 
-    def setLookupCacheStaticFilePath(self, lookupCacheStaticFilePath: str):
-        """Set path to static lookup cache to use for lookup decoding (not updated by generation)"""
-        return self._set(lookupCacheStaticFilePath=lookupCacheStaticFilePath)
+    # def setLookupCacheStaticFilePath(self, lookupCacheStaticFilePath: str):
+    #     """Set path to static lookup cache to use for lookup decoding (not updated by generation)"""
+    #     return self._set(lookupCacheStaticFilePath=lookupCacheStaticFilePath)
 
-    def setLookupCacheDynamicFilePath(self, lookupCacheDynamicFilePath: str):
-        """Set path to dynamic lookup cache to use for lookup decoding (updated by generation)"""
-        return self._set(lookupCacheDynamicFilePath=lookupCacheDynamicFilePath)
+    # def setLookupCacheDynamicFilePath(self, lookupCacheDynamicFilePath: str):
+    #     """Set path to dynamic lookup cache to use for lookup decoding (updated by generation)"""
+    #     return self._set(lookupCacheDynamicFilePath=lookupCacheDynamicFilePath)
 
     def setEmbedding(self, embedding: bool):
         """Whether to load model with embedding support"""
@@ -1098,9 +1132,9 @@ class HasLlamaCppProperties:
         """Whether to enable Flash Attention"""
         return self._set(flashAttention=flashAttention)
 
-    def setInputPrefixBos(self, inputPrefixBos: bool):
-        """Whether to add prefix BOS to user inputs, preceding the `--in-prefix` string"""
-        return self._set(inputPrefixBos=inputPrefixBos)
+    # def setInputPrefixBos(self, inputPrefixBos: bool):
+    #     """Whether to add prefix BOS to user inputs, preceding the `--in-prefix` string"""
+    #     return self._set(inputPrefixBos=inputPrefixBos)
 
     def setUseMmap(self, useMmap: bool):
         """Whether to use memory-map model (faster load but may increase pageouts if not using mlock)"""
@@ -1260,9 +1294,9 @@ class HasLlamaCppProperties:
         """Set token id bias"""
         return self._call_java("setTokenBias", tokenBias)
 
-    def setLoraAdapters(self, loraAdapters: Dict[str, float]):
-        """Set LoRA adapters with their scaling factors"""
-        return self._call_java("setLoraAdapters", loraAdapters)
+    # def setLoraAdapters(self, loraAdapters: Dict[str, float]):
+    #     """Set LoRA adapters with their scaling factors"""
+    #     return self._call_java("setLoraAdapters", loraAdapters)
 
     def getMetadata(self):
         """Gets the metadata of the model"""

sparknlp/internal/__init__.py

@@ -1182,4 +1182,13 @@ class _E5VEmbeddingsLoader(ExtendedJavaWrapper):
             path,
             jspark,
             use_openvino
+        )
+
+class _Phi4Loader(ExtendedJavaWrapper):
+    def __init__(self, path, jspark, use_openvino=False):
+        super(_Phi4Loader, self).__init__(
+            "com.johnsnowlabs.nlp.annotators.seq2seq.Phi4Transformer.loadSavedModel",
+            path,
+            jspark,
+            use_openvino,
         )

sparknlp/reader/reader2doc.py

@@ -0,0 +1,194 @@
+#  Copyright 2017-2025 John Snow Labs
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+from pyspark import keyword_only
+from pyspark.ml.param import TypeConverters, Params, Param
+
+from sparknlp.common import AnnotatorType
+from sparknlp.internal import AnnotatorTransformer
+from sparknlp.partition.partition_properties import *
+
+
+class Reader2Doc(
+    AnnotatorTransformer,
+    HasEmailReaderProperties,
+    HasExcelReaderProperties,
+    HasHTMLReaderProperties,
+    HasPowerPointProperties,
+    HasTextReaderProperties,
+):
+    """
+    The Reader2Doc annotator allows you to use reading files more smoothly within existing
+    Spark NLP workflows, enabling seamless reuse of your pipelines.
+
+    Reader2Doc can be used for extracting structured content from various document types
+    using Spark NLP readers. It supports reading from many file types and returns parsed
+    output as a structured Spark DataFrame.
+
+    Supported formats include:
+    
+    - Plain text
+    - HTML
+    - Word (.doc/.docx)
+    - Excel (.xls/.xlsx)
+    - PowerPoint (.ppt/.pptx)
+    - Email files (.eml, .msg)
+    - PDFs
+
+    Examples
+    --------
+    >>> from johnsnowlabs.reader import Reader2Doc
+    >>> from johnsnowlabs.nlp.base import DocumentAssembler
+    >>> from pyspark.ml import Pipeline
+    >>> # Initialize Reader2Doc for PDF files
+    >>> reader2doc = Reader2Doc() \\
+    ...     .setContentType("application/pdf") \\
+    ...     .setContentPath(f"{pdf_directory}/")
+    >>> # Build the pipeline with the Reader2Doc stage
+    >>> pipeline = Pipeline(stages=[reader2doc])
+    >>> # Fit the pipeline to an empty DataFrame
+    >>> pipeline_model = pipeline.fit(empty_data_set)
+    >>> result_df = pipeline_model.transform(empty_data_set)
+    >>> # Show the resulting DataFrame
+    >>> result_df.show()
+    +------------------------------------------------------------------------------------------------------------------------------------+
+    |document                                                                                                                            |
+    +------------------------------------------------------------------------------------------------------------------------------------+
+    |[{'document', 0, 14, 'This is a Title', {'pageNumber': 1, 'elementType': 'Title', 'fileName': 'pdf-title.pdf'}, []}]               |
+    |[{'document', 15, 38, 'This is a narrative text', {'pageNumber': 1, 'elementType': 'NarrativeText', 'fileName': 'pdf-title.pdf'}, []}]|
+    |[{'document', 39, 68, 'This is another narrative text', {'pageNumber': 1, 'elementType': 'NarrativeText', 'fileName': 'pdf-title.pdf'}, []}]|
+    +------------------------------------------------------------------------------------------------------------------------------------+
+"""
+
+    name = "Reader2Doc"
+    outputAnnotatorType = AnnotatorType.DOCUMENT
+
+    contentPath = Param(
+        Params._dummy(),
+        "contentPath",
+        "contentPath path to files to read",
+        typeConverter=TypeConverters.toString,
+    )
+
+    outputCol = Param(
+        Params._dummy(),
+        "outputCol",
+        "output column name",
+        typeConverter=TypeConverters.toString,
+    )
+
+    contentType = Param(
+        Params._dummy(),
+        "contentType",
+        "Set the content type to load following MIME specification",
+        typeConverter=TypeConverters.toString,
+    )
+
+    explodeDocs = Param(
+        Params._dummy(),
+        "explodeDocs",
+        "whether to explode the documents into separate rows",
+        typeConverter=TypeConverters.toBoolean,
+    )
+
+    flattenOutput = Param(
+        Params._dummy(),
+        "flattenOutput",
+        "If true, output is flattened to plain text with minimal metadata",
+        typeConverter=TypeConverters.toBoolean,
+    )
+
+    titleThreshold = Param(
+        Params._dummy(),
+        "titleThreshold",
+        "Minimum font size threshold for title detection in PDF docs",
+        typeConverter=TypeConverters.toFloat,
+    )
+
+    @keyword_only
+    def __init__(self):
+        super(Reader2Doc, self).__init__(classname="com.johnsnowlabs.reader.Reader2Doc")
+        self._setDefault(
+            outputCol="document",
+            explodeDocs=False,
+            contentType="",
+            flattenOutput=False,
+            titleThreshold=18
+        )
+    @keyword_only
+
+    def setParams(self):
+        kwargs = self._input_kwargs
+        return self._set(**kwargs)
+
+    def setContentPath(self, value):
+        """Sets content path.
+
+        Parameters
+        ----------
+        value : str
+            contentPath path to files to read
+        """
+        return self._set(contentPath=value)
+
+    def setContentType(self, value):
+        """
+        Set the content type to load following MIME specification
+
+        Parameters
+        ----------
+        value : str
+            content type to load following MIME specification
+        """
+        return self._set(contentType=value)
+
+    def setExplodeDocs(self, value):
+        """Sets whether to explode the documents into separate rows.
+
+
+        Parameters
+        ----------
+        value : boolean
+        Whether to explode the documents into separate rows
+        """
+        return self._set(explodeDocs=value)
+
+    def setOutputCol(self, value):
+        """Sets output column name.
+
+        Parameters
+        ----------
+        value : str
+            Name of the Output Column
+        """
+        return self._set(outputCol=value)
+
+    def setFlattenOutput(self, value):
+        """Sets whether to flatten the output to plain text with minimal metadata.
+
+        Parameters
+        ----------
+        value : bool
+            If true, output is flattened to plain text with minimal metadata
+        """
+        return self._set(flattenOutput=value)
+
+    def setTitleThreshold(self, value):
+        """Sets the minimum font size threshold for title detection in PDF documents.
+
+        Parameters
+        ----------
+        value : float
+            Minimum font size threshold for title detection in PDF docs
+        """
+        return self._set(titleThreshold=value)

sparknlp/reader/sparknlp_reader.py

@@ -413,4 +413,49 @@ class SparkNLPReader(ExtendedJavaWrapper):
         if not isinstance(filePath, str):
             raise TypeError("filePath must be a string")
         jdf = self._java_obj.md(filePath)
+        return self.getDataFrame(self.spark, jdf)
+
+    def csv(self, csvPath):
+        """Reads CSV files and returns a Spark DataFrame.
+
+        Parameters
+        ----------
+        docPath : str
+            Path to an CSV file or a directory containing CSV files.
+
+        Returns
+        -------
+        pyspark.sql.DataFrame
+            A DataFrame containing parsed CSV content.
+
+        Examples
+        --------
+        >>> from sparknlp.reader import SparkNLPReader
+        >>> csv_df = SparkNLPReader(spark).csv("home/user/csv-directory")
+
+        You can use SparkNLP for one line of code
+
+        >>> import sparknlp
+        >>> csv_df = sparknlp.read().csv("home/user/csv-directory")
+        >>> csv_df.show(truncate=False)
+        +-----------------------------------------------------------------------------------------------------------------------------------------+
+        |csv                                                                                                                                      |
+        +-----------------------------------------------------------------------------------------------------------------------------------------+
+        |[{NarrativeText, Alice 100 Bob 95, {}}, {Table, <table><tr><td>Alice</td><td>100</td></tr><tr><td>Bob</td><td>95</td></tr></table>, {}}] |
+        +-----------------------------------------------------------------------------------------------------------------------------------------+
+
+        >>> csv_df.printSchema()
+        root
+         |-- path: string (nullable = true)
+         |-- csv: array (nullable = true)
+         |    |-- element: struct (containsNull = true)
+         |    |    |-- elementType: string (nullable = true)
+         |    |    |-- content: string (nullable = true)
+         |    |    |-- metadata: map (nullable = true)
+         |    |    |    |-- key: string
+         |    |    |    |-- value: string (valueContainsNull = true)
+        """
+        if not isinstance(csvPath, str):
+            raise TypeError("docPath must be a string")
+        jdf = self._java_obj.csv(csvPath)
         return self.getDataFrame(self.spark, jdf)