♻️ Don't convert to camelCase when generating inference types

Author: SBrandeis

packages/tasks/scripts/inference-codegen.ts

@@ -57,7 +57,7 @@ async function generateTypescript(inputData: InputData): Promise<SerializedRende
 		indentation: "\t",
 		rendererOptions: {
 			"just-types": true,
-			"nice-property-names": true,
+			"nice-property-names": false,
 			"prefer-unions": true,
 			"prefer-const-values": true,
 			"prefer-unknown": true,

packages/tasks/src/tasks/audio-classification/inference.ts

@@ -23,11 +23,11 @@ export interface AudioClassificationInput {
  * Additional inference parameters for Audio Classification
  */
 export interface AudioClassificationParameters {
-	functionToApply?: ClassificationOutputTransform;
+	function_to_apply?: ClassificationOutputTransform;
 	/**
 	 * When specified, limits the output to the top K most probable classes.
 	 */
-	topK?: number;
+	top_k?: number;
 	[property: string]: unknown;
 }
 /**

packages/tasks/src/tasks/automatic-speech-recognition/inference.ts

@@ -32,7 +32,7 @@ export interface AutomaticSpeechRecognitionParameters {
 	/**
 	 * Whether to output corresponding timestamps with the generated text
 	 */
-	returnTimestamps?: boolean;
+	return_timestamps?: boolean;
 	[property: string]: unknown;
 }
 
@@ -45,18 +45,18 @@ export interface GenerationParameters {
 	/**
 	 * Whether to use sampling instead of greedy decoding when generating new tokens.
 	 */
-	doSample?: boolean;
+	do_sample?: boolean;
 	/**
 	 * Controls the stopping condition for beam-based methods.
 	 */
-	earlyStopping?: EarlyStoppingUnion;
+	early_stopping?: EarlyStoppingUnion;
 	/**
 	 * If set to float strictly between 0 and 1, only tokens with a conditional probability
 	 * greater than epsilon_cutoff will be sampled. In the paper, suggested values range from
 	 * 3e-4 to 9e-4, depending on the size of the model. See [Truncation Sampling as Language
 	 * Model Desmoothing](https://hf.co/papers/2210.15191) for more details.
 	 */
-	epsilonCutoff?: number;
+	epsilon_cutoff?: number;
 	/**
 	 * Eta sampling is a hybrid of locally typical sampling and epsilon sampling. If set to
 	 * float strictly between 0 and 1, a token is only considered if it is greater than either
@@ -66,62 +66,62 @@ export interface GenerationParameters {
 	 * See [Truncation Sampling as Language Model Desmoothing](https://hf.co/papers/2210.15191)
 	 * for more details.
 	 */
-	etaCutoff?: number;
+	eta_cutoff?: number;
 	/**
 	 * The maximum length (in tokens) of the generated text, including the input.
 	 */
-	maxLength?: number;
+	max_length?: number;
 	/**
 	 * The maximum number of tokens to generate. Takes precedence over maxLength.
 	 */
-	maxNewTokens?: number;
+	max_new_tokens?: number;
 	/**
 	 * The minimum length (in tokens) of the generated text, including the input.
 	 */
-	minLength?: number;
+	min_length?: number;
 	/**
 	 * The minimum number of tokens to generate. Takes precedence over maxLength.
 	 */
-	minNewTokens?: number;
+	min_new_tokens?: number;
 	/**
 	 * Number of groups to divide num_beams into in order to ensure diversity among different
 	 * groups of beams. See [this paper](https://hf.co/papers/1610.02424) for more details.
 	 */
-	numBeamGroups?: number;
+	num_beam_groups?: number;
 	/**
 	 * Number of beams to use for beam search.
 	 */
-	numBeams?: number;
+	num_beams?: number;
 	/**
 	 * The value balances the model confidence and the degeneration penalty in contrastive
 	 * search decoding.
 	 */
-	penaltyAlpha?: number;
+	penalty_alpha?: number;
 	/**
 	 * The value used to modulate the next token probabilities.
 	 */
 	temperature?: number;
 	/**
 	 * The number of highest probability vocabulary tokens to keep for top-k-filtering.
 	 */
-	topK?: number;
+	top_k?: number;
 	/**
 	 * If set to float < 1, only the smallest set of most probable tokens with probabilities
 	 * that add up to top_p or higher are kept for generation.
 	 */
-	topP?: number;
+	top_p?: number;
 	/**
 	 * Local typicality measures how similar the conditional probability of predicting a target
 	 * token next is to the expected conditional probability of predicting a random token next,
 	 * given the partial text already generated. If set to float < 1, the smallest set of the
 	 * most locally typical tokens with probabilities that add up to typical_p or higher are
 	 * kept for generation. See [this paper](https://hf.co/papers/2202.00666) for more details.
 	 */
-	typicalP?: number;
+	typical_p?: number;
 	/**
 	 * Whether the model should use the past last key/values attentions to speed up decoding
 	 */
-	useCache?: boolean;
+	use_cache?: boolean;
 	[property: string]: unknown;
 }
 

packages/tasks/src/tasks/depth-estimation/inference.ts

@@ -30,6 +30,6 @@ export interface DepthEstimationParameters {
 	/**
 	 * When specified, limits the output to the top K most probable classes.
 	 */
-	topK?: number;
+	top_k?: number;
 	[property: string]: unknown;
 }

packages/tasks/src/tasks/document-question-answering/inference.ts

@@ -42,11 +42,11 @@ export interface DocumentQuestionAnsweringParameters {
 	 * be split in several chunks with some overlap. This argument controls the size of that
 	 * overlap.
 	 */
-	docStride?: number;
+	doc_stride?: number;
 	/**
 	 * Whether to accept impossible as an answer
 	 */
-	handleImpossibleAnswer?: boolean;
+	handle_impossible_answer?: boolean;
 	/**
 	 * Language to use while running OCR. Defaults to english.
 	 */
@@ -55,27 +55,27 @@ export interface DocumentQuestionAnsweringParameters {
 	 * The maximum length of predicted answers (e.g., only answers with a shorter length are
 	 * considered).
 	 */
-	maxAnswerLen?: number;
+	max_answer_len?: number;
 	/**
 	 * The maximum length of the question after tokenization. It will be truncated if needed.
 	 */
-	maxQuestionLen?: number;
+	max_question_len?: number;
 	/**
 	 * The maximum length of the total sentence (context + question) in tokens of each chunk
 	 * passed to the model. The context will be split in several chunks (using doc_stride as
 	 * overlap) if needed.
 	 */
-	maxSeqLen?: number;
+	max_seq_len?: number;
 	/**
 	 * The number of answers to return (will be chosen by order of likelihood). Can return less
 	 * than top_k answers if there are not enough options available within the context.
 	 */
-	topK?: number;
+	top_k?: number;
 	/**
 	 * A list of words and bounding boxes (normalized 0->1000). If provided, the inference will
 	 * skip the OCR step and use the provided bounding boxes instead.
 	 */
-	wordBoxes?: WordBox[];
+	word_boxes?: WordBox[];
 	[property: string]: unknown;
 }
 export type WordBox = number[] | string;

packages/tasks/src/tasks/fill-mask/inference.ts

@@ -33,18 +33,14 @@ export interface FillMaskParameters {
 	/**
 	 * When passed, overrides the number of predictions to return.
 	 */
-	topK?: number;
+	top_k?: number;
 	[property: string]: unknown;
 }
 export type FillMaskOutput = FillMaskOutputElement[];
 /**
  * Outputs of inference for the Fill Mask task
  */
 export interface FillMaskOutputElement {
-	/**
-	 * The predicted token (to replace the masked one).
-	 */
-	fillMaskOutputTokenStr?: string;
 	/**
 	 * The corresponding probability
 	 */
@@ -58,5 +54,9 @@ export interface FillMaskOutputElement {
 	 */
 	token: number;
 	tokenStr: unknown;
+	/**
+	 * The predicted token (to replace the masked one).
+	 */
+	token_str?: string;
 	[property: string]: unknown;
 }

packages/tasks/src/tasks/image-classification/inference.ts

@@ -23,11 +23,11 @@ export interface ImageClassificationInput {
  * Additional inference parameters for Image Classification
  */
 export interface ImageClassificationParameters {
-	functionToApply?: ClassificationOutputTransform;
+	function_to_apply?: ClassificationOutputTransform;
 	/**
 	 * When specified, limits the output to the top K most probable classes.
 	 */
-	topK?: number;
+	top_k?: number;
 	[property: string]: unknown;
 }
 /**

packages/tasks/src/tasks/image-segmentation/inference.ts

@@ -26,11 +26,11 @@ export interface ImageSegmentationParameters {
 	/**
 	 * Threshold to use when turning the predicted masks into binary values.
 	 */
-	maskThreshold?: number;
+	mask_threshold?: number;
 	/**
 	 * Mask overlap threshold to eliminate small, disconnected segments.
 	 */
-	overlapMaskAreaThreshold?: number;
+	overlap_mask_area_threshold?: number;
 	/**
 	 * Segmentation task to be performed, depending on model capabilities.
 	 */

packages/tasks/src/tasks/image-to-image/inference.ts

@@ -29,20 +29,20 @@ export interface ImageToImageParameters {
 	 * For diffusion models. A higher guidance scale value encourages the model to generate
 	 * images closely linked to the text prompt at the expense of lower image quality.
 	 */
-	guidanceScale?: number;
+	guidance_scale?: number;
 	/**
 	 * One or several prompt to guide what NOT to include in image generation.
 	 */
-	negativePrompt?: string[];
+	negative_prompt?: string[];
 	/**
 	 * For diffusion models. The number of denoising steps. More denoising steps usually lead to
 	 * a higher quality image at the expense of slower inference.
 	 */
-	numInferenceSteps?: number;
+	num_inference_steps?: number;
 	/**
 	 * The size in pixel of the output image
 	 */
-	targetSize?: TargetSize;
+	target_size?: TargetSize;
 	[property: string]: unknown;
 }
 

packages/tasks/src/tasks/image-to-text/inference.ts

@@ -32,7 +32,7 @@ export interface ImageToTextParameters {
 	/**
 	 * The amount of maximum tokens to generate.
 	 */
-	maxNewTokens?: number;
+	max_new_tokens?: number;
 	[property: string]: unknown;
 }
 
@@ -45,18 +45,18 @@ export interface GenerationParameters {
 	/**
 	 * Whether to use sampling instead of greedy decoding when generating new tokens.
 	 */
-	doSample?: boolean;
+	do_sample?: boolean;
 	/**
 	 * Controls the stopping condition for beam-based methods.
 	 */
-	earlyStopping?: EarlyStoppingUnion;
+	early_stopping?: EarlyStoppingUnion;
 	/**
 	 * If set to float strictly between 0 and 1, only tokens with a conditional probability
 	 * greater than epsilon_cutoff will be sampled. In the paper, suggested values range from
 	 * 3e-4 to 9e-4, depending on the size of the model. See [Truncation Sampling as Language
 	 * Model Desmoothing](https://hf.co/papers/2210.15191) for more details.
 	 */
-	epsilonCutoff?: number;
+	epsilon_cutoff?: number;
 	/**
 	 * Eta sampling is a hybrid of locally typical sampling and epsilon sampling. If set to
 	 * float strictly between 0 and 1, a token is only considered if it is greater than either
@@ -66,62 +66,62 @@ export interface GenerationParameters {
 	 * See [Truncation Sampling as Language Model Desmoothing](https://hf.co/papers/2210.15191)
 	 * for more details.
 	 */
-	etaCutoff?: number;
+	eta_cutoff?: number;
 	/**
 	 * The maximum length (in tokens) of the generated text, including the input.
 	 */
-	maxLength?: number;
+	max_length?: number;
 	/**
 	 * The maximum number of tokens to generate. Takes precedence over maxLength.
 	 */
-	maxNewTokens?: number;
+	max_new_tokens?: number;
 	/**
 	 * The minimum length (in tokens) of the generated text, including the input.
 	 */
-	minLength?: number;
+	min_length?: number;
 	/**
 	 * The minimum number of tokens to generate. Takes precedence over maxLength.
 	 */
-	minNewTokens?: number;
+	min_new_tokens?: number;
 	/**
 	 * Number of groups to divide num_beams into in order to ensure diversity among different
 	 * groups of beams. See [this paper](https://hf.co/papers/1610.02424) for more details.
 	 */
-	numBeamGroups?: number;
+	num_beam_groups?: number;
 	/**
 	 * Number of beams to use for beam search.
 	 */
-	numBeams?: number;
+	num_beams?: number;
 	/**
 	 * The value balances the model confidence and the degeneration penalty in contrastive
 	 * search decoding.
 	 */
-	penaltyAlpha?: number;
+	penalty_alpha?: number;
 	/**
 	 * The value used to modulate the next token probabilities.
 	 */
 	temperature?: number;
 	/**
 	 * The number of highest probability vocabulary tokens to keep for top-k-filtering.
 	 */
-	topK?: number;
+	top_k?: number;
 	/**
 	 * If set to float < 1, only the smallest set of most probable tokens with probabilities
 	 * that add up to top_p or higher are kept for generation.
 	 */
-	topP?: number;
+	top_p?: number;
 	/**
 	 * Local typicality measures how similar the conditional probability of predicting a target
 	 * token next is to the expected conditional probability of predicting a random token next,
 	 * given the partial text already generated. If set to float < 1, the smallest set of the
 	 * most locally typical tokens with probabilities that add up to typical_p or higher are
 	 * kept for generation. See [this paper](https://hf.co/papers/2202.00666) for more details.
 	 */
-	typicalP?: number;
+	typical_p?: number;
 	/**
 	 * Whether the model should use the past last key/values attentions to speed up decoding
 	 */
-	useCache?: boolean;
+	use_cache?: boolean;
 	[property: string]: unknown;
 }
 
@@ -138,6 +138,6 @@ export interface ImageToTextOutput {
 	/**
 	 * The generated text.
 	 */
-	imageToTextOutputGeneratedText?: string;
+	generated_text?: string;
 	[property: string]: unknown;
 }