Source code for sparknlp.annotator.seq2seq.phi4_transformer
# 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 *
[docs]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.]|
+------------------------------------------------+
"""
[docs] configProtoBytes = Param(Params._dummy(),
"configProtoBytes",
"ConfigProto from tensorflow, serialized into byte array. Get with config_proto.SerializeToString()",
TypeConverters.toListInt)
[docs] minOutputLength = Param(Params._dummy(), "minOutputLength", "Minimum length of the sequence to be generated",
typeConverter=TypeConverters.toInt)
[docs] maxOutputLength = Param(Params._dummy(), "maxOutputLength", "Maximum length of output text",
typeConverter=TypeConverters.toInt)
[docs] doSample = Param(Params._dummy(), "doSample", "Whether or not to use sampling; use greedy decoding otherwise",
typeConverter=TypeConverters.toBoolean)
[docs] temperature = Param(Params._dummy(), "temperature", "The value used to module the next token probabilities",
typeConverter=TypeConverters.toFloat)
[docs] topK = Param(Params._dummy(), "topK",
"The number of highest probability vocabulary tokens to keep for top-k-filtering",
typeConverter=TypeConverters.toInt)
[docs] 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)
[docs] 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)
[docs] noRepeatNgramSize = Param(Params._dummy(), "noRepeatNgramSize",
"If set to int > 0, all ngrams of that size can only occur once",
typeConverter=TypeConverters.toInt)
[docs] ignoreTokenIds = Param(Params._dummy(), "ignoreTokenIds",
"A list of token ids which are ignored in the decoder's output",
typeConverter=TypeConverters.toListInt)
[docs] beamSize = Param(Params._dummy(), "beamSize",
"The number of beams to use for beam search",
typeConverter=TypeConverters.toInt)
[docs] stopTokenIds = Param(Params._dummy(), "stopTokenIds",
"A list of token ids which are considered as stop tokens in the decoder's output",
typeConverter=TypeConverters.toListInt)
[docs] 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)
[docs] 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)
[docs] 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)
[docs] def setMaxOutputLength(self, value):
"""Sets maximum length of output text.
Parameters
----------
value : int
Maximum length of output text
"""
return self._set(maxOutputLength=value)
[docs] 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)
[docs] 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)
[docs] 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)
[docs] 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)
[docs] 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)
[docs] 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)
[docs] 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)
[docs] 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
[docs] 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
[docs] 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)