Transformers.js

Demo

Don’t believe us? Play around with some of these models:

Task:

Installation

If you use npm,
you can install it using:

npm i @xenova/transformers

Alternatively, you can use it in a

Basic example

It's super easy to translate from existing code!

from transformers import pipeline

# Allocate a pipeline for sentiment-analysis
pipe = pipeline('sentiment-analysis')

out = pipe('I love transformers!')
# [{'label': 'POSITIVE', 'score': 0.999806941}]

import { pipeline } from "@xenova/transformers";

// Allocate a pipeline for sentiment-analysis
let pipe = await pipeline('sentiment-analysis');

let out = await pipe('I love transformers!');
// [{'label': 'POSITIVE', 'score': 0.999817686}]

In the same way as the Python library, you can use a different model by providing its
name as the second argument to the pipeline function. For example:

// Use a different model for sentiment-analysis
let pipe = await pipeline('sentiment-analysis', 'nlptown/bert-base-multilingual-uncased-sentiment');

Custom setup

By default, Transformers.js uses hosted models
precompiled
WASM binaries,
which should work out-of-the-box. You can override this behaviour as follows:

import { env } from "@xenova/transformers";

// Use a different host for models.
// - `remoteURL` defaults to use the HuggingFace Hub
// - `localURL` defaults to '/models/onnx/quantized/'
env.remoteURL = 'https://www.example.com/';
env.localURL = '/path/to/models/';

// Set whether to use remote or local models. Defaults to true.
//  - If true, use the path specified by `env.remoteURL`.
//  - If false, use the path specified by `env.localURL`.
env.remoteModels = false;

// Set parent path of .wasm files. Defaults to use a CDN.
env.onnx.wasm.wasmPaths = '/path/to/files/';

Usage

We currently support the following
tasks and
models, which can be used with the
pipeline
function.

1. sentiment-analysis (a.k.a. text-classification)

   Supported models: distilbert-base-uncased-finetuned-sst-2-english,
   nlptown/bert-base-multilingual-uncased-sentiment,
   distilgpt2.
   For more information, check out the
   
   Text Classification docs
   .

2. question-answering

   Supported models: distilbert-base-cased-distilled-squad,
   distilbert-base-uncased-distilled-squad.
   For more information, check out the
   
   Question Answering docs
   .

3. fill-mask

   Supported models: xlm-roberta-base, albert-large-v2,
   albert-base-v2, distilroberta-base, roberta-base,
   bert-base-cased, bert-base-uncased,
   bert-base-multilingual-uncased, bert-base-multilingual-cased,
   distilbert-base-cased, distilbert-base-uncased.
   For more information, check out the
   
   Language Modelling docs
   .

4. summarization

   Supported models:
   t5-small,
   t5-base,
   t5-v1_1-small,
   t5-v1_1-base,
   sshleifer/distilbart-cnn-6-6,
   sshleifer/distilbart-cnn-12-6.
   For more information, check out the
   
   Summarization docs
   .

5. translation (translation_en_to_de, translation_en_to_fr,
   translation_en_to_ro)

   Supported models:
   t5-small,
   t5-base,
   t5-v1_1-small,
   t5-v1_1-base.
   For more information, check out the
   
   Translation docs
   .

6. text2text-generation

   Supported models:
   google/flan-t5-small,
   google/flan-t5-base,
   t5-small,
   t5-base,
   google/t5-v1_1-small,
   google/t5-v1_1-base,
   sshleifer/distilbart-cnn-6-6,
   sshleifer/distilbart-cnn-12-6.
   For more information, check out the
   
   Text Generation docs
   .

7. text-generation

   Supported models:
   gpt2,
   distilgpt2,
   Salesforce/codegen-350M-mono,
   Salesforce/codegen-350M-multi,
   Salesforce/codegen-350M-nl.
   For more information, check out the
   
   Text Generation docs
   .

8. automatic-speech-recognition

   Supported models:
   openai/whisper-tiny.en,
   openai/whisper-tiny,
   openai/whisper-small.en,
   openai/whisper-small,
   openai/whisper-base.en,
   openai/whisper-base.
   For more information, check out the
   
   Automatic Speech Recognition docs
   .

9. image-to-text

   Supported models:
   nlpconnect/vit-gpt2-image-captioning.
   For more information, check out the
   
   Image-to-Text docs
   .

10. image-classification

    Supported models:
    google/vit-base-patch16-224.
    For more information, check out the
    
    Image Classification docs
    .

11. zero-shot-image-classification

    Supported models:
    openai/clip-vit-base-patch16,
    openai/clip-vit-base-patch32.
    For more information, check out the
    
    Zero-Shot Image Classification
    .

12. embeddings

    Supported models:
    sentence-transformers/all-MiniLM-L6-v2,
    sentence-transformers/all-MiniLM-L12-v2,
    sentence-transformers/all-distilroberta-v1,
    sentence-transformers/paraphrase-albert-base-v2,
    sentence-transformers/paraphrase-albert-small-v2.
    For more information, check out the
    
    Embeddings docs
    .

The following
model types
are supported:

1. BERT

   Tasks: Masked language modelling
   (AutoModelForMaskedLM),
   question answering
   (AutoModelForQuestionAnswering), and
   sequence classification
   (AutoModelForSequenceClassification).
   For more information, check out the BERT docs.

2. ALBERT

   Tasks: Masked language modelling
   (AutoModelForMaskedLM).
   For more information, check out the ALBERT docs.

3. DistilBERT

   Tasks: Masked language modelling
   (AutoModelForMaskedLM),
   question answering
   (AutoModelForQuestionAnswering), and
   sequence classification
   (AutoModelForSequenceClassification).
   For more information, check out the DistilBERT
   docs.

4. T5

   Tasks: Sequence-to-sequence for
   translation/summarization
   (AutoModelForSeq2SeqLM).
   For more information, check out the T5 docs.

5. T5v1.1

   Tasks: Sequence-to-sequence
   (AutoModelForSeq2SeqLM).
   For more information, check out the T5v1.1 docs.

6. FLAN-T5

   Tasks: Sequence-to-sequence for over 1000 tasks
   (AutoModelForSeq2SeqLM).
   For more information, check out the FLAN-T5 docs.

7. GPT2/DistilGPT2

   Tasks: Text generation
   (AutoModelForCausalLM).
   For more information, check out the
   GPT2 docs or
   DistilGPT2 docs.

8. BART

   Tasks: Sequence-to-sequence for summarization
   (AutoModelForSeq2SeqLM).
   For more information, check out the BART docs.

9. CodeGen

   Tasks: Text generation
   (AutoModelForCausalLM).
   For more information, check out the
   CodeGen docs.

10. Whisper

    Tasks: Automatic speech recognition
    (AutoModelForSeq2SeqLM).
    For more information, check out the Whisper docs.

11. CLIP

    Tasks: Zero-shot Image classification
    (AutoModel).
    For more information, check out the CLIP
    docs.

12. Vision Transformer (ViT)

    Tasks: Image classification
    (AutoModelForImageClassification).
    For more information, check out the Vision Transformer
    docs.

Coming soon... In the meantime, check out the source code for the demo
here.