spark-nlp 6.0.5__tar.gz → 6.1.0__tar.gz

{spark_nlp-6.0.5 → spark_nlp-6.1.0}/PKG-INFO

@@ -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 → spark_nlp-6.1.0}/README.md

@@ -63,7 +63,7 @@ $ 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:
@@ -129,7 +129,7 @@ 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 |
 |-----------|--------------------|--------------------|--------------------|--------------------|--------------------|--------------------|--------------------|--------------------|
@@ -159,7 +159,7 @@ 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**            |
 |--------------------|--------------------|
@@ -176,7 +176,7 @@ 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 → spark_nlp-6.1.0}/setup.py

@@ -41,7 +41,7 @@ setup(
     # project code, see
     # https://packaging.python.org/en/latest/single_source_version.html
 
-    version='6.0.5',  # Required
+    version='6.1.0',  # Required
 
     # This is a one-line description or tagline of what your project does. This
     # corresponds to the 'Summary' metadata field:

{spark_nlp-6.0.5 → spark_nlp-6.1.0}/spark_nlp.egg-info/PKG-INFO

@@ -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 → spark_nlp-6.1.0}/spark_nlp.egg-info/SOURCES.txt

@@ -190,6 +190,7 @@ sparknlp/annotator/seq2seq/nllb_transformer.py
 sparknlp/annotator/seq2seq/olmo_transformer.py
 sparknlp/annotator/seq2seq/phi2_transformer.py
 sparknlp/annotator/seq2seq/phi3_transformer.py
+sparknlp/annotator/seq2seq/phi4_transformer.py
 sparknlp/annotator/seq2seq/qwen_transformer.py
 sparknlp/annotator/seq2seq/starcoder_transformer.py
 sparknlp/annotator/seq2seq/t5_transformer.py
@@ -253,6 +254,7 @@ sparknlp/pretrained/utils.py
 sparknlp/reader/__init__.py
 sparknlp/reader/enums.py
 sparknlp/reader/pdf_to_text.py
+sparknlp/reader/reader2doc.py
 sparknlp/reader/sparknlp_reader.py
 sparknlp/training/__init__.py
 sparknlp/training/conll.py

{spark_nlp-6.0.5 → spark_nlp-6.1.0}/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,

{spark_nlp-6.0.5 → spark_nlp-6.1.0}/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 *

{spark_nlp-6.0.5 → spark_nlp-6.1.0}/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

spark_nlp-6.1.0/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)