@huggingface/transformers 4.0.0-next.0 → 4.0.0-next.10

package/types/models/glm_moe_dsa/modeling_glm_moe_dsa.d.ts

@@ -0,0 +1,8 @@
+export class GlmMoeDsaPreTrainedModel extends PreTrainedModel {
+}
+export class GlmMoeDsaModel extends GlmMoeDsaPreTrainedModel {
+}
+export class GlmMoeDsaForCausalLM extends GlmMoeDsaPreTrainedModel {
+}
+import { PreTrainedModel } from '../modeling_utils.js';
+//# sourceMappingURL=modeling_glm_moe_dsa.d.ts.map

package/types/models/glm_moe_dsa/modeling_glm_moe_dsa.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"modeling_glm_moe_dsa.d.ts","sourceRoot":"","sources":["../../../src/models/glm_moe_dsa/modeling_glm_moe_dsa.js"],"names":[],"mappings":"AAEA;CAAgE;AAChE;CAA+D;AAC/D;CAAqE;gCAJrC,sBAAsB"}

package/types/models/glm_ocr/modeling_glm_ocr.d.ts

@@ -0,0 +1,26 @@
+export class GlmOcrForConditionalGeneration extends Qwen2_5_VLForConditionalGeneration {
+    /**
+     * Compute 3D positional indices for vision tokens.
+     * Temporal is constant, height is repeat-interleaved, width tiles.
+     * @param {number} start_position
+     * @param {number[]} grid_thw [T, H, W]
+     * @param {number} temp_merge_size
+     * @param {number} spatial_merge_size
+     * @returns {number[]} Flat array of length 3 * seq_len: [temporal..., height..., width...]
+     */
+    get_vision_position_ids(start_position: number, grid_thw: number[], temp_merge_size: number, spatial_merge_size: number): number[];
+    /**
+     * GlmOcr uses mm_token_type_ids-style grouping (image tokens identified by image_token_id)
+     * instead of vision_start_token_id scanning used by Qwen2VL.
+     * After a vision segment, position advances by max(h, w) / spatial_merge_size.
+     */
+    _get_multimodal_rope_positions({ filtered_ids, image_grid_thw_list, video_grid_thw_list, spatial_merge_size, state, }: {
+        filtered_ids: any;
+        image_grid_thw_list: any;
+        video_grid_thw_list: any;
+        spatial_merge_size: any;
+        state: any;
+    }): number[][];
+}
+import { Qwen2_5_VLForConditionalGeneration } from '../qwen2_5_vl/modeling_qwen2_5_vl.js';
+//# sourceMappingURL=modeling_glm_ocr.d.ts.map

package/types/models/glm_ocr/modeling_glm_ocr.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"modeling_glm_ocr.d.ts","sourceRoot":"","sources":["../../../src/models/glm_ocr/modeling_glm_ocr.js"],"names":[],"mappings":"AAEA;IACI;;;;;;;;OAQG;IACH,wCANW,MAAM,YACN,MAAM,EAAE,mBACR,MAAM,sBACN,MAAM,GACJ,MAAM,EAAE,CAgBpB;IAED;;;;OAIG;IACH;;;;;;mBA2CC;CACJ;mDA7EkD,sCAAsC"}

package/types/models/granite_speech/feature_extraction_granite_speech.d.ts

@@ -0,0 +1,16 @@
+export class GraniteSpeechFeatureExtractor extends FeatureExtractor {
+    constructor(config: any);
+    mel_filters: number[][];
+    window: Float64Array<any>;
+    /**
+     * Extract mel spectrogram features from audio, matching the Python GraniteSpeechFeatureExtractor.
+     * @param {Float32Array|Float64Array} audio The audio waveform.
+     * @returns {Promise<{input_features: Tensor}>}
+     */
+    _call(audio: Float32Array | Float64Array): Promise<{
+        input_features: Tensor;
+    }>;
+}
+import { FeatureExtractor } from '../../feature_extraction_utils.js';
+import { Tensor } from '../../utils/tensor.js';
+//# sourceMappingURL=feature_extraction_granite_speech.d.ts.map

package/types/models/granite_speech/feature_extraction_granite_speech.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"feature_extraction_granite_speech.d.ts","sourceRoot":"","sources":["../../../src/models/granite_speech/feature_extraction_granite_speech.js"],"names":[],"mappings":"AAIA;IACI,yBAsBC;IAhBG,wBAQC;IAKD,0BAAqC;IAKzC;;;;OAIG;IACH,aAHW,YAAY,GAAC,YAAY,GACvB,OAAO,CAAC;QAAC,cAAc,EAAE,MAAM,CAAA;KAAC,CAAC,CAwB7C;CACJ;iCAzDuD,mCAAmC;uBAEpE,uBAAuB"}

package/types/models/granite_speech/modeling_granite_speech.d.ts

@@ -0,0 +1,4 @@
+export class GraniteSpeechForConditionalGeneration extends UltravoxModel {
+}
+import { UltravoxModel } from '../ultravox/modeling_ultravox.js';
+//# sourceMappingURL=modeling_granite_speech.d.ts.map

package/types/models/granite_speech/modeling_granite_speech.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"modeling_granite_speech.d.ts","sourceRoot":"","sources":["../../../src/models/granite_speech/modeling_granite_speech.js"],"names":[],"mappings":"AAEA;CAEC;8BAJ6B,kCAAkC"}

package/types/models/granite_speech/processing_granite_speech.d.ts

@@ -0,0 +1,19 @@
+export class GraniteSpeechProcessor extends Processor {
+    static tokenizer_class: typeof AutoTokenizer;
+    static feature_extractor_class: typeof AutoFeatureExtractor;
+    /**
+     * Compute the number of audio tokens for a given raw audio length.
+     * @param {number} audioLength Raw audio sample count.
+     * @returns {number} Number of projector output tokens.
+     */
+    _get_num_audio_features(audioLength: number): number;
+    /**
+     * @param {string} text The text input to process.
+     * @param {Float32Array} audio The audio input to process.
+     */
+    _call(text: string, audio?: Float32Array, kwargs?: {}): Promise<any>;
+}
+import { Processor } from '../../processing_utils.js';
+import { AutoTokenizer } from '../auto/tokenization_auto.js';
+import { AutoFeatureExtractor } from '../auto/feature_extraction_auto.js';
+//# sourceMappingURL=processing_granite_speech.d.ts.map

package/types/models/granite_speech/processing_granite_speech.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"processing_granite_speech.d.ts","sourceRoot":"","sources":["../../../src/models/granite_speech/processing_granite_speech.js"],"names":[],"mappings":"AAKA;IACI,6CAAuC;IACvC,4DAAsD;IAGtD;;;;OAIG;IACH,qCAHW,MAAM,GACJ,MAAM,CAUlB;IAED;;;OAGG;IACH,YAHW,MAAM,UACN,YAAY,6BAiCtB;CACJ;0BA3DyB,2BAA2B;8BADvB,8BAA8B;qCADvB,oCAAoC"}

package/types/models/grounding_dino/image_processing_grounding_dino.d.ts

@@ -1,5 +1,5 @@
 /**
- * @typedef {object} GroundingDinoFeatureExtractorResultProps
+ * @typedef {Object} GroundingDinoFeatureExtractorResultProps
  * @property {import('../../utils/tensor.js').Tensor} pixel_mask
  * @typedef {import('../../image_processors_utils.js').ImageProcessorResult & GroundingDinoFeatureExtractorResultProps} GroundingDinoFeatureExtractorResult
  */

package/types/models/idefics3/modeling_idefics3.d.ts

@@ -1,23 +1,7 @@
-export class Idefics3PreTrainedModel extends PreTrainedModel {
-}
 /**
  * The Idefics3 model which consists of a vision backbone and a language model.
  */
-export class Idefics3ForConditionalGeneration extends Idefics3PreTrainedModel {
-    encode_image({ pixel_values, pixel_attention_mask }: {
-        pixel_values: any;
-        pixel_attention_mask: any;
-    }): Promise<any>;
-    _merge_input_ids_with_image_features(kwargs: any): {
-        inputs_embeds: any;
-        attention_mask: any;
-    };
-}
-/**
- * The SmolVLM Model with a language modeling head.
- * It is made up a SigLIP vision encoder, with a language modeling head on top.
- */
-export class SmolVLMForConditionalGeneration extends Idefics3ForConditionalGeneration {
+export class Idefics3ForConditionalGeneration extends LlavaForConditionalGeneration {
 }
-import { PreTrainedModel } from '../modeling_utils.js';
+import { LlavaForConditionalGeneration } from '../llava/modeling_llava.js';
 //# sourceMappingURL=modeling_idefics3.d.ts.map

package/types/models/idefics3/modeling_idefics3.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"modeling_idefics3.d.ts","sourceRoot":"","sources":["../../../src/models/idefics3/modeling_idefics3.js"],"names":[],"mappings":"AAGA;CASC;AAED;;GAEG;AACH;IACI;;;qBAIC;IAED;;;MAUC;CACJ;AAED;;;GAGG;AACH;CAAwF;gCAzCX,sBAAsB"}
+{"version":3,"file":"modeling_idefics3.d.ts","sourceRoot":"","sources":["../../../src/models/idefics3/modeling_idefics3.js"],"names":[],"mappings":"AAEA;;GAEG;AACH;CASC;8CAd6C,4BAA4B"}

package/types/models/image_processors.d.ts

@@ -1,6 +1,7 @@
 export * from "./beit/image_processing_beit.js";
 export * from "./bit/image_processing_bit.js";
 export * from "./chinese_clip/image_processing_chinese_clip.js";
+export * from "./chmv2/image_processing_chmv2.js";
 export * from "./clip/image_processing_clip.js";
 export * from "./convnext/image_processing_convnext.js";
 export * from "./deit/image_processing_deit.js";
@@ -9,11 +10,14 @@ export * from "./dinov3_vit/image_processing_dinov3_vit.js";
 export * from "./donut/image_processing_donut.js";
 export * from "./dpt/image_processing_dpt.js";
 export * from "./efficientnet/image_processing_efficientnet.js";
+export * from "./gemma3/image_processing_gemma3.js";
+export * from "./glm46v/image_processing_glm46v.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";
+export * from "./lfm2_vl/image_processing_lfm2_vl.js";
 export * from "./llava_onevision/image_processing_llava_onevision.js";
 export * from "./mask2former/image_processing_mask2former.js";
 export * from "./maskformer/image_processing_maskformer.js";

package/types/models/lfm2_vl/image_processing_lfm2_vl.d.ts

@@ -0,0 +1,41 @@
+export class Lfm2VlImageProcessor extends ImageProcessor {
+    constructor(config: Record<string, any>);
+    downsample_factor: any;
+    do_image_splitting: any;
+    min_tiles: any;
+    max_tiles: any;
+    use_thumbnail: any;
+    min_image_tokens: any;
+    max_image_tokens: any;
+    encoder_patch_size: any;
+    tile_size: any;
+    max_pixels_tolerance: any;
+    return_row_col_info: any;
+    max_num_patches: number;
+    /**
+     * Check if the image is too large to be processed as a single tile.
+     * @param {number} height
+     * @param {number} width
+     * @returns {boolean}
+     */
+    _is_image_too_large(height: number, width: number): boolean;
+    /**
+     * Get the grid layout for tiling a large image.
+     * @param {number} height
+     * @param {number} width
+     * @returns {{ grid_width: number, grid_height: number, target_width: number, target_height: number }}
+     */
+    _get_grid_layout(height: number, width: number): {
+        grid_width: number;
+        grid_height: number;
+        target_width: number;
+        target_height: number;
+    };
+    /** @param {RawImage|RawImage[]|RawImage[][]} images */
+    _call(images: RawImage | RawImage[] | RawImage[][], { return_row_col_info }?: {
+        return_row_col_info?: any;
+    }): Promise<Record<string, any>>;
+}
+export type RawImage = import("../../utils/image.js").RawImage;
+import { ImageProcessor } from '../../image_processors_utils.js';
+//# sourceMappingURL=image_processing_lfm2_vl.d.ts.map

package/types/models/lfm2_vl/image_processing_lfm2_vl.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"image_processing_lfm2_vl.d.ts","sourceRoot":"","sources":["../../../src/models/lfm2_vl/image_processing_lfm2_vl.js"],"names":[],"mappings":"AAiIA;IACI,oBAAuB,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,EAiBzC;IAfG,uBAAsD;IACtD,wBAA2D;IAC3D,eAAsC;IACtC,eAAuC;IACvC,mBAAiD;IACjD,sBAAqD;IACrD,sBAAsD;IACtD,wBAA8E;IAC9E,eAAwC;IACxC,0BAA8D;IAC9D,yBAA8D;IAI9D,wBAAyE;IAG7E;;;;;OAKG;IACH,4BAJW,MAAM,SACN,MAAM,GACJ,OAAO,CAUnB;IAED;;;;;OAKG;IACH,yBAJW,MAAM,SACN,MAAM,GACJ;QAAE,UAAU,EAAE,MAAM,CAAC;QAAC,WAAW,EAAE,MAAM,CAAC;QAAC,YAAY,EAAE,MAAM,CAAC;QAAC,aAAa,EAAE,MAAM,CAAA;KAAE,CAiBpG;IAED,uDAAuD;IAEvD,cAFY,QAAQ,GAAC,QAAQ,EAAE,GAAC,QAAQ,EAAE,EAAE;;qCAmH3C;CACJ;uBA5SY,OAAO,sBAAsB,EAAE,QAAQ;+BAJP,iCAAiC"}

package/types/models/lfm2_vl/modeling_lfm2_vl.d.ts

@@ -0,0 +1,4 @@
+export class Lfm2VlForConditionalGeneration extends LlavaForConditionalGeneration {
+}
+import { LlavaForConditionalGeneration } from '../llava/modeling_llava.js';
+//# sourceMappingURL=modeling_lfm2_vl.d.ts.map

package/types/models/lfm2_vl/modeling_lfm2_vl.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"modeling_lfm2_vl.d.ts","sourceRoot":"","sources":["../../../src/models/lfm2_vl/modeling_lfm2_vl.js"],"names":[],"mappings":"AAEA;CAUC;8CAZ6C,4BAA4B"}

package/types/models/lfm2_vl/processing_lfm2_vl.d.ts

@@ -0,0 +1,18 @@
+/**
+ * @typedef {import('../../utils/image.js').RawImage} RawImage
+ */
+export class Lfm2VlProcessor extends Processor {
+    static tokenizer_class: typeof AutoTokenizer;
+    static image_processor_class: typeof AutoImageProcessor;
+    /**
+     * @param {RawImage|RawImage[]} images
+     * @param {string|string[]|null} [text]
+     * @param {Record<string, any>} [kwargs]
+     */
+    _call(images: RawImage | RawImage[], text?: string | string[] | null, kwargs?: Record<string, any>): Promise<any>;
+}
+export type RawImage = import("../../utils/image.js").RawImage;
+import { Processor } from '../../processing_utils.js';
+import { AutoTokenizer } from '../auto/tokenization_auto.js';
+import { AutoImageProcessor } from '../auto/image_processing_auto.js';
+//# sourceMappingURL=processing_lfm2_vl.d.ts.map

package/types/models/lfm2_vl/processing_lfm2_vl.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"processing_lfm2_vl.d.ts","sourceRoot":"","sources":["../../../src/models/lfm2_vl/processing_lfm2_vl.js"],"names":[],"mappings":"AAIA;;GAEG;AAEH;IACI,6CAAuC;IACvC,wDAAkD;IAElD;;;;OAIG;IACH,cAJW,QAAQ,GAAC,QAAQ,EAAE,SACnB,MAAM,GAAC,MAAM,EAAE,GAAC,IAAI,WACpB,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,gBA4D7B;CACJ;uBAvEY,OAAO,sBAAsB,EAAE,QAAQ;0BAL1B,2BAA2B;8BAEvB,8BAA8B;mCADzB,kCAAkC"}

package/types/models/lighton_ocr/modeling_lighton_ocr.d.ts

@@ -0,0 +1,4 @@
+export class LightOnOcrForConditionalGeneration extends LlavaForConditionalGeneration {
+}
+import { LlavaForConditionalGeneration } from '../llava/modeling_llava.js';
+//# sourceMappingURL=modeling_lighton_ocr.d.ts.map

package/types/models/lighton_ocr/modeling_lighton_ocr.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"modeling_lighton_ocr.d.ts","sourceRoot":"","sources":["../../../src/models/lighton_ocr/modeling_lighton_ocr.js"],"names":[],"mappings":"AAEA;CAAwF;8CAF1C,4BAA4B"}

package/types/models/marian/tokenization_marian.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"tokenization_marian.d.ts","sourceRoot":"","sources":["../../../src/models/marian/tokenization_marian.js"],"names":[],"mappings":"AAGA;;;GAGG;AACH;IASQ,sBAAqC;IAErC,mCAA6G;CAoCpH;oCAtDmC,6BAA6B"}
+{"version":3,"file":"tokenization_marian.d.ts","sourceRoot":"","sources":["../../../src/models/marian/tokenization_marian.js"],"names":[],"mappings":"AAIA;;;GAGG;AACH;IASQ,sBAAqC;IAErC,mCAA6G;CAoCpH;oCAvDmC,6BAA6B"}

package/types/models/mistral3/modeling_mistral3.d.ts

@@ -1,4 +1,4 @@
-export class Mistral3ForConditionalGeneration extends LlavaQwen2ForCausalLM {
+export class Mistral3ForConditionalGeneration extends LlavaForConditionalGeneration {
 }
-import { LlavaQwen2ForCausalLM } from '../llava/modeling_llava.js';
+import { LlavaForConditionalGeneration } from '../llava/modeling_llava.js';
 //# sourceMappingURL=modeling_mistral3.d.ts.map

package/types/models/mistral3/modeling_mistral3.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"modeling_mistral3.d.ts","sourceRoot":"","sources":["../../../src/models/mistral3/modeling_mistral3.js"],"names":[],"mappings":"AAEA;CAA8E;sCAFxC,4BAA4B"}
+{"version":3,"file":"modeling_mistral3.d.ts","sourceRoot":"","sources":["../../../src/models/mistral3/modeling_mistral3.js"],"names":[],"mappings":"AAEA;CAAsF;8CAFxC,4BAA4B"}

package/types/models/mistral4/modeling_mistral4.d.ts

@@ -0,0 +1,8 @@
+export class Mistral4PreTrainedModel extends PreTrainedModel {
+}
+export class Mistral4Model extends Mistral4PreTrainedModel {
+}
+export class Mistral4ForCausalLM extends Mistral4PreTrainedModel {
+}
+import { PreTrainedModel } from '../modeling_utils.js';
+//# sourceMappingURL=modeling_mistral4.d.ts.map

package/types/models/mistral4/modeling_mistral4.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"modeling_mistral4.d.ts","sourceRoot":"","sources":["../../../src/models/mistral4/modeling_mistral4.js"],"names":[],"mappings":"AAEA;CAA+D;AAC/D;CAA6D;AAC7D;CAAmE;gCAJnC,sBAAsB"}

package/types/models/modeling_utils.d.ts

@@ -10,6 +10,18 @@ export function registerTaskMappings(mappings: any): void;
  * @private
  */
 export function boolTensor(value: boolean): Tensor;
+/**
+ * Get the session configuration for a given model type.
+ * @param {number} modelType The model type enum value.
+ * @param {Object} config The model config.
+ * @param {Object} [options] Loading options.
+ * @returns {{ sessions: Record<string, string>, cache_sessions?: Record<string, true>, optional_configs?: Record<string, string> }}
+ */
+export function getSessionsConfig(modelType: number, config: any, options?: any): {
+    sessions: Record<string, string>;
+    cache_sessions?: Record<string, true>;
+    optional_configs?: Record<string, string>;
+};
 /**
  * Perform forward pass on the seq2seq model (both encoder and decoder).
  * @param {Object} self The seq2seq model object.
@@ -41,28 +53,28 @@ export function decoder_forward(self: any, model_inputs: any, is_encoder_decoder
  * @param {Object} params Additional parameters.
  * @param {Function} [params.encode_function] The function to encode the modality values.
  * @param {Function} [params.merge_function] The function to merge the modality features with the input embeddings.
- * @param {string} [params.modality_input_name] The modality input name.
+ * @param {string[]} [params.modality_input_names] The modality input name.
  * @param {string} [params.modality_output_name] The modality output name.
  * @param {Tensor} [params.input_ids=null]
  * @param {Tensor} [params.attention_mask=null]
  * @param {Tensor} [params.position_ids=null]
  * @param {Tensor} [params.inputs_embeds=null]
- * @param {Tensor} [params.past_key_values=null]
+ * @param {DynamicCache} [params.past_key_values=null]
  * @param {Object} [params.generation_config=null]
  * @param {Object} [params.logits_processor=null]
  * @returns {Promise<Tensor>} The model's output tensor
  * @private
  */
-export function generic_text_to_text_forward(self: any, { encode_function, merge_function, modality_input_name, modality_output_name, input_ids, attention_mask, position_ids, inputs_embeds, past_key_values, generation_config, logits_processor, ...kwargs }: {
+export function generic_text_to_text_forward(self: any, { encode_function, merge_function, modality_input_names, modality_output_name, input_ids, attention_mask, position_ids, inputs_embeds, past_key_values, generation_config, logits_processor, ...kwargs }: {
     encode_function?: Function;
     merge_function?: Function;
-    modality_input_name?: string;
+    modality_input_names?: string[];
     modality_output_name?: string;
     input_ids?: Tensor;
     attention_mask?: Tensor;
     position_ids?: Tensor;
     inputs_embeds?: Tensor;
-    past_key_values?: Tensor;
+    past_key_values?: DynamicCache;
     generation_config?: any;
     logits_processor?: any;
 }): Promise<Tensor>;
@@ -168,6 +180,7 @@ export namespace MODEL_TYPES {
     let ImageAudioTextToText: number;
     let Supertonic: number;
     let Chatterbox: number;
+    let VoxtralRealtime: number;
 }
 export const MODEL_TYPE_MAPPING: Map<any, any>;
 export const MODEL_NAME_TO_CLASS_MAPPING: Map<any, any>;
@@ -209,8 +222,8 @@ export class PreTrainedModel extends PreTrainedModel_base {
     config: import("../configs.js").PretrainedConfig;
     sessions: Record<string, any>;
     configs: Record<string, any>;
-    can_generate: boolean;
-    _forward: typeof decoder_forward | typeof auto_encoder_forward | typeof encoder_forward;
+    can_generate: any;
+    _forward: any;
     _prepare_inputs_for_generation: any;
     /** @type {import('../configs.js').TransformersJSConfig} */
     custom_config: import("../configs.js").TransformersJSConfig;
@@ -257,9 +270,9 @@ export class PreTrainedModel extends PreTrainedModel_base {
     /**
      *
      * @param {GenerationConfig} generation_config
-     * @param {StoppingCriteriaList} [stopping_criteria=null]
+     * @param {import('../generation/stopping_criteria.js').StoppingCriteria|import('../generation/stopping_criteria.js').StoppingCriteria[]|StoppingCriteriaList} [stopping_criteria=null]
      */
-    _get_stopping_criteria(generation_config: GenerationConfig, stopping_criteria?: StoppingCriteriaList): StoppingCriteriaList;
+    _get_stopping_criteria(generation_config: GenerationConfig, stopping_criteria?: import("../generation/stopping_criteria.js").StoppingCriteria | import("../generation/stopping_criteria.js").StoppingCriteria[] | StoppingCriteriaList): StoppingCriteriaList;
     /**
      * Confirms that the model class is compatible with generation.
      * If not, raises an exception that points to the right class to use.
@@ -287,7 +300,7 @@ export class PreTrainedModel extends PreTrainedModel_base {
      * @param {Tensor} [params.inputs=null]
      * @param {number} [params.bos_token_id=null]
      * @param {Record<string, Tensor|number[]>} [params.model_kwargs]
-     * @returns {{inputs_tensor: Tensor, model_inputs: Record<string, Tensor>, model_input_name: string}} The model-specific inputs for generation.
+     * @returns {{inputs_tensor: Tensor, model_inputs: Record<string, Tensor> & {past_key_values?: DynamicCache}, model_input_name: string}} The model-specific inputs for generation.
      */
     _prepare_model_inputs({ inputs, bos_token_id, model_kwargs }: {
         inputs?: Tensor;
@@ -295,7 +308,9 @@ export class PreTrainedModel extends PreTrainedModel_base {
         model_kwargs?: Record<string, Tensor | number[]>;
     }): {
         inputs_tensor: Tensor;
-        model_inputs: Record<string, Tensor>;
+        model_inputs: Record<string, Tensor> & {
+            past_key_values?: DynamicCache;
+        };
         model_input_name: string;
     };
     _prepare_encoder_decoder_kwargs_for_generation({ inputs_tensor, model_inputs, model_input_name, generation_config, }: {
@@ -319,13 +334,14 @@ export class PreTrainedModel extends PreTrainedModel_base {
      */
     generate({ inputs, generation_config, logits_processor, stopping_criteria, streamer, ...kwargs }: import("../generation/parameters.js").GenerationFunctionParameters): Promise<ModelOutput | Tensor>;
     /**
-     * Returns an object containing past key values from the given decoder results object.
+     * Returns a DynamicCache containing past key values from the given decoder results object.
      *
      * @param {Object} decoderResults The decoder results object.
-     * @param {Object} pastKeyValues The previous past key values.
-     * @returns {Object} An object containing past key values.
+     * @param {DynamicCache} pastKeyValues The previous past key values.
+     * @param {boolean} [disposeEncoderPKVs=false] Whether to dispose encoder past key values.
+     * @returns {DynamicCache} A new DynamicCache containing the updated past key values.
      */
-    getPastKeyValues(decoderResults: any, pastKeyValues: any, disposeEncoderPKVs?: boolean): any;
+    getPastKeyValues(decoderResults: any, pastKeyValues: DynamicCache, disposeEncoderPKVs?: boolean): DynamicCache;
     /**
      * Returns an object containing attentions from the given model output object.
      *
@@ -338,22 +354,25 @@ export class PreTrainedModel extends PreTrainedModel_base {
     /**
      * Adds past key values to the decoder feeds object. If pastKeyValues is null, creates new tensors for past key values.
      *
-     * @param {Object} decoderFeeds The decoder feeds object to add past key values to.
-     * @param {Object} pastKeyValues An object containing past key values.
+     * @param {Record<string, any>} decoderFeeds The decoder feeds object to add past key values to.
+     * @param {DynamicCache|null} pastKeyValues The cache containing past key values.
      */
-    addPastKeyValues(decoderFeeds: any, pastKeyValues: any): void;
-    encode_image({ pixel_values }: {
-        pixel_values: any;
-    }): Promise<any>;
-    encode_text({ input_ids }: {
-        input_ids: any;
-    }): Promise<any>;
-    encode_audio({ audio_values }: {
-        audio_values: any;
-    }): Promise<any>;
+    addPastKeyValues(decoderFeeds: Record<string, any>, pastKeyValues: DynamicCache | null): void;
+    /**
+     * Helper function to select valid inputs and run through the appropriate encoder (vision, text, audio) based on the input type.
+     * @param {string} sessionName
+     * @param {Record<string, Tensor>} inputs
+     * @param {string} outputName
+     * @private
+     */
+    private _encode_input;
+    encode_image(inputs: any): Promise<any>;
+    encode_text(inputs: any): Promise<any>;
+    encode_audio(inputs: any): Promise<any>;
 }
 import { Tensor } from '../utils/tensor.js';
 import { Seq2SeqLMOutput } from './modeling_outputs.js';
+import { DynamicCache } from '../cache_utils.js';
 import { GenerationConfig } from '../generation/configuration_utils.js';
 import { StoppingCriteriaList } from '../generation/stopping_criteria.js';
 import { ModelOutput } from './modeling_outputs.js';

package/types/models/modeling_utils.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"modeling_utils.d.ts","sourceRoot":"","sources":["../../src/models/modeling_utils.js"],"names":[],"mappings":"AASA;;;GAGG;AACH,0DAEC;AAyDD;;;;;GAKG;AACH,kCAJW,OAAO,GACL,MAAM,CAKlB;AAotCD;;;;;;GAMG;AACH,+DAHa,OAAO,CAAC,eAAe,CAAC,CAoBpC;AAED;;;;;;GAMG;AACH,+DAHa,OAAO,KAAQ,CAgC3B;AAED,iFAIC;AAED;;;;;;GAMG;AACH,6FAHa,OAAO,KAAQ,CA2B3B;AAED;;;;;;;;;;;;;;;;;GAiBG;AACH,iQAdG;IAA0B,eAAe;IACf,cAAc;IAChB,mBAAmB,GAAnC,MAAM;IACU,oBAAoB,GAApC,MAAM;IACU,SAAS,GAAzB,MAAM;IACU,cAAc,GAA9B,MAAM;IACU,YAAY,GAA5B,MAAM;IACU,aAAa,GAA7B,MAAM;IACU,eAAe,GAA/B,MAAM;IACU,iBAAiB;IACjB,gBAAgB;CACxC,GAAU,OAAO,CAAC,MAAM,CAAC,CAsF3B;AAED;;;;;;GAMG;AACH,oEAHa,OAAO,CAAC,MAAM,CAAC,CAW3B;AAED;;;;;;GAMG;AACH,oEAHa,OAAO,CAAC,MAAM,CAAC,CAW3B;AAED;;;;;;;;GAQG;AACH,mDAHW,MAAM,yBACJ;IAAC,IAAI,EAAE,aAAa,CAAC;IAAC,IAAI,EAAE,MAAM,EAAE,CAAA;CAAC,CAsBjD;AAED;;;;;;;;;;;GAWG;AACH,4GAUC;AAED,iIA2CC;AAED,yIASC;AAED,sGAMC;AAED;;;;;;;;;EA6BC;AAED;;;;;;;;;EAcC;AAED;;;;;;;;;EAcC;AAED;;;;;;;GAOG;AACH,oEANW,MAAM,SACN,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,WACtB,OAAO,iBAAiB,EAAE,sBAAsB,GAC9C,OAAO,CAAC,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC,CAYxC;AAptDD,oCAAsC;;;;;;;;;;;;;;;;;;;AA8JtC,+CAA4C;AAC5C,wDAAqD;AACrD,wDAAqD;;KArKb,GAAG;UAClC,GAAE;;AAsKX;;GAEG;AACH;IAiDI;;;;;;;;;;;;;;OAcG;IACH,sDATW,MAAM,gKAKN,OAAO,iBAAiB,EAAE,sBAAsB,GAE9C,OAAO,CAAC,eAAe,CAAC,CAwRpC;IAjVD;;;;;OAKG;IACH,oBAJW,OAAO,eAAe,EAAE,gBAAgB,YACxC,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,WACnB,MAAM,CAAC,MAAM,MAAS,EAyBhC;IAjCD,wBAA8B;IAC9B,yBAAiD;IAEjD,mCAAqC;IAUjC,iDAAoB;IACpB,8BAAwB;IACxB,6BAAsB;IAQtB,sBAA2C;IAC3C,wFAAkC;IAClC,oCAA+D;IAM/D,2DAA2D;IAC3D,eADW,OAAO,eAAe,EAAE,oBAAoB,CACS;IAGpE;;;;OAIG;IACH,WAHa,OAAO,CAAC,OAAO,EAAE,CAAC,CAS9B;IAySD;;;;OAIG;IACH,0BAFa,OAAO,KAAQ,CAI3B;IAED;;;;;;OAMG;IACH,4BAHa,OAAO,KAAQ,CAK3B;IAED;;;OAGG;IACH,yBAFa,gBAAgB,GAAC,IAAI,CAIjC;IAED;;;;;OAKG;IACH,8BA2JC;IAED;;;;;;OAMG;IACH,8CAJW,gBAAgB,GAAC,IAAI,+CAEnB,gBAAgB,CA+B5B;IAED;;;;OAIG;IACH,0CAHW,gBAAgB,sBAChB,oBAAoB,wBAqB9B;IAED;;;OAGG;IACH,8BA2BC;IAED,mDAKC;IAED;;;;;;;;OAQG;IACH,wGANG;QAA2B,mBAAmB,EAAtC,MAAM,EAAE,EAAE;QACK,OAAO;QACP,YAAY;QACX,kBAAkB,EAAlC,OAAO;KACf,OAuBF;IAED;;;;;;;OAOG;IACH,8DALG;QAAwB,MAAM,GAAtB,MAAM;QACU,YAAY,GAA5B,MAAM;QACmC,YAAY,GAArD,MAAM,CAAC,MAAM,EAAE,MAAM,GAAC,MAAM,EAAE,CAAC;KACvC,GAAU;QAAC,aAAa,EAAE,MAAM,CAAC;QAAC,YAAY,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;QAAC,gBAAgB,EAAE,MAAM,CAAA;KAAC,CAmBnG;IAED;;;;;qBAmDC;IAED;;;OAGG;IACH,oJAFW,GAAC;;;MAwDX;IAED;;;;OAIG;IACH,kGAHW,OAAO,6BAA6B,EAAE,4BAA4B,GAChE,OAAO,CAAC,WAAW,GAAC,MAAM,CAAC,CAgOvC;IAED;;;;;;OAMG;IACH,6FAkCC;IAED;;;;;OAKG;IACH,kCAFa;QAAC,gBAAgB,CAAC,EAAE,MAAM,EAAE,CAAA;KAAC,CAgBzC;IAED;;;;;OAKG;IACH,8DAeC;IAED;;qBAGC;IAED;;qBAGC;IAED;;qBAGC;CACJ;uBA/xCmE,oBAAoB;gCAexD,uBAAuB;iCActB,sCAAsC;qCACG,oCAAoC;4BAGlF,uBAAuB"}
+{"version":3,"file":"modeling_utils.d.ts","sourceRoot":"","sources":["../../src/models/modeling_utils.js"],"names":[],"mappings":"AASA;;;GAGG;AACH,0DAEC;AA2DD;;;;;GAKG;AACH,kCAJW,OAAO,GACL,MAAM,CAKlB;AAuLD;;;;;;GAMG;AACH,6CALW,MAAM,+BAGJ;IAAE,QAAQ,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IAAC,cAAc,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC;IAAC,gBAAgB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAA;CAAE,CASlI;AAg9BD;;;;;;GAMG;AACH,+DAHa,OAAO,CAAC,eAAe,CAAC,CAyBpC;AAED;;;;;;GAMG;AACH,+DAHa,OAAO,KAAQ,CAgC3B;AAED,iFAIC;AAED;;;;;;GAMG;AACH,6FAHa,OAAO,KAAQ,CAoC3B;AAED;;;;;;;;;;;;;;;;;GAiBG;AACH,kQAdG;IAA0B,eAAe;IACf,cAAc;IACd,oBAAoB,GAAtC,MAAM,EAAE;IACQ,oBAAoB,GAApC,MAAM;IACU,SAAS,GAAzB,MAAM;IACU,cAAc,GAA9B,MAAM;IACU,YAAY,GAA5B,MAAM;IACU,aAAa,GAA7B,MAAM;IACgB,eAAe,GAArC,YAAY;IACI,iBAAiB;IACjB,gBAAgB;CACxC,GAAU,OAAO,CAAC,MAAM,CAAC,CAyG3B;AAED;;;;;;GAMG;AACH,oEAHa,OAAO,CAAC,MAAM,CAAC,CAW3B;AAED;;;;;;GAMG;AACH,oEAHa,OAAO,CAAC,MAAM,CAAC,CAW3B;AAED;;;;;;;;GAQG;AACH,mDAHW,MAAM,yBACJ;IAAC,IAAI,EAAE,aAAa,CAAC;IAAC,IAAI,EAAE,MAAM,EAAE,CAAA;CAAC,CAsBjD;AAED;;;;;;;;;;;GAWG;AACH,4GAUC;AAED,iIAkDC;AAED,yIASC;AAED,sGAMC;AAED;;;;;;;;;EAiCC;AAED;;;;;;;;;EAcC;AAED;;;;;;;;;EAcC;AAED;;;;;;;GAOG;AACH,oEANW,MAAM,SACN,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,WACtB,OAAO,iBAAiB,EAAE,sBAAsB,GAC9C,OAAO,CAAC,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC,CAYxC;AAnsDD,oCAAsC;;;;;;;;;;;;;;;;;;;;AA+StC,+CAA4C;AAC5C,wDAAqD;AACrD,wDAAqD;;KAtTb,GAAG;UAClC,GAAE;;AAuTX;;GAEG;AACH;IA8CI;;;;;;;;;;;;;;OAcG;IACH,sDATW,MAAM,gKAKN,OAAO,iBAAiB,EAAE,sBAAsB,GAE9C,OAAO,CAAC,eAAe,CAAC,CA0DpC;IAhHD;;;;;OAKG;IACH,oBAJW,OAAO,eAAe,EAAE,gBAAgB,YACxC,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,WACnB,MAAM,CAAC,MAAM,MAAS,EAsBhC;IA9BD,wBAA8B;IAC9B,yBAAiD;IAEjD,mCAAqC;IAUjC,iDAAoB;IACpB,8BAAwB;IACxB,6BAAsB;IAKtB,kBAA2C;IAC3C,cAAkC;IAClC,oCAA+D;IAM/D,2DAA2D;IAC3D,eADW,OAAO,eAAe,EAAE,oBAAoB,CACS;IAGpE;;;;OAIG;IACH,WAHa,OAAO,CAAC,OAAO,EAAE,CAAC,CAS9B;IA2ED;;;;OAIG;IACH,0BAFa,OAAO,KAAQ,CAI3B;IAED;;;;;;OAMG;IACH,4BAHa,OAAO,KAAQ,CAK3B;IAED;;;OAGG;IACH,yBAFa,gBAAgB,GAAC,IAAI,CAIjC;IAED;;;;;OAKG;IACH,8BA2JC;IAED;;;;;;OAMG;IACH,8CAJW,gBAAgB,GAAC,IAAI,+CAEnB,gBAAgB,CA+B5B;IAED;;;;OAIG;IACH,0CAHW,gBAAgB,sBAChB,OAAO,oCAAoC,EAAE,gBAAgB,GAAC,OAAO,oCAAoC,EAAE,gBAAgB,EAAE,GAAC,oBAAoB,wBAqB5J;IAED;;;OAGG;IACH,8BA2BC;IAED,mDAKC;IAED;;;;;;;;OAQG;IACH,wGANG;QAA2B,mBAAmB,EAAtC,MAAM,EAAE,EAAE;QACK,OAAO;QACP,YAAY;QACX,kBAAkB,EAAlC,OAAO;KACf,OA0BF;IAED;;;;;;;OAOG;IACH,8DALG;QAAwB,MAAM,GAAtB,MAAM;QACU,YAAY,GAA5B,MAAM;QACmC,YAAY,GAArD,MAAM,CAAC,MAAM,EAAE,MAAM,GAAC,MAAM,EAAE,CAAC;KACvC,GAAU;QAAC,aAAa,EAAE,MAAM,CAAC;QAAC,YAAY,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,GAAG;YAAC,eAAe,CAAC,EAAE,YAAY,CAAA;SAAC,CAAC;QAAC,gBAAgB,EAAE,MAAM,CAAA;KAAC,CAmBtI;IAED;;;;;qBAmDC;IAED;;;OAGG;IACH,oJAFW,GAAC;;;MAwDX;IAED;;;;OAIG;IACH,kGAHW,OAAO,6BAA6B,EAAE,4BAA4B,GAChE,OAAO,CAAC,WAAW,GAAC,MAAM,CAAC,CAgOvC;IAED;;;;;;;OAOG;IACH,qDAJW,YAAY,uBACZ,OAAO,GACL,YAAY,CAsCxB;IAED;;;;;OAKG;IACH,kCAFa;QAAC,gBAAgB,CAAC,EAAE,MAAM,EAAE,CAAA;KAAC,CAgBzC;IAED;;;;;OAKG;IACH,+BAHW,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,iBACnB,YAAY,GAAC,IAAI,QAiB3B;IAED;;;;;;OAMG;IACH,sBAOC;IAED,wCAEC;IAED,uCAEC;IAED,wCAEC;CACJ;uBAluCmE,oBAAoB;gCAexD,uBAAuB;6BAoB1B,mBAAmB;iCANf,sCAAsC;qCACG,oCAAoC;4BAGlF,uBAAuB"}

package/types/models/models.d.ts

@@ -1,7 +1,8 @@
 export * from "./albert/modeling_albert.js";
 export * from "./apertus/modeling_apertus.js";
+export * from "./afmoe/modeling_afmoe.js";
 export * from "./arcee/modeling_arcee.js";
-export * from "./ast/modeling_ast.js";
+export * from "./audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.js";
 export * from "./bart/modeling_bart.js";
 export * from "./beit/modeling_beit.js";
 export * from "./bert/modeling_bert.js";
@@ -11,17 +12,21 @@ export * from "./bloom/modeling_bloom.js";
 export * from "./camembert/modeling_camembert.js";
 export * from "./chatterbox/modeling_chatterbox.js";
 export * from "./chinese_clip/modeling_chinese_clip.js";
+export * from "./chmv2/modeling_chmv2.js";
 export * from "./clap/modeling_clap.js";
 export * from "./clip/modeling_clip.js";
 export * from "./clipseg/modeling_clipseg.js";
 export * from "./codegen/modeling_codegen.js";
 export * from "./cohere/modeling_cohere.js";
+export * from "./cohere2/modeling_cohere2.js";
+export * from "./cohere_asr/modeling_cohere_asr.js";
 export * from "./convbert/modeling_convbert.js";
 export * from "./convnext/modeling_convnext.js";
 export * from "./convnextv2/modeling_convnextv2.js";
 export * from "./d_fine/modeling_d_fine.js";
 export * from "./dac/modeling_dac.js";
 export * from "./deberta/modeling_deberta.js";
+export * from "./deepseek_v3/modeling_deepseek_v3.js";
 export * from "./deberta_v2/modeling_deberta_v2.js";
 export * from "./decision_transformer/modeling_decision_transformer.js";
 export * from "./deit/modeling_deit.js";
@@ -39,6 +44,7 @@ export * from "./efficientnet/modeling_efficientnet.js";
 export * from "./electra/modeling_electra.js";
 export * from "./ernie4_5/modeling_ernie4_5.js";
 export * from "./esm/modeling_esm.js";
+export * from "./eurobert/modeling_eurobert.js";
 export * from "./exaone/modeling_exaone.js";
 export * from "./falcon/modeling_falcon.js";
 export * from "./falcon_h1/modeling_falcon_h1.js";
@@ -49,6 +55,8 @@ export * from "./gemma2/modeling_gemma2.js";
 export * from "./gemma3/modeling_gemma3.js";
 export * from "./gemma3n/modeling_gemma3n.js";
 export * from "./glm/modeling_glm.js";
+export * from "./glm_moe_dsa/modeling_glm_moe_dsa.js";
+export * from "./glm_ocr/modeling_glm_ocr.js";
 export * from "./glpn/modeling_glpn.js";
 export * from "./gpt_bigcode/modeling_gpt_bigcode.js";
 export * from "./gpt_neo/modeling_gpt_neo.js";
@@ -58,6 +66,7 @@ export * from "./gpt2/modeling_gpt2.js";
 export * from "./gptj/modeling_gptj.js";
 export * from "./granite/modeling_granite.js";
 export * from "./granitemoehybrid/modeling_granitemoehybrid.js";
+export * from "./granite_speech/modeling_granite_speech.js";
 export * from "./grounding_dino/modeling_grounding_dino.js";
 export * from "./groupvit/modeling_groupvit.js";
 export * from "./helium/modeling_helium.js";
@@ -69,7 +78,9 @@ export * from "./ijepa/modeling_ijepa.js";
 export * from "./jais/modeling_jais.js";
 export * from "./jina_clip/modeling_jina_clip.js";
 export * from "./lfm2/modeling_lfm2.js";
+export * from "./lighton_ocr/modeling_lighton_ocr.js";
 export * from "./lfm2_moe/modeling_lfm2_moe.js";
+export * from "./lfm2_vl/modeling_lfm2_vl.js";
 export * from "./llama/modeling_llama.js";
 export * from "./llama4/modeling_llama4.js";
 export * from "./llava/modeling_llava.js";
@@ -84,6 +95,7 @@ export * from "./metric3dv2/modeling_metric3dv2.js";
 export * from "./mgp_str/modeling_mgp_str.js";
 export * from "./mimi/modeling_mimi.js";
 export * from "./mistral/modeling_mistral.js";
+export * from "./mistral4/modeling_mistral4.js";
 export * from "./mobilebert/modeling_mobilebert.js";
 export * from "./mobilellm/modeling_mobilellm.js";
 export * from "./mobilenet_v1/modeling_mobilenet_v1.js";
@@ -101,11 +113,13 @@ export * from "./mt5/modeling_mt5.js";
 export * from "./multi_modality/modeling_multi_modality.js";
 export * from "./musicgen/modeling_musicgen.js";
 export * from "./nanochat/modeling_nanochat.js";
+export * from "./nemotron_h/modeling_nemotron_h.js";
 export * from "./neobert/modeling_neobert.js";
 export * from "./nomic_bert/modeling_nomic_bert.js";
 export * from "./olmo/modeling_olmo.js";
 export * from "./olmo2/modeling_olmo2.js";
 export * from "./olmo3/modeling_olmo3.js";
+export * from "./olmo_hybrid/modeling_olmo_hybrid.js";
 export * from "./openelm/modeling_openelm.js";
 export * from "./opt/modeling_opt.js";
 export * from "./owlv2/modeling_owlv2.js";
@@ -120,8 +134,16 @@ export * from "./phi3_v/modeling_phi3_v.js";
 export * from "./pvt/modeling_pvt.js";
 export * from "./pyannote/modeling_pyannote.js";
 export * from "./qwen2/modeling_qwen2.js";
+export * from "./qwen2_moe/modeling_qwen2_moe.js";
 export * from "./qwen2_vl/modeling_qwen2_vl.js";
+export * from "./qwen2_5_vl/modeling_qwen2_5_vl.js";
 export * from "./qwen3/modeling_qwen3.js";
+export * from "./qwen3_moe/modeling_qwen3_moe.js";
+export * from "./qwen3_next/modeling_qwen3_next.js";
+export * from "./qwen3_vl/modeling_qwen3_vl.js";
+export * from "./qwen3_vl_moe/modeling_qwen3_vl_moe.js";
+export * from "./qwen3_5/modeling_qwen3_5.js";
+export * from "./qwen3_5_moe/modeling_qwen3_5_moe.js";
 export * from "./resnet/modeling_resnet.js";
 export * from "./rf_detr/modeling_rf_detr.js";
 export * from "./roberta/modeling_roberta.js";
@@ -135,6 +157,7 @@ export * from "./segformer/modeling_segformer.js";
 export * from "./siglip/modeling_siglip.js";
 export * from "./smollm3/modeling_smollm3.js";
 export * from "./snac/modeling_snac.js";
+export * from "./solar_open/modeling_solar_open.js";
 export * from "./speecht5/modeling_speecht5.js";
 export * from "./squeezebert/modeling_squeezebert.js";
 export * from "./stablelm/modeling_stablelm.js";
@@ -157,6 +180,8 @@ export * from "./vit_msn/modeling_vit_msn.js";
 export * from "./vitmatte/modeling_vitmatte.js";
 export * from "./vitpose/modeling_vitpose.js";
 export * from "./vits/modeling_vits.js";
+export * from "./voxtral/modeling_voxtral.js";
+export * from "./voxtral_realtime/modeling_voxtral_realtime.js";
 export * from "./wav2vec2/modeling_wav2vec2.js";
 export * from "./wav2vec2_bert/modeling_wav2vec2_bert.js";
 export * from "./wavlm/modeling_wavlm.js";

package/types/models/nemotron_h/modeling_nemotron_h.d.ts

@@ -0,0 +1,8 @@
+export class NemotronHPreTrainedModel extends PreTrainedModel {
+}
+export class NemotronHModel extends NemotronHPreTrainedModel {
+}
+export class NemotronHForCausalLM extends NemotronHPreTrainedModel {
+}
+import { PreTrainedModel } from '../modeling_utils.js';
+//# sourceMappingURL=modeling_nemotron_h.d.ts.map

package/types/models/nemotron_h/modeling_nemotron_h.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"modeling_nemotron_h.d.ts","sourceRoot":"","sources":["../../../src/models/nemotron_h/modeling_nemotron_h.js"],"names":[],"mappings":"AAEA;CAAgE;AAChE;CAA+D;AAC/D;CAAqE;gCAJrC,sBAAsB"}

package/types/models/olmo_hybrid/modeling_olmo_hybrid.d.ts

@@ -0,0 +1,8 @@
+export class OlmoHybridPreTrainedModel extends PreTrainedModel {
+}
+export class OlmoHybridModel extends OlmoHybridPreTrainedModel {
+}
+export class OlmoHybridForCausalLM extends OlmoHybridPreTrainedModel {
+}
+import { PreTrainedModel } from '../modeling_utils.js';
+//# sourceMappingURL=modeling_olmo_hybrid.d.ts.map

package/types/models/olmo_hybrid/modeling_olmo_hybrid.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"modeling_olmo_hybrid.d.ts","sourceRoot":"","sources":["../../../src/models/olmo_hybrid/modeling_olmo_hybrid.js"],"names":[],"mappings":"AAEA;CAAiE;AACjE;CAAiE;AACjE;CAAuE;gCAJvC,sBAAsB"}

package/types/models/paligemma/modeling_paligemma.d.ts

@@ -1,10 +1,4 @@
-export class PaliGemmaPreTrainedModel extends PreTrainedModel {
+export class PaliGemmaForConditionalGeneration extends LlavaForConditionalGeneration {
 }
-export class PaliGemmaForConditionalGeneration extends PaliGemmaPreTrainedModel {
-    _merge_input_ids_with_image_features(kwargs: any): {
-        inputs_embeds: any;
-        attention_mask: any;
-    };
-}
-import { PreTrainedModel } from '../modeling_utils.js';
+import { LlavaForConditionalGeneration } from '../llava/modeling_llava.js';
 //# sourceMappingURL=modeling_paligemma.d.ts.map