2.16.1

What's new?

• Add support for the image-feature-extraction pipeline in #650.

  Example: Perform image feature extraction with Xenova/vit-base-patch16-224-in21k.

  const image_feature_extractor = await pipeline('image-feature-extraction', 'Xenova/vit-base-patch16-224-in21k');
  const url = 'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/cats.png';
  const features = await image_feature_extractor(url);
  // Tensor {
  //   dims: [ 1, 197, 768 ],
  //   type: 'float32',
  //   data: Float32Array(151296) [ ... ],
  //   size: 151296
  // }

  Example: Compute image embeddings with Xenova/clip-vit-base-patch32.

  const image_feature_extractor = await pipeline('image-feature-extraction', 'Xenova/clip-vit-base-patch32');
  const url = 'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/cats.png';
  const features = await image_feature_extractor(url);
  // Tensor {
  //   dims: [ 1, 512 ],
  //   type: 'float32',
  //   data: Float32Array(512) [ ... ],
  //   size: 512
  // }

• Fix channel format when padding non-square images for certain models in #655. This means you can now perform super-resolution for non-square images with APISR models:

  Example: Upscale an image with Xenova/4x_APISR_GRL_GAN_generator-onnx.

  import { pipeline } from '@xenova/transformers';
  
  // Create image-to-image pipeline
  const upscaler = await pipeline('image-to-image', 'Xenova/4x_APISR_GRL_GAN_generator-onnx', {
      quantized: false,
  });
  
  // Upscale an image
  const url = 'https://huggingface.co/datasets/Xenova/transformers.js-docs/resolve/main/anime.png';
  const output = await upscaler(url);
  // RawImage {
  //   data: Uint8Array(16588800) [ ... ],
  //   width: 2560,
  //   height: 1920,
  //   channels: 3
  // }
  
  // (Optional) Save the upscaled image
  output.save('upscaled.png');

  See example output

  Input image:
  

  Output image:
  

• Update tokenizer apply_chat_template functionality in #647. This PR added functionality to support the new C4AI Command-R tokenizer.

  See example tool usage

  import { AutoTokenizer } from "@xenova/transformers";
  
  const tokenizer = await AutoTokenizer.from_pretrained("Xenova/c4ai-command-r-v01-tokenizer")
  
  // define conversation input:
  const conversation = [
    { role: "user", content: "Whats the biggest penguin in the world?" }
  ]
  // Define tools available for the model to use:
  const tools = [
    {
      name: "internet_search",
      description: "Returns a list of relevant document snippets for a textual query retrieved from the internet",
      parameter_definitions: {
        query: {
          description: "Query to search the internet with",
          type: "str",
          required: true
        }
      }
    },
    {
      name: "directly_answer",
      description: "Calls a standard (un-augmented) AI chatbot to generate a response given the conversation history",
      parameter_definitions: {}
    }
  ]
  
  
  // render the tool use prompt as a string:
  const tool_use_prompt = tokenizer.apply_chat_template(
    conversation,
    {
      chat_template: "tool_use",
      tokenize: false,
      add_generation_prompt: true,
      tools,
    }
  )
  console.log(tool_use_prompt)

  See example RAG usage

  import { AutoTokenizer } from "@xenova/transformers";
  
  const tokenizer = await AutoTokenizer.from_pretrained("Xenova/c4ai-command-r-v01-tokenizer")
  
  // define conversation input:
  const conversation = [
    { role: "user", content: "Whats the biggest penguin in the world?" }
  ]
  // define documents to ground on:
  const documents = [
    { title: "Tall penguins", text: "Emperor penguins are the tallest growing up to 122 cm in height." },
    { title: "Penguin habitats", text: "Emperor penguins only live in Antarctica." }
  ]
  
  // render the RAG prompt as a string:
  const grounded_generation_prompt = tokenizer.apply_chat_template(
    conversation,
    {
      chat_template: "rag",
      tokenize: false,
      add_generation_prompt: true,
  
      documents,
      citation_mode: "accurate", // or "fast"
    }
  )
  console.log(grounded_generation_prompt);

• Add support for EfficientNet in #639.

  Example: Classify images with chriamue/bird-species-classifier

  import { pipeline } from '@xenova/transformers';
  
  // Create image classification pipeline
  const classifier = await pipeline('image-classification', 'chriamue/bird-species-classifier', {
      quantized: false,      // Quantized model doesn't work
      revision: 'refs/pr/1', // Needed until the model author merges the PR
  });
  
  // Classify an image
  const url = 'https://upload.wikimedia.org/wikipedia/commons/7/73/Short_tailed_Albatross1.jpg';
  const output = await classifier(url);
  console.log(output)
  // [{ label: 'ALBATROSS', score: 0.9999023079872131 }]

Full Changelog: 2.16.0...2.16.1