sparknlp.annotator.seq2seq.llama3_transformer#

Classes#

class LLAMA3Transformer(classname='com.johnsnowlabs.nlp.annotators.seq2seq.LLAMA3Transformer', java_model=None)[source]#

Llama 3: Cutting-Edge Foundation and Fine-Tuned Chat Models

The Llama 3 release introduces a new family of pretrained and fine-tuned LLMs, ranging in scale from 8B and 70B parameters. Llama 3 models are designed with enhanced efficiency, performance, and safety, making them more capable than previous versions. These models are trained on a more diverse and expansive dataset, offering improved contextual understanding and generation quality.

The fine-tuned models, known as Llama 3-instruct, are optimized for dialogue applications using an advanced version of Reinforcement Learning from Human Feedback (RLHF). Llama 3-instruct models demonstrate superior performance across multiple benchmarks, outperforming Llama 2 and even matching or exceeding the capabilities of some closed-source models.

Pretrained models can be loaded with pretrained() of the companion object:

>>> llama3 = LLAMA3Transformer.pretrained() \
...     .setInputCols(["document"]) \
...     .setOutputCol("generation")

The default model is "llama3-7b", if no name is provided. For available pretrained models please see the Models Hub.

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.

References

• Llama 3: Cutting-Edge Foundation and Fine-Tuned Chat Models

• facebookresearch/llama

Paper Abstract:

Llama 3 is the latest iteration of large language models from Meta, offering a range of models from 1 billion to 70 billion parameters. The fine-tuned versions, known as Llama 3-Chat, are specifically designed for dialogue applications and have been optimized using advanced techniques such as RLHF. Llama 3 models show remarkable improvements in both safety and performance, making them a leading choice in both open-source and commercial settings. Our comprehensive approach to training and fine-tuning these models is aimed at encouraging responsible AI development and fostering community collaboration.

Examples

>>> import sparknlp
>>> from sparknlp.base import *
>>> from sparknlp.annotator import *
>>> from pyspark.ml import Pipeline
>>> documentAssembler = DocumentAssembler() \
...     .setInputCol("text") \
...     .setOutputCol("documents")
>>> llama3 = LLAMA3Transformer.pretrained("llama_3_7b_chat_hf_int8") \
...     .setInputCols(["documents"]) \
...     .setMaxOutputLength(60) \
...     .setOutputCol("generation")
>>> pipeline = Pipeline().setStages([documentAssembler, llama3])
>>> data = spark.createDataFrame([
...     (
...         1,
...         "<|start_header_id|>system<|end_header_id|> \n"+         ...         "You are a minion chatbot who always responds in minion speak! \n" +         ...         "<|start_header_id|>user<|end_header_id|> \n" +         ...         "Who are you? \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                                          |
+------------------------------------------------+
|[Oooh, me am Minion! Me help you with things! Me speak Minion language, yeah! Bana-na-na!]|
+------------------------------------------------+

setIgnoreTokenIds(value)[source]#

A list of token ids which are ignored in the decoder’s output.

Parameters:

valueList[int]

The words to be filtered out

setConfigProtoBytes(b)[source]#

Sets configProto from tensorflow, serialized into byte array.

Parameters:

bList[int]

ConfigProto from tensorflow, serialized into byte array

setMinOutputLength(value)[source]#

Sets minimum length of the sequence to be generated.

Parameters:

valueint

Minimum length of the sequence to be generated

setMaxOutputLength(value)[source]#

Sets maximum length of output text.

Parameters:

valueint

Maximum length of output text

setDoSample(value)[source]#

Sets whether or not to use sampling, use greedy decoding otherwise.

Parameters:

valuebool

Whether or not to use sampling; use greedy decoding otherwise

setTemperature(value)[source]#

Sets the value used to module the next token probabilities.

Parameters:

valuefloat

The value used to module the next token probabilities

setTopK(value)[source]#

Sets the number of highest probability vocabulary tokens to keep for top-k-filtering.

Parameters:

valueint

Number of highest probability vocabulary tokens to keep

setTopP(value)[source]#

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:

valuefloat

Cumulative probability for vocabulary tokens

setRepetitionPenalty(value)[source]#

Sets the parameter for repetition penalty. 1.0 means no penalty.

Parameters:

valuefloat

The repetition penalty

References

See Ctrl: A Conditional Transformer Language Model For Controllable Generation for more details.

setNoRepeatNgramSize(value)[source]#

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:

valueint

N-gram size can only occur once

setBeamSize(value)[source]#

Sets the number of beams to use for beam search.

Parameters:

valueint

The number of beams to use for beam search

setStopTokenIds(value)[source]#

Sets a list of token ids which are considered as stop tokens in the decoder’s output.

Parameters:

valueList[int]

The words to be considered as stop tokens

static loadSavedModel(folder, spark_session, use_openvino=False)[source]#

Loads a locally saved model.

Parameters:

folderstr

Folder of the saved model

spark_sessionpyspark.sql.SparkSession

The current SparkSession

Returns:

LLAMA3Transformer

The restored model

static pretrained(name='llama_3_7b_chat_hf_int4', lang='en', remote_loc=None)[source]#

Downloads and loads a pretrained model.

Parameters:

namestr, optional

Name of the pretrained model, by default “llama_2_7b_chat_hf_int4”

langstr, optional

Language of the pretrained model, by default “en”

remote_locstr, optional

Optional remote address of the resource, by default None. Will use Spark NLPs repositories otherwise.

Returns:

LLAMA3Transformer

The restored model