@huggingface/transformers 3.2.3 → 3.3.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@huggingface/transformers",
-  "version": "3.2.3",
+  "version": "3.3.0",
   "description": "State-of-the-art Machine Learning for the web. Run 🤗 Transformers directly in your browser, with no need for a server!",
   "main": "./src/transformers.js",
   "types": "./types/transformers.d.ts",
@@ -24,7 +24,7 @@
   "scripts": {
     "format": "prettier --write .",
     "format:check": "prettier --check .",
-    "typegen": "tsc ./src/transformers.js --allowJs --declaration --emitDeclarationOnly --declarationMap --outDir types",
+    "typegen": "tsc --build",
     "dev": "webpack serve --no-client-overlay",
     "build": "webpack && npm run typegen",
     "test": "node --experimental-vm-modules node_modules/jest/bin/jest.js --verbose",
@@ -57,7 +57,7 @@
   "dependencies": {
     "@huggingface/jinja": "^0.3.2",
     "onnxruntime-node": "1.20.1",
-    "onnxruntime-web": "1.21.0-dev.20241205-d27fecd3d3",
+    "onnxruntime-web": "1.21.0-dev.20250111-73f5b0c597",
     "sharp": "^0.33.5"
   },
   "devDependencies": {

package/src/base/feature_extraction_utils.js

@@ -17,23 +17,23 @@ export class FeatureExtractor extends Callable {
     }
 
     /**
-     * Instantiate one of the processor classes of the library from a pretrained model.
+     * Instantiate one of the feature extractor classes of the library from a pretrained model.
      * 
-     * The processor class to instantiate is selected based on the `image_processor_type` (or `feature_extractor_type`; legacy)
-     * property of the config object (either passed as an argument or loaded from `pretrained_model_name_or_path` if possible)
+     * The feature extractor class to instantiate is selected based on the `feature_extractor_type` property of
+     * the config object (either passed as an argument or loaded from `pretrained_model_name_or_path` if possible)
      * 
      * @param {string} pretrained_model_name_or_path The name or path of the pretrained model. Can be either:
-     * - A string, the *model id* of a pretrained processor hosted inside a model repo on huggingface.co.
+     * - A string, the *model id* of a pretrained feature_extractor hosted inside a model repo on huggingface.co.
      *   Valid model ids can be located at the root-level, like `bert-base-uncased`, or namespaced under a
      *   user or organization name, like `dbmdz/bert-base-german-cased`.
-     * - A path to a *directory* containing processor files, e.g., `./my_model_directory/`.
-     * @param {import('../utils/hub.js').PretrainedOptions} options Additional options for loading the processor.
+     * - A path to a *directory* containing feature_extractor files, e.g., `./my_model_directory/`.
+     * @param {import('../utils/hub.js').PretrainedOptions} options Additional options for loading the feature_extractor.
      * 
-     * @returns {Promise<FeatureExtractor>} A new instance of the Processor class.
+     * @returns {Promise<FeatureExtractor>} A new instance of the Feature Extractor class.
      */
     static async from_pretrained(pretrained_model_name_or_path, options) {
-        const preprocessorConfig = await getModelJSON(pretrained_model_name_or_path, FEATURE_EXTRACTOR_NAME, true, options);
-        return new this(preprocessorConfig);
+        const config = await getModelJSON(pretrained_model_name_or_path, FEATURE_EXTRACTOR_NAME, true, options);
+        return new this(config);
     }
 }
 

package/src/base/image_processors_utils.js

@@ -68,7 +68,7 @@ function enforce_size_divisibility([width, height], divisor) {
  * @param {number[]} arr The coordinate for the center of the box and its width, height dimensions (center_x, center_y, width, height)
  * @returns {number[]} The coodinates for the top-left and bottom-right corners of the box (top_left_x, top_left_y, bottom_right_x, bottom_right_y)
  */
-function center_to_corners_format([centerX, centerY, width, height]) {
+export function center_to_corners_format([centerX, centerY, width, height]) {
     return [
         centerX - width / 2,
         centerY - height / 2,
@@ -604,14 +604,20 @@ export class ImageProcessor extends Callable {
         this.do_thumbnail = config.do_thumbnail;
         this.size = config.size ?? config.image_size;
         this.do_resize = config.do_resize ?? (this.size !== undefined);
+        // @ts-expect-error TS2339
         this.size_divisibility = config.size_divisibility ?? config.size_divisor;
 
         this.do_center_crop = config.do_center_crop;
+        // @ts-expect-error TS2339
         this.crop_size = config.crop_size;
+        // @ts-expect-error TS2339
         this.do_convert_rgb = config.do_convert_rgb ?? true;
+        // @ts-expect-error TS2339
         this.do_crop_margin = config.do_crop_margin;
 
+        // @ts-expect-error TS2339
         this.pad_size = config.pad_size;
+        // @ts-expect-error TS2339
         this.do_pad = config.do_pad;
 
         if (this.do_pad && !this.pad_size && this.size && this.size.width !== undefined && this.size.height !== undefined) {
@@ -820,6 +826,7 @@ export class ImageProcessor extends Callable {
         // Support both formats for backwards compatibility
         else if (Number.isInteger(size)) {
             shortest_edge = size;
+            // @ts-expect-error TS2339
             longest_edge = this.config.max_size ?? shortest_edge;
 
         } else if (size !== undefined) {
@@ -888,6 +895,7 @@ export class ImageProcessor extends Callable {
         } else if (size.min_pixels !== undefined && size.max_pixels !== undefined) {
             // Custom resize logic for Qwen2-VL models
             const { min_pixels, max_pixels } = size;
+            // @ts-expect-error TS2339
             const factor = this.config.patch_size * this.config.merge_size;
             return smart_resize(srcHeight, srcWidth, factor, min_pixels, max_pixels);
         } else {
@@ -903,6 +911,7 @@ export class ImageProcessor extends Callable {
     async resize(image) {
         const [newWidth, newHeight] = this.get_resize_output_image_size(image, this.size);
         return await image.resize(newWidth, newHeight, {
+            // @ts-expect-error TS2322
             resample: this.resample,
         });
     }
@@ -953,6 +962,7 @@ export class ImageProcessor extends Callable {
 
         // Resize the image using thumbnail method.
         if (this.do_thumbnail) {
+            // @ts-expect-error TS2345
             image = await this.thumbnail(image, this.size, this.resample);
         }
 
@@ -977,6 +987,7 @@ export class ImageProcessor extends Callable {
         // NOTE: All pixel-level manipulation (i.e., modifying `pixelData`)
         // occurs with data in the hwc format (height, width, channels), 
         // to emulate the behavior of the original Python code (w/ numpy).
+        /** @type {Float32Array} */
         let pixelData = Float32Array.from(image.data);
         let imgDims = [image.height, image.width, image.channels];
 

package/src/base/processing_utils.js

@@ -28,6 +28,7 @@ import { getModelJSON } from '../utils/hub.js';
 /**
  * @typedef {Object} ProcessorProperties Additional processor-specific properties.
  * @typedef {import('../utils/hub.js').PretrainedOptions & ProcessorProperties} PretrainedProcessorOptions
+ * @typedef {import('../tokenizers.js').PreTrainedTokenizer} PreTrainedTokenizer
  */
 
 
@@ -61,7 +62,7 @@ export class Processor extends Callable {
     }
 
     /**
-     * @returns {import('../tokenizers.js').PreTrainedTokenizer|undefined} The tokenizer of the processor, if it exists.
+     * @returns {PreTrainedTokenizer|undefined} The tokenizer of the processor, if it exists.
      */
     get tokenizer() {
         return this.components.tokenizer;
@@ -74,6 +75,11 @@ export class Processor extends Callable {
         return this.components.feature_extractor;
     }
 
+    /**
+     * @param {Parameters<PreTrainedTokenizer['apply_chat_template']>[0]} messages
+     * @param {Parameters<PreTrainedTokenizer['apply_chat_template']>[1]} options
+     * @returns {ReturnType<PreTrainedTokenizer['apply_chat_template']>}
+     */
     apply_chat_template(messages, options = {}) {
         if (!this.tokenizer) {
             throw new Error('Unable to apply chat template without a tokenizer.');
@@ -84,6 +90,10 @@ export class Processor extends Callable {
         });
     }
 
+    /**
+     * @param {Parameters<PreTrainedTokenizer['batch_decode']>} args
+     * @returns {ReturnType<PreTrainedTokenizer['batch_decode']>}
+     */
     batch_decode(...args) {
         if (!this.tokenizer) {
             throw new Error('Unable to decode without a tokenizer.');
@@ -91,6 +101,17 @@ export class Processor extends Callable {
         return this.tokenizer.batch_decode(...args);
     }
 
+    /**
+     * @param {Parameters<PreTrainedTokenizer['decode']>} args
+     * @returns {ReturnType<PreTrainedTokenizer['decode']>}
+     */
+    decode(...args) {
+        if (!this.tokenizer) {
+            throw new Error('Unable to decode without a tokenizer.');
+        }
+        return this.tokenizer.decode(...args);
+    }
+
 
     /**
      * Calls the feature_extractor function with the given input.
@@ -111,8 +132,8 @@ export class Processor extends Callable {
     /**
      * Instantiate one of the processor classes of the library from a pretrained model.
      * 
-     * The processor class to instantiate is selected based on the `feature_extractor_type` property of the config object
-     * (either passed as an argument or loaded from `pretrained_model_name_or_path` if possible)
+     * The processor class to instantiate is selected based on the `image_processor_type` (or `feature_extractor_type`; legacy)
+     * property of the config object (either passed as an argument or loaded from `pretrained_model_name_or_path` if possible)
      * 
      * @param {string} pretrained_model_name_or_path The name or path of the pretrained model. Can be either:
      * - A string, the *model id* of a pretrained processor hosted inside a model repo on huggingface.co.

package/src/configs.js

@@ -70,15 +70,19 @@ function getNormalizedConfig(config) {
         case 'florence2':
         case 'llava_onevision':
         case 'idefics3':
+            // @ts-expect-error TS2339
             init_normalized_config = getNormalizedConfig(config.text_config);
             break;
         case 'moondream1':
+            // @ts-expect-error TS2339
             init_normalized_config = getNormalizedConfig(config.phi_config);
             break;
         case 'musicgen':
+            // @ts-expect-error TS2339
             init_normalized_config = getNormalizedConfig(config.decoder);
             break;
         case 'multi_modality':
+            // @ts-expect-error TS2339
             init_normalized_config = getNormalizedConfig(config.language_config);
             break;
 
@@ -199,6 +203,7 @@ function getNormalizedConfig(config) {
             break;
 
         case 'vision-encoder-decoder':
+            // @ts-expect-error TS2339
             const decoderConfig = getNormalizedConfig(config.decoder);
 
             const add_encoder_pkv = 'num_decoder_layers' in decoderConfig;

package/src/env.js

@@ -26,7 +26,7 @@ import fs from 'fs';
 import path from 'path';
 import url from 'url';
 
-const VERSION = '3.2.3';
+const VERSION = '3.3.0';
 
 // Check if various APIs are available (depends on environment)
 const IS_BROWSER_ENV = typeof window !== "undefined" && typeof window.document !== "undefined";
@@ -160,4 +160,3 @@ export const env = {
 function isEmpty(obj) {
     return Object.keys(obj).length === 0;
 }
-

package/src/generation/streamers.js

@@ -37,6 +37,7 @@ export class TextStreamer extends BaseStreamer {
      * @param {import('../tokenizers.js').PreTrainedTokenizer} tokenizer
      * @param {Object} options
      * @param {boolean} [options.skip_prompt=false] Whether to skip the prompt tokens
+     * @param {boolean} [options.skip_special_tokens=true] Whether to skip special tokens when decoding
      * @param {function(string): void} [options.callback_function=null] Function to call when a piece of text is ready to display
      * @param {function(bigint[]): void} [options.token_callback_function=null] Function to call when a new token is generated
      * @param {Object} [options.decode_kwargs={}] Additional keyword arguments to pass to the tokenizer's decode method
@@ -45,6 +46,7 @@ export class TextStreamer extends BaseStreamer {
         skip_prompt = false,
         callback_function = null,
         token_callback_function = null,
+        skip_special_tokens = true,
         decode_kwargs = {},
         ...kwargs
     } = {}) {
@@ -53,7 +55,7 @@ export class TextStreamer extends BaseStreamer {
         this.skip_prompt = skip_prompt;
         this.callback_function = callback_function ?? stdout_write;
         this.token_callback_function = token_callback_function;
-        this.decode_kwargs = { ...decode_kwargs, ...kwargs };
+        this.decode_kwargs = { skip_special_tokens, ...decode_kwargs, ...kwargs };
 
         // variables used in the streaming process
         this.token_cache = [];
@@ -169,9 +171,10 @@ export class WhisperTextStreamer extends TextStreamer {
     } = {}) {
         super(tokenizer, {
             skip_prompt,
+            skip_special_tokens,
             callback_function,
             token_callback_function,
-            decode_kwargs: { skip_special_tokens, ...decode_kwargs },
+            decode_kwargs,
         });
         this.timestamp_begin = tokenizer.timestamp_begin;
 

package/src/models/auto/feature_extraction_auto.js

@@ -6,22 +6,6 @@ import * as AllFeatureExtractors from '../feature_extractors.js';
 
 export class AutoFeatureExtractor {
 
-    /**
-     * Instantiate one of the feature extractor classes of the library from a pretrained model.
-     * 
-     * The processor class to instantiate is selected based on the `feature_extractor_type` property of
-     * the config object (either passed as an argument or loaded from `pretrained_model_name_or_path` if possible)
-     * 
-     * @param {string} pretrained_model_name_or_path The name or path of the pretrained model. Can be either:
-     * - A string, the *model id* of a pretrained processor hosted inside a model repo on huggingface.co.
-     *   Valid model ids can be located at the root-level, like `bert-base-uncased`, or namespaced under a
-     *   user or organization name, like `dbmdz/bert-base-german-cased`.
-     * - A path to a *directory* containing processor files, e.g., `./my_model_directory/`.
-     * @param {import('../../utils/hub.js').PretrainedOptions} options Additional options for loading the processor.
-     * 
-     * @returns {Promise<AllFeatureExtractors.ImageProcessor>} A new instance of the Processor class.
-     */
-
     /** @type {typeof FeatureExtractor.from_pretrained} */
     static async from_pretrained(pretrained_model_name_or_path, options={}) {
 

package/src/models/auto/processing_auto.js

@@ -40,22 +40,6 @@ import * as AllFeatureExtractors from '../feature_extractors.js';
  */
 export class AutoProcessor {
 
-    /**
-     * Instantiate one of the processor classes of the library from a pretrained model.
-     * 
-     * The processor class to instantiate is selected based on the `image_processor_type` (or `feature_extractor_type`; legacy)
-     * property of the config object (either passed as an argument or loaded from `pretrained_model_name_or_path` if possible)
-     * 
-     * @param {string} pretrained_model_name_or_path The name or path of the pretrained model. Can be either:
-     * - A string, the *model id* of a pretrained processor hosted inside a model repo on huggingface.co.
-     *   Valid model ids can be located at the root-level, like `bert-base-uncased`, or namespaced under a
-     *   user or organization name, like `dbmdz/bert-base-german-cased`.
-     * - A path to a *directory* containing processor files, e.g., `./my_model_directory/`.
-     * @param {import('../../utils/hub.js').PretrainedOptions} options Additional options for loading the processor.
-     * 
-     * @returns {Promise<Processor>} A new instance of the Processor class.
-     */
-
     /** @type {typeof Processor.from_pretrained} */
     static async from_pretrained(pretrained_model_name_or_path, options={}) {
 

package/src/models/convnext/image_processing_convnext.js

@@ -9,6 +9,7 @@ export class ConvNextImageProcessor extends ImageProcessor {
         /**
          * Percentage of the image to crop. Only has an effect if this.size < 384.
          */
+        // @ts-expect-error TS2339
         this.crop_pct = this.config.crop_pct ?? (224 / 256);
     }
 

package/src/models/efficientnet/image_processing_efficientnet.js

@@ -5,6 +5,7 @@ import {
 export class EfficientNetImageProcessor extends ImageProcessor {
     constructor(config) {
         super(config);
+        // @ts-expect-error TS2339
         this.include_top = this.config.include_top ?? true;
         if (this.include_top) {
             this.image_std = this.image_std.map(x => x * x);

package/src/models/florence2/processing_florence2.js

@@ -10,8 +10,11 @@ export class Florence2Processor extends Processor {
         super(config, components);
 
         const {
+            // @ts-expect-error TS2339
             tasks_answer_post_processing_type,
+            // @ts-expect-error TS2339
             task_prompts_without_inputs,
+            // @ts-expect-error TS2339
             task_prompts_with_input,
         } = this.image_processor.config;
 

package/src/models/grounding_dino/image_processing_grounding_dino.js

@@ -0,0 +1,29 @@
+
+import { 
+    ImageProcessor,
+} from "../../base/image_processors_utils.js";
+import { ones } from '../../utils/tensor.js';
+
+
+/**
+ * @typedef {object} GroundingDinoFeatureExtractorResultProps
+ * @property {import('../../utils/tensor.js').Tensor} pixel_mask
+ * @typedef {import('../../base/image_processors_utils.js').ImageProcessorResult & GroundingDinoFeatureExtractorResultProps} GroundingDinoFeatureExtractorResult
+ */
+
+export class GroundingDinoImageProcessor extends ImageProcessor {
+    /**
+     * Calls the feature extraction process on an array of images, preprocesses
+     * each image, and concatenates the resulting features into a single Tensor.
+     * @param {import('../../utils/image.js').RawImage[]} images The image(s) to extract features from.
+     * @returns {Promise<GroundingDinoFeatureExtractorResult>} An object containing the concatenated pixel values of the preprocessed images.
+     */
+    async _call(images) {
+        const result = await super._call(images);
+
+        const dims = result.pixel_values.dims;
+        const pixel_mask = ones([dims[0], dims[2], dims[3]]);
+
+        return { ...result, pixel_mask };
+    }
+}

package/src/models/grounding_dino/processing_grounding_dino.js

@@ -0,0 +1,101 @@
+import { Processor } from "../../base/processing_utils.js";
+import { AutoImageProcessor } from "../auto/image_processing_auto.js";
+import { AutoTokenizer } from "../../tokenizers.js";
+import { center_to_corners_format } from "../../base/image_processors_utils.js";
+
+/**
+ * Get token ids of phrases from posmaps and input_ids.
+ * @param {import('../../utils/tensor.js').Tensor} posmaps A boolean tensor of unbatched text-thresholded logits related to the detected bounding boxes of shape `(hidden_size, )`.
+ * @param {import('../../utils/tensor.js').Tensor} input_ids A tensor of token ids of shape `(sequence_length, )`.
+ */
+function get_phrases_from_posmap(posmaps, input_ids) {
+
+    const left_idx = 0;
+    const right_idx = posmaps.dims.at(-1) - 1;
+
+    const posmaps_list = posmaps.tolist();
+    posmaps_list.fill(false, 0, left_idx + 1);
+    posmaps_list.fill(false, right_idx);
+
+    const input_ids_list = input_ids.tolist();
+    return posmaps_list
+        .map((val, idx) => val ? idx : null)
+        .filter(idx => idx !== null)
+        .map(i => input_ids_list[i]);
+}
+
+export class GroundingDinoProcessor extends Processor {
+    static tokenizer_class = AutoTokenizer
+    static image_processor_class = AutoImageProcessor
+
+    /**
+     * @typedef {import('../../utils/image.js').RawImage} RawImage
+     */
+    /**
+     * 
+     * @param {RawImage|RawImage[]|RawImage[][]} images  
+     * @param {string|string[]} text 
+     * @returns {Promise<any>}
+     */
+    async _call(images, text, options = {}) {
+
+        const image_inputs = images ? await this.image_processor(images, options) : {};
+        const text_inputs = text ? this.tokenizer(text, options) : {};
+
+        return {
+            ...text_inputs,
+            ...image_inputs,
+        }
+    }
+    post_process_grounded_object_detection(outputs, input_ids, {
+        box_threshold = 0.25,
+        text_threshold = 0.25,
+        target_sizes = null
+    } = {}) {
+        const { logits, pred_boxes } = outputs;
+        const batch_size = logits.dims[0];
+
+        if (target_sizes !== null && target_sizes.length !== batch_size) {
+            throw Error("Make sure that you pass in as many target sizes as the batch dimension of the logits")
+        }
+        const num_queries = logits.dims.at(1);
+
+        const probs = logits.sigmoid(); // (batch_size, num_queries, 256)
+        const scores = probs.max(-1).tolist(); // (batch_size, num_queries)
+
+        // Convert to [x0, y0, x1, y1] format
+        const boxes = pred_boxes.tolist() // (batch_size, num_queries, 4)
+            .map(batch => batch.map(box => center_to_corners_format(box)));
+
+        const results = [];
+        for (let i = 0; i < batch_size; ++i) {
+            const target_size = target_sizes !== null ? target_sizes[i] : null;
+
+            // Convert from relative [0, 1] to absolute [0, height] coordinates
+            if (target_size !== null) {
+                boxes[i] = boxes[i].map(box => box.map((x, j) => x * target_size[(j + 1) % 2]));
+            }
+
+            const batch_scores = scores[i];
+            const final_scores = [];
+            const final_phrases = [];
+            const final_boxes = [];
+            for (let j = 0; j < num_queries; ++j) {
+                const score = batch_scores[j];
+                if (score <= box_threshold) {
+                    continue;
+                }
+                const box = boxes[i][j];
+                const prob = probs[i][j];
+
+                final_scores.push(score);
+                final_boxes.push(box);
+
+                const phrases = get_phrases_from_posmap(prob.gt(text_threshold), input_ids[i]);
+                final_phrases.push(phrases);
+            }
+            results.push({ scores: final_scores, boxes: final_boxes, labels: this.batch_decode(final_phrases) });
+        }
+        return results;
+    }
+}

package/src/models/idefics3/image_processing_idefics3.js

@@ -146,6 +146,8 @@ export class Idefics3ImageProcessor extends ImageProcessor {
 
                     const start_offset = i * pixel_attention_mask_stride + num_patches * h * w;
                     const end_offset = (i + 1) * pixel_attention_mask_stride;
+
+                    // @ts-expect-error
                     pixel_attention_mask_data.fill(false, start_offset, end_offset);
                 }
             }

package/src/models/image_processors.js

@@ -10,6 +10,7 @@ export * from './donut/image_processing_donut.js'
 export * from './dpt/image_processing_dpt.js'
 export * from './efficientnet/image_processing_efficientnet.js'
 export * from './glpn/image_processing_glpn.js'
+export * from './grounding_dino/image_processing_grounding_dino.js'
 export * from './idefics3/image_processing_idefics3.js'
 export * from './janus/image_processing_janus.js'
 export * from './jina_clip/image_processing_jina_clip.js'

package/src/models/janus/image_processing_janus.js

@@ -13,6 +13,7 @@ export class VLMImageProcessor extends ImageProcessor {
             },
             ...config,
         });
+        // @ts-expect-error TS2339
         this.constant_values = this.config.background_color.map(x => x * this.rescale_factor)
     }
 

package/src/models/mgp_str/processing_mgp_str.js

@@ -119,6 +119,8 @@ export class MgpstrProcessor extends Processor {
      * - bpe_preds: The list of BPE decoded sentences.
      * - wp_preds: The list of wp decoded sentences.
      */
+    // @ts-expect-error The type of this method is not compatible with the one
+    // in the base class. It might be a good idea to fix this.
     batch_decode([char_logits, bpe_logits, wp_logits]) {
         const [char_preds, char_scores] = this._decode_helper(char_logits, 'char');
         const [bpe_preds, bpe_scores] = this._decode_helper(bpe_logits, 'bpe');

package/src/models/paligemma/processing_paligemma.js

@@ -41,6 +41,7 @@ export class PaliGemmaProcessor extends Processor {
         }
 
         const bos_token = this.tokenizer.bos_token;
+        // @ts-expect-error TS2339
         const image_seq_length = this.image_processor.config.image_seq_length;
         let input_strings;
         if (text.some((t) => t.includes(IMAGE_TOKEN))) {

package/src/models/phi3_v/processing_phi3_v.js

@@ -14,7 +14,7 @@ export class Phi3VProcessor extends Processor {
      * 
      * @param {string|string[]} text 
      * @param {RawImage|RawImage[]} images 
-     * @param  {...any} args 
+     * @param  { { padding?: boolean, truncation?: boolean, num_crops?: number } | undefined } options
      * @returns {Promise<any>}
      */
     async _call(text, images = null, {

package/src/models/processors.js

@@ -1,9 +1,10 @@
 export * from './florence2/processing_florence2.js';
-export * from './mgp_str/processing_mgp_str.js';
-export * from './moonshine/processing_moonshine.js';
+export * from './grounding_dino/processing_grounding_dino.js';
 export * from './idefics3/processing_idefics3.js';
 export * from './janus/processing_janus.js';
 export * from './jina_clip/processing_jina_clip.js';
+export * from './mgp_str/processing_mgp_str.js';
+export * from './moonshine/processing_moonshine.js';
 export * from './owlvit/processing_owlvit.js';
 export * from './phi3_v/processing_phi3_v.js';
 export * from './paligemma/processing_paligemma.js';

package/src/models/pyannote/feature_extraction_pyannote.js

@@ -52,6 +52,7 @@ export class PyAnnoteFeatureExtractor extends FeatureExtractor {
 
             let current_speaker = -1;
             for (let i = 0; i < scores.length; ++i) {
+                /** @type {number[]} */
                 const probabilities = softmax(scores[i]);
                 const [score, id] = max(probabilities);
                 const [start, end] = [i, i + 1];

package/src/models/qwen2_vl/processing_qwen2_vl.js

@@ -28,6 +28,7 @@ export class Qwen2VLProcessor extends Processor {
         }
 
         if (image_grid_thw) {
+            // @ts-expect-error TS2551
             let merge_length = this.image_processor.config.merge_size ** 2;
             let index = 0;
 

package/src/models/seamless_m4t/feature_extraction_seamless_m4t.js

@@ -133,8 +133,8 @@ export class SeamlessM4TFeatureExtractor extends FeatureExtractor {
                         'int64',
                         new BigInt64Array(numPaddedFrames),
                         [1, numPaddedFrames],
-                    )
-                    padded_attention_mask.data.fill(1n, 0, num_frames);
+                    );
+                    /** @type {BigInt64Array} */ (padded_attention_mask.data).fill(1n, 0, num_frames);
                 }
             }
         }

package/src/models/whisper/feature_extraction_whisper.js

@@ -44,7 +44,7 @@ export class WhisperFeatureExtractor extends FeatureExtractor {
         )
 
         const data = features.data;
-        const maxValue = max(data)[0];
+        const maxValue = max(/** @type {Float32Array} */(data))[0];
 
         for (let i = 0; i < data.length; ++i) {
             data[i] = (Math.max(data[i], maxValue - 8.0) + 4.0) / 4.0;