Source code for sparknlp.annotator.seq2seq.olmo_transformer
# Copyright 2017-2022 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 OLMoTransformer."""
from sparknlp.common import *
[docs]class OLMoTransformer(AnnotatorModel, HasBatchedAnnotate, HasEngine):
"""OLMo: Open Language Models
OLMo is a series of Open Language Models designed to enable the science of language models.
The OLMo models are trained on the Dolma dataset. We release all code, checkpoints, logs
(coming soon), and details involved in training these models.
Pretrained models can be loaded with :meth:`.pretrained` of the companion
object:
>>> olmo = OLMoTransformer.pretrained() \\
... .setInputCols(["document"]) \\
... .setOutputCol("generation")
The default model is ``"olmo_1b_int4"``, if no name is provided. For available
pretrained models please see the `Models Hub
<https://sparknlp.org/models?q=olmo>`__.
====================== ======================
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 20
doSample
Whether or not to use sampling; use greedy decoding otherwise, by default False
temperature
The value used to module the next token probabilities, by default 1.0
topK
The number of highest probability vocabulary tokens to keep for
top-k-filtering, by default 50
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
sequence. The use of an accelerator such as GPU is recommended.
References
----------
- `OLMo Project Page.
<https://allenai.org/olmo>`__
- `OLMO GitHub Repository.
<https://github.com/allenai/OLMo>`__
- `OLMo: Accelerating the Science of Language Models
<https://arxiv.org/pdf/2402.00838.pdf>`__
**Paper Abstract:**
*Language models (LMs) have become ubiquitous in both NLP research and in commercial product offerings.
As their commercial importance has surged, the most powerful models have become closed off, gated behind
proprietary interfaces, with important details of their training data, architectures, and development
undisclosed. Given the importance of these details in scientifically studying these models, including
their biases and potential risks, we believe it is essential for the research community to have access
to powerful, truly open LMs. To this end, this technical report details the first release of OLMo,
a state-of-the-art, truly Open Language Model and its framework to build and study the science of
language modeling. Unlike most prior efforts that have only released model weights and inference code,
we release OLMo and the whole framework, including training data and training and evaluation code.
We hope this release will empower and strengthen the open research community and inspire a new wave
of innovation.*
Examples
--------
>>> import sparknlp
>>> from sparknlp.base import *
>>> from sparknlp.annotator import *
>>> from pyspark.ml import Pipeline
>>> documentAssembler = DocumentAssembler() \\
... .setInputCol("text") \\
... .setOutputCol("documents")
>>> olmo = OLMoTransformer.pretrained("olmo-7b") \\
... .setInputCols(["documents"]) \\
... .setMaxOutputLength(50) \\
... .setOutputCol("generation")
>>> pipeline = Pipeline().setStages([documentAssembler, olmo])
>>> data = spark.createDataFrame([["My name is Leonardo."]]).toDF("text")
>>> result = pipeline.fit(data).transform(data)
>>> result.select("summaries.generation").show(truncate=False)
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|result |
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|[My name is Leonardo . I am a student of the University of California, Berkeley. I am interested in the field of Artificial Intelligence and its applications in the real world. I have a strong |
| passion for learning and am always looking for ways to improve my knowledge and skills] |
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
"""
name = "OLMoTransformer"
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)
[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)
@keyword_only
def __init__(self, classname="com.johnsnowlabs.nlp.annotators.seq2seq.OLMoTransformer", java_model=None):
super(OLMoTransformer, self).__init__(classname=classname, java_model=java_model)
self._setDefault(minOutputLength=0, maxOutputLength=20, doSample=False, temperature=0.6, topK=50, topP=0.9,
repetitionPenalty=1.0, noRepeatNgramSize=0, ignoreTokenIds=[], batchSize=1)
[docs] @staticmethod
def loadSavedModel(folder, spark_session):
"""Loads a locally saved model.
Parameters
----------
folder : str
Folder of the saved model
spark_session : pyspark.sql.SparkSession
The current SparkSession
Returns
-------
OLMoTransformer
The restored model
"""
from sparknlp.internal import _OLMoLoader
jModel = _OLMoLoader(folder, spark_session._jsparkSession)._java_obj
return OLMoTransformer(java_model=jModel)
[docs] @staticmethod
def pretrained(name="olmo_1b_int4", lang="en", remote_loc=None):
"""Downloads and loads a pretrained model.
Parameters
----------
name : str, optional
Name of the pretrained model, by default "olmo-7b"
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
-------
OLMoTransformer
The restored model
"""
from sparknlp.pretrained import ResourceDownloader
return ResourceDownloader.downloadModel(OLMoTransformer, name, lang, remote_loc)