Informers

🔥 Fast transformer inference for Ruby

For non-ONNX models, check out Transformers.rb 🙂

Installation

Add this line to your application’s Gemfile:

gem "informers"

Getting Started

• Models
• Pipelines

Models

Embedding

• sentence-transformers/all-MiniLM-L6-v2
• sentence-transformers/multi-qa-MiniLM-L6-cos-v1
• sentence-transformers/all-mpnet-base-v2
• sentence-transformers/paraphrase-MiniLM-L6-v2
• mixedbread-ai/mxbai-embed-large-v1
• Supabase/gte-small
• intfloat/e5-base-v2
• nomic-ai/nomic-embed-text-v1
• BAAI/bge-base-en-v1.5
• jinaai/jina-embeddings-v2-base-en
• Snowflake/snowflake-arctic-embed-m-v1.5

Reranking

• mixedbread-ai/mxbai-rerank-base-v1
• jinaai/jina-reranker-v1-turbo-en
• BAAI/bge-reranker-base
• Xenova/ms-marco-MiniLM-L-6-v2

sentence-transformers/all-MiniLM-L6-v2

Docs

sentences = ["This is an example sentence", "Each sentence is converted"]

model = Informers.pipeline("embedding", "sentence-transformers/all-MiniLM-L6-v2")
embeddings = model.(sentences)

sentence-transformers/multi-qa-MiniLM-L6-cos-v1

Docs

query = "How many people live in London?"
docs = ["Around 9 Million people live in London", "London is known for its financial district"]

model = Informers.pipeline("embedding", "sentence-transformers/multi-qa-MiniLM-L6-cos-v1")
query_embedding = model.(query)
doc_embeddings = model.(docs)
scores = doc_embeddings.map { |e| e.zip(query_embedding).sum { |d, q| d * q } }
doc_score_pairs = docs.zip(scores).sort_by { |d, s| -s }

sentence-transformers/all-mpnet-base-v2

Docs

sentences = ["This is an example sentence", "Each sentence is converted"]

model = Informers.pipeline("embedding", "sentence-transformers/all-mpnet-base-v2")
embeddings = model.(sentences)

sentence-transformers/paraphrase-MiniLM-L6-v2

Docs

sentences = ["This is an example sentence", "Each sentence is converted"]

model = Informers.pipeline("embedding", "sentence-transformers/paraphrase-MiniLM-L6-v2")
embeddings = model.(sentences, normalize: false)

mixedbread-ai/mxbai-embed-large-v1

Docs

query_prefix = "Represent this sentence for searching relevant passages: "

input = [
  "The dog is barking",
  "The cat is purring",
  query_prefix + "puppy"
]

model = Informers.pipeline("embedding", "mixedbread-ai/mxbai-embed-large-v1")
embeddings = model.(input)

Supabase/gte-small

Docs

sentences = ["That is a happy person", "That is a very happy person"]

model = Informers.pipeline("embedding", "Supabase/gte-small")
embeddings = model.(sentences)

intfloat/e5-base-v2

Docs

doc_prefix = "passage: "
query_prefix = "query: "

input = [
  doc_prefix + "Ruby is a programming language created by Matz",
  query_prefix + "Ruby creator"
]

model = Informers.pipeline("embedding", "intfloat/e5-base-v2")
embeddings = model.(input)

nomic-ai/nomic-embed-text-v1

Docs

doc_prefix = "search_document: "
query_prefix = "search_query: "

input = [
  doc_prefix + "The dog is barking",
  doc_prefix + "The cat is purring",
  query_prefix + "puppy"
]

model = Informers.pipeline("embedding", "nomic-ai/nomic-embed-text-v1")
embeddings = model.(input)

BAAI/bge-base-en-v1.5

Docs

query_prefix = "Represent this sentence for searching relevant passages: "

input = [
  "The dog is barking",
  "The cat is purring",
  query_prefix + "puppy"
]

model = Informers.pipeline("embedding", "BAAI/bge-base-en-v1.5")
embeddings = model.(input)

jinaai/jina-embeddings-v2-base-en

Docs

sentences = ["How is the weather today?", "What is the current weather like today?"]

model = Informers.pipeline("embedding", "jinaai/jina-embeddings-v2-base-en", model_file_name: "../model")
embeddings = model.(sentences)

Snowflake/snowflake-arctic-embed-m-v1.5

Docs

query_prefix = "Represent this sentence for searching relevant passages: "

input = [
  "The dog is barking",
  "The cat is purring",
  query_prefix + "puppy"
]

model = Informers.pipeline("embedding", "Snowflake/snowflake-arctic-embed-m-v1.5")
embeddings = model.(input, model_output: "sentence_embedding", pooling: "none")

mixedbread-ai/mxbai-rerank-base-v1

Docs

query = "How many people live in London?"
docs = ["Around 9 Million people live in London", "London is known for its financial district"]

model = Informers.pipeline("reranking", "mixedbread-ai/mxbai-rerank-base-v1")
result = model.(query, docs)

jinaai/jina-reranker-v1-turbo-en

Docs

query = "How many people live in London?"
docs = ["Around 9 Million people live in London", "London is known for its financial district"]

model = Informers.pipeline("reranking", "jinaai/jina-reranker-v1-turbo-en")
result = model.(query, docs)

BAAI/bge-reranker-base

Docs

query = "How many people live in London?"
docs = ["Around 9 Million people live in London", "London is known for its financial district"]

model = Informers.pipeline("reranking", "BAAI/bge-reranker-base")
result = model.(query, docs)

Xenova/ms-marco-MiniLM-L-6-v2

Docs

query = "How many people live in London?"
docs = ["Around 9 Million people live in London", "London is known for its financial district"]

model = Informers.pipeline("reranking", "Xenova/ms-marco-MiniLM-L-6-v2")
result = model.(query, docs)

Other

The model must include a .onnx file (example). If the file is not at onnx/model.onnx, use the model_file_name option to specify the location.

Pipelines

• Text
• Vision
• Audio
• Multimodel

Text

Embedding

embed = Informers.pipeline("embedding")
embed.("We are very happy to show you the 🤗 Transformers library.")

Reranking

rerank = Informers.pipeline("reranking")
rerank.("Who created Ruby?", ["Matz created Ruby", "Another doc"])

Named-entity recognition

ner = Informers.pipeline("ner")
ner.("Ruby is a programming language created by Matz")

Sentiment analysis

classifier = Informers.pipeline("sentiment-analysis")
classifier.("We are very happy to show you the 🤗 Transformers library.")

Question answering

qa = Informers.pipeline("question-answering")
qa.("Who invented Ruby?", "Ruby is a programming language created by Matz")

Zero-shot classification

classifier = Informers.pipeline("zero-shot-classification")
classifier.("text", ["label1", "label2", "label3"])

Text generation

generator = Informers.pipeline("text-generation")
generator.("I enjoy walking with my cute dog,")

Text-to-text generation

text2text = Informers.pipeline("text2text-generation")
text2text.("translate from English to French: I'm very happy")

Translation

translator = Informers.pipeline("translation", "Xenova/nllb-200-distilled-600M")
translator.("जीवन एक चॉकलेट बॉक्स की तरह है।", src_lang: "hin_Deva", tgt_lang: "fra_Latn")

Summarization

summarizer = Informers.pipeline("summarization")
summarizer.("Many paragraphs of text")

Fill mask

unmasker = Informers.pipeline("fill-mask")
unmasker.("Paris is the [MASK] of France.")

Feature extraction

extractor = Informers.pipeline("feature-extraction")
extractor.("We are very happy to show you the 🤗 Transformers library.")

Vision

Note: ruby-vips is required to load images

Image classification

classifier = Informers.pipeline("image-classification")
classifier.("image.jpg")

Zero-shot image classification

classifier = Informers.pipeline("zero-shot-image-classification")
classifier.("image.jpg", ["label1", "label2", "label3"])

Image segmentation

segmenter = Informers.pipeline("image-segmentation")
segmenter.("image.jpg")

Object detection

detector = Informers.pipeline("object-detection")
detector.("image.jpg")

Zero-shot object detection

detector = Informers.pipeline("zero-shot-object-detection")
detector.("image.jpg", ["label1", "label2", "label3"])

Depth estimation

estimator = Informers.pipeline("depth-estimation")
estimator.("image.jpg")

Image-to-image

upscaler = Informers.pipeline("image-to-image")
upscaler.("image.jpg")

Image feature extraction

extractor = Informers.pipeline("image-feature-extraction")
extractor.("image.jpg")

Audio

Note: ffmpeg is required to load audio files

Audio classification

classifier = Informers.pipeline("audio-classification")
classifier.("audio.wav")

Multimodal

Image captioning

captioner = Informers.pipeline("image-to-text")
captioner.("image.jpg")

Document question answering

qa = Informers.pipeline("document-question-answering")
qa.("image.jpg", "What is the invoice number?")

Reference

Specify a variant of the model if available (fp32, fp16, int8, uint8, q8, q4, q4f16, or bnb4)

Informers.pipeline("embedding", "Xenova/all-MiniLM-L6-v2", dtype: "fp16")

Specify a device (cpu, cuda, or coreml)

Informers.pipeline("embedding", device: "cuda")

Note: Follow these instructions for cuda

Specify ONNX Runtime session options

Informers.pipeline("embedding", session_options: {log_severity_level: 2})

Credits

This library was ported from Transformers.js and is available under the same license.

Upgrading

1.0

Task classes have been replaced with the pipeline method.

# before
model = Informers::SentimentAnalysis.new("sentiment-analysis.onnx")
model.predict("This is super cool")

# after
model = Informers.pipeline("sentiment-analysis")
model.("This is super cool")

History

View the changelog

Contributing

Everyone is encouraged to help improve this project. Here are a few ways you can help:

• Report bugs
• Fix bugs and submit pull requests
• Write, clarify, or fix documentation
• Suggest or add new features

To get started with development:

git clone https://github.com/ankane/informers.git
cd informers
bundle install
bundle exec rake download:files
bundle exec rake test