@huggingface/tasks 0.13.4 → 0.13.6

package/dist/esm/tasks/zero-shot-image-classification/inference.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"inference.d.ts","sourceRoot":"","sources":["../../../../src/tasks/zero-shot-image-classification/inference.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AACH;;GAEG;AACH,MAAM,WAAW,gCAAgC;IAChD;;OAEG;IACH,MAAM,EAAE,MAAM,CAAC;IACf;;OAEG;IACH,UAAU,EAAE,qCAAqC,CAAC;IAClD,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC;CAC5B;AACD;;;;GAIG;AACH,MAAM,WAAW,qCAAqC;IACrD;;OAEG;IACH,gBAAgB,EAAE,MAAM,EAAE,CAAC;IAC3B;;;OAGG;IACH,mBAAmB,CAAC,EAAE,MAAM,CAAC;IAC7B,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC;CAC5B;AACD,MAAM,MAAM,iCAAiC,GAAG,wCAAwC,EAAE,CAAC;AAC3F;;GAEG;AACH,MAAM,WAAW,wCAAwC;IACxD;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC;CAC5B"}
+{"version":3,"file":"inference.d.ts","sourceRoot":"","sources":["../../../../src/tasks/zero-shot-image-classification/inference.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AACH;;GAEG;AACH,MAAM,WAAW,gCAAgC;IAChD;;OAEG;IACH,MAAM,EAAE,MAAM,CAAC;IACf;;OAEG;IACH,UAAU,EAAE,qCAAqC,CAAC;IAClD,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC;CAC5B;AACD;;GAEG;AACH,MAAM,WAAW,qCAAqC;IACrD;;OAEG;IACH,gBAAgB,EAAE,MAAM,EAAE,CAAC;IAC3B;;;OAGG;IACH,mBAAmB,CAAC,EAAE,MAAM,CAAC;IAC7B,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC;CAC5B;AACD,MAAM,MAAM,iCAAiC,GAAG,wCAAwC,EAAE,CAAC;AAC3F;;GAEG;AACH,MAAM,WAAW,wCAAwC;IACxD;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC;CAC5B"}

package/dist/esm/tasks/zero-shot-object-detection/inference.d.ts

@@ -12,14 +12,12 @@ export interface ZeroShotObjectDetectionInput {
      */
     inputs: string;
     /**
-     * Additional inference parameters
+     * Additional inference parameters for Zero Shot Object Detection
      */
     parameters: ZeroShotObjectDetectionParameters;
     [property: string]: unknown;
 }
 /**
- * Additional inference parameters
- *
  * Additional inference parameters for Zero Shot Object Detection
  */
 export interface ZeroShotObjectDetectionParameters {

package/dist/esm/tasks/zero-shot-object-detection/inference.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"inference.d.ts","sourceRoot":"","sources":["../../../../src/tasks/zero-shot-object-detection/inference.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AACH;;GAEG;AACH,MAAM,WAAW,4BAA4B;IAC5C;;OAEG;IACH,MAAM,EAAE,MAAM,CAAC;IACf;;OAEG;IACH,UAAU,EAAE,iCAAiC,CAAC;IAC9C,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC;CAC5B;AACD;;;;GAIG;AACH,MAAM,WAAW,iCAAiC;IACjD;;OAEG;IACH,gBAAgB,EAAE,MAAM,EAAE,CAAC;IAC3B,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC;CAC5B;AACD;;;GAGG;AACH,MAAM,WAAW,WAAW;IAC3B,IAAI,EAAE,MAAM,CAAC;IACb,IAAI,EAAE,MAAM,CAAC;IACb,IAAI,EAAE,MAAM,CAAC;IACb,IAAI,EAAE,MAAM,CAAC;IACb,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC;CAC5B;AACD,MAAM,MAAM,6BAA6B,GAAG,oCAAoC,EAAE,CAAC;AACnF;;GAEG;AACH,MAAM,WAAW,oCAAoC;IACpD;;;OAGG;IACH,GAAG,EAAE,WAAW,CAAC;IACjB;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC;CAC5B"}
+{"version":3,"file":"inference.d.ts","sourceRoot":"","sources":["../../../../src/tasks/zero-shot-object-detection/inference.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AACH;;GAEG;AACH,MAAM,WAAW,4BAA4B;IAC5C;;OAEG;IACH,MAAM,EAAE,MAAM,CAAC;IACf;;OAEG;IACH,UAAU,EAAE,iCAAiC,CAAC;IAC9C,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC;CAC5B;AACD;;GAEG;AACH,MAAM,WAAW,iCAAiC;IACjD;;OAEG;IACH,gBAAgB,EAAE,MAAM,EAAE,CAAC;IAC3B,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC;CAC5B;AACD;;;GAGG;AACH,MAAM,WAAW,WAAW;IAC3B,IAAI,EAAE,MAAM,CAAC;IACb,IAAI,EAAE,MAAM,CAAC;IACb,IAAI,EAAE,MAAM,CAAC;IACb,IAAI,EAAE,MAAM,CAAC;IACb,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC;CAC5B;AACD,MAAM,MAAM,6BAA6B,GAAG,oCAAoC,EAAE,CAAC;AACnF;;GAEG;AACH,MAAM,WAAW,oCAAoC;IACpD;;;OAGG;IACH,GAAG,EAAE,WAAW,CAAC;IACjB;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC;CAC5B"}

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@huggingface/tasks",
   "packageManager": "pnpm@8.10.5",
-  "version": "0.13.4",
+  "version": "0.13.6",
   "description": "List of ML tasks for huggingface.co/tasks",
   "repository": "https://github.com/huggingface/huggingface.js.git",
   "publishConfig": {

package/src/hardware.ts

@@ -452,7 +452,7 @@ export const SKUS = {
 				memory: [8, 16, 24],
 			},
 			"Apple M2 Pro": {
-				tflops: 13.6,
+				tflops: 6.8,
 				memory: [16, 24, 32],
 			},
 			"Apple M2 Max": {
@@ -464,17 +464,29 @@ export const SKUS = {
 				memory: [64, 96, 128, 192],
 			},
 			"Apple M3": {
-				tflops: 2.84,
+				tflops: 4.1,
 				memory: [8, 16, 24],
 			},
 			"Apple M3 Pro": {
-				tflops: 14,
+				tflops: 7.4,
 				memory: [18, 36],
 			},
 			"Apple M3 Max": {
 				tflops: 14.2,
 				memory: [36, 48, 64, 96, 128],
 			},
+			"Apple M4": {
+				tflops: 4.6,
+				memory: [16, 24, 32],
+			},
+			"Apple M4 Pro": {
+				tflops: 9.2,
+				memory: [24, 48],
+			},
+			"Apple M4 Max": {
+				tflops: 18.4,
+				memory: [36, 48, 64, 128],
+			},
 		},
 	},
 } satisfies Record<string, Record<string, Record<string, HardwareSpec>>>;

package/src/local-apps.spec.ts

@@ -0,0 +1,123 @@
+import { describe, expect, it } from "vitest";
+import { LOCAL_APPS } from "./local-apps.js";
+import type { ModelData } from "./model-data.js";
+
+describe("local-apps", () => {
+	it("llama.cpp conversational", async () => {
+		const { snippet: snippetFunc } = LOCAL_APPS["llama.cpp"];
+		const model: ModelData = {
+			id: "bartowski/Llama-3.2-3B-Instruct-GGUF",
+			tags: ["conversational"],
+			inference: "",
+		};
+		const snippet = snippetFunc(model);
+
+		expect(snippet[0].content).toEqual(`# Load and run the model:
+llama-cli \\
+  --hf-repo "bartowski/Llama-3.2-3B-Instruct-GGUF" \\
+  --hf-file {{GGUF_FILE}} \\
+  -p "You are a helpful assistant" \\
+  --conversation`);
+	});
+
+	it("llama.cpp non-conversational", async () => {
+		const { snippet: snippetFunc } = LOCAL_APPS["llama.cpp"];
+		const model: ModelData = {
+			id: "mlabonne/gemma-2b-GGUF",
+			tags: [],
+			inference: "",
+		};
+		const snippet = snippetFunc(model);
+
+		expect(snippet[0].content).toEqual(`# Load and run the model:
+llama-cli \\
+  --hf-repo "mlabonne/gemma-2b-GGUF" \\
+  --hf-file {{GGUF_FILE}} \\
+  -p "Once upon a time,"`);
+	});
+
+	it("vLLM conversational llm", async () => {
+		const { snippet: snippetFunc } = LOCAL_APPS["vllm"];
+		const model: ModelData = {
+			id: "meta-llama/Llama-3.2-3B-Instruct",
+			pipeline_tag: "text-generation",
+			tags: ["conversational"],
+			inference: "",
+		};
+		const snippet = snippetFunc(model);
+
+		expect((snippet[0].content as string[]).join("\n")).toEqual(`# Load and run the model:
+vllm serve "meta-llama/Llama-3.2-3B-Instruct"
+# Call the server using curl:
+curl -X POST "http://localhost:8000/v1/chat/completions" \\
+	-H "Content-Type: application/json" \\
+	--data '{
+		"model": "meta-llama/Llama-3.2-3B-Instruct",
+		"messages": [
+			{
+				"role": "user",
+				"content": "What is the capital of France?"
+			}
+		]
+	}'`);
+	});
+
+	it("vLLM non-conversational llm", async () => {
+		const { snippet: snippetFunc } = LOCAL_APPS["vllm"];
+		const model: ModelData = {
+			id: "meta-llama/Llama-3.2-3B",
+			tags: [""],
+			inference: "",
+		};
+		const snippet = snippetFunc(model);
+
+		expect((snippet[0].content as string[]).join("\n")).toEqual(`# Load and run the model:
+vllm serve "meta-llama/Llama-3.2-3B"
+# Call the server using curl:
+curl -X POST "http://localhost:8000/v1/completions" \\
+	-H "Content-Type: application/json" \\
+	--data '{
+		"model": "meta-llama/Llama-3.2-3B",
+		"prompt": "Once upon a time,",
+		"max_tokens": 512,
+		"temperature": 0.5
+	}'`);
+	});
+
+	it("vLLM conversational vlm", async () => {
+		const { snippet: snippetFunc } = LOCAL_APPS["vllm"];
+		const model: ModelData = {
+			id: "meta-llama/Llama-3.2-11B-Vision-Instruct",
+			pipeline_tag: "image-text-to-text",
+			tags: ["conversational"],
+			inference: "",
+		};
+		const snippet = snippetFunc(model);
+
+		expect((snippet[0].content as string[]).join("\n")).toEqual(`# Load and run the model:
+vllm serve "meta-llama/Llama-3.2-11B-Vision-Instruct"
+# Call the server using curl:
+curl -X POST "http://localhost:8000/v1/chat/completions" \\
+	-H "Content-Type: application/json" \\
+	--data '{
+		"model": "meta-llama/Llama-3.2-11B-Vision-Instruct",
+		"messages": [
+			{
+				"role": "user",
+				"content": [
+					{
+						"type": "text",
+						"text": "Describe this image in one sentence."
+					},
+					{
+						"type": "image_url",
+						"image_url": {
+							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
+						}
+					}
+				]
+			}
+		]
+	}'`);
+	});
+});

package/src/local-apps.ts

@@ -1,6 +1,9 @@
 import { parseGGUFQuantLabel } from "./gguf.js";
 import type { ModelData } from "./model-data.js";
 import type { PipelineType } from "./pipelines.js";
+import { stringifyMessages } from "./snippets/common.js";
+import { getModelInputSnippet } from "./snippets/inputs.js";
+import type { ChatCompletionInputMessage } from "./tasks/index.js";
 
 export interface LocalAppSnippet {
 	/**
@@ -92,15 +95,20 @@ function isMlxModel(model: ModelData) {
 }
 
 const snippetLlamacpp = (model: ModelData, filepath?: string): LocalAppSnippet[] => {
-	const command = (binary: string) =>
-		[
+	const command = (binary: string) => {
+		const snippet = [
 			"# Load and run the model:",
 			`${binary} \\`,
 			`  --hf-repo "${model.id}" \\`,
 			`  --hf-file ${filepath ?? "{{GGUF_FILE}}"} \\`,
-			'  -p "You are a helpful assistant" \\',
-			"  --conversation",
-		].join("\n");
+			`  -p "${model.tags.includes("conversational") ? "You are a helpful assistant" : "Once upon a time,"}"`,
+		];
+		if (model.tags.includes("conversational")) {
+			snippet[snippet.length - 1] += " \\";
+			snippet.push("  --conversation");
+		}
+		return snippet.join("\n");
+	};
 	return [
 		{
 			title: "Install from brew",
@@ -178,22 +186,33 @@ const snippetLocalAI = (model: ModelData, filepath?: string): LocalAppSnippet[]
 };
 
 const snippetVllm = (model: ModelData): LocalAppSnippet[] => {
-	const runCommand = [
-		"# Call the server using curl:",
-		`curl -X POST "http://localhost:8000/v1/chat/completions" \\`,
-		`	-H "Content-Type: application/json" \\`,
-		`	--data '{`,
-		`		"model": "${model.id}",`,
-		`		"messages": [`,
-		`			{"role": "user", "content": "Hello!"}`,
-		`		]`,
-		`	}'`,
-	];
+	const messages = getModelInputSnippet(model) as ChatCompletionInputMessage[];
+	const runCommandInstruct = `# Call the server using curl:
+curl -X POST "http://localhost:8000/v1/chat/completions" \\
+	-H "Content-Type: application/json" \\
+	--data '{
+		"model": "${model.id}",
+		"messages": ${stringifyMessages(messages, {
+			indent: "\t\t",
+			attributeKeyQuotes: true,
+			customContentEscaper: (str) => str.replace(/'/g, "'\\''"),
+		})}
+	}'`;
+	const runCommandNonInstruct = `# Call the server using curl:
+curl -X POST "http://localhost:8000/v1/completions" \\
+	-H "Content-Type: application/json" \\
+	--data '{
+		"model": "${model.id}",
+		"prompt": "Once upon a time,",
+		"max_tokens": 512,
+		"temperature": 0.5
+	}'`;
+	const runCommand = model.tags.includes("conversational") ? runCommandInstruct : runCommandNonInstruct;
 	return [
 		{
 			title: "Install from pip",
 			setup: ["# Install vLLM from pip:", "pip install vllm"].join("\n"),
-			content: [`# Load and run the model:\nvllm serve "${model.id}"`, runCommand.join("\n")],
+			content: [`# Load and run the model:\nvllm serve "${model.id}"`, runCommand],
 		},
 		{
 			title: "Use Docker images",
@@ -210,7 +229,7 @@ const snippetVllm = (model: ModelData): LocalAppSnippet[] => {
 			].join("\n"),
 			content: [
 				`# Load and run the model:\ndocker exec -it my_vllm_container bash -c "vllm serve ${model.id}"`,
-				runCommand.join("\n"),
+				runCommand,
 			],
 		},
 	];

package/src/model-libraries-snippets.spec.ts

@@ -0,0 +1,54 @@
+import { describe, expect, it } from "vitest";
+import type { ModelData } from "./model-data.js";
+import { llama_cpp_python } from "./model-libraries-snippets.js";
+
+describe("model-libraries-snippets", () => {
+	it("llama_cpp_python conversational", async () => {
+		const model: ModelData = {
+			id: "bartowski/Llama-3.2-3B-Instruct-GGUF",
+			pipeline_tag: "text-generation",
+			tags: ["conversational"],
+			inference: "",
+		};
+		const snippet = llama_cpp_python(model);
+
+		expect(snippet.join("\n")).toEqual(`from llama_cpp import Llama
+
+llm = Llama.from_pretrained(
+	repo_id="bartowski/Llama-3.2-3B-Instruct-GGUF",
+	filename="{{GGUF_FILE}}",
+)
+
+llm.create_chat_completion(
+	messages = [
+		{
+			"role": "user",
+			"content": "What is the capital of France?"
+		}
+	]
+)`);
+	});
+
+	it("llama_cpp_python non-conversational", async () => {
+		const model: ModelData = {
+			id: "mlabonne/gemma-2b-GGUF",
+			tags: [""],
+			inference: "",
+		};
+		const snippet = llama_cpp_python(model);
+
+		expect(snippet.join("\n")).toEqual(`from llama_cpp import Llama
+
+llm = Llama.from_pretrained(
+	repo_id="mlabonne/gemma-2b-GGUF",
+	filename="{{GGUF_FILE}}",
+)
+
+output = llm(
+	"Once upon a time,",
+	max_tokens=512,
+	echo=True
+)
+print(output)`);
+	});
+});

package/src/model-libraries-snippets.ts

@@ -1,6 +1,9 @@
 import type { ModelData } from "./model-data.js";
 import type { WidgetExampleTextInput, WidgetExampleSentenceSimilarityInput } from "./widget-example.js";
 import { LIBRARY_TASK_MAPPING } from "./library-to-tasks.js";
+import { getModelInputSnippet } from "./snippets/inputs.js";
+import type { ChatCompletionInputMessage } from "./tasks/index.js";
+import { stringifyMessages } from "./snippets/common.js";
 
 const TAG_CUSTOM_CODE = "custom_code";
 
@@ -418,23 +421,33 @@ model = keras_hub.models.CausalLM.from_preset("hf://${model.id}", dtype="bfloat1
 `,
 ];
 
-export const llama_cpp_python = (model: ModelData): string[] => [
-	`from llama_cpp import Llama
+export const llama_cpp_python = (model: ModelData): string[] => {
+	const snippets = [
+		`from llama_cpp import Llama
 
 llm = Llama.from_pretrained(
 	repo_id="${model.id}",
 	filename="{{GGUF_FILE}}",
 )
+`,
+	];
 
-llm.create_chat_completion(
-	messages = [
-		{
-			"role": "user",
-			"content": "What is the capital of France?"
-		}
-	]
-)`,
-];
+	if (model.tags.includes("conversational")) {
+		const messages = getModelInputSnippet(model) as ChatCompletionInputMessage[];
+		snippets.push(`llm.create_chat_completion(
+	messages = ${stringifyMessages(messages, { attributeKeyQuotes: true, indent: "\t" })}
+)`);
+	} else {
+		snippets.push(`output = llm(
+	"Once upon a time,",
+	max_tokens=512,
+	echo=True
+)
+print(output)`);
+	}
+
+	return snippets;
+};
 
 export const tf_keras = (model: ModelData): string[] => [
 	`# Note: 'keras<3.x' or 'tf_keras' must be installed (legacy)

package/src/snippets/curl.ts

@@ -9,7 +9,7 @@ export const snippetBasic = (model: ModelDataMinimal, accessToken: string): Infe
 	-X POST \\
 	-d '{"inputs": ${getModelInputSnippet(model, true)}}' \\
 	-H 'Content-Type: application/json' \\
-	-H "Authorization: Bearer ${accessToken || `{API_TOKEN}`}"`,
+	-H 'Authorization: Bearer ${accessToken || `{API_TOKEN}`}'`,
 });
 
 export const snippetTextGeneration = (
@@ -36,7 +36,7 @@ export const snippetTextGeneration = (
 		};
 		return {
 			content: `curl 'https://api-inference.huggingface.co/models/${model.id}/v1/chat/completions' \\
--H "Authorization: Bearer ${accessToken || `{API_TOKEN}`}" \\
+-H 'Authorization: Bearer ${accessToken || `{API_TOKEN}`}' \\
 -H 'Content-Type: application/json' \\
 --data '{
     "model": "${model.id}",
@@ -63,14 +63,14 @@ export const snippetZeroShotClassification = (model: ModelDataMinimal, accessTok
 	-X POST \\
 	-d '{"inputs": ${getModelInputSnippet(model, true)}, "parameters": {"candidate_labels": ["refund", "legal", "faq"]}}' \\
 	-H 'Content-Type: application/json' \\
-	-H "Authorization: Bearer ${accessToken || `{API_TOKEN}`}"`,
+	-H 'Authorization: Bearer ${accessToken || `{API_TOKEN}`}'`,
 });
 
 export const snippetFile = (model: ModelDataMinimal, accessToken: string): InferenceSnippet => ({
 	content: `curl https://api-inference.huggingface.co/models/${model.id} \\
 	-X POST \\
 	--data-binary '@${getModelInputSnippet(model, true, true)}' \\
-	-H "Authorization: Bearer ${accessToken || `{API_TOKEN}`}"`,
+	-H 'Authorization: Bearer ${accessToken || `{API_TOKEN}`}'`,
 });
 
 export const curlSnippets: Partial<

package/src/snippets/js.ts

@@ -10,7 +10,7 @@ export const snippetBasic = (model: ModelDataMinimal, accessToken: string): Infe
 		"https://api-inference.huggingface.co/models/${model.id}",
 		{
 			headers: {
-				Authorization: "Bearer ${accessToken || `{API_TOKEN}`}"
+				Authorization: "Bearer ${accessToken || `{API_TOKEN}`}",
 				"Content-Type": "application/json",
 			},
 			method: "POST",
@@ -151,7 +151,7 @@ export const snippetZeroShotClassification = (model: ModelDataMinimal, accessTok
 		"https://api-inference.huggingface.co/models/${model.id}",
 		{
 			headers: {
-				Authorization: "Bearer ${accessToken || `{API_TOKEN}`}"
+				Authorization: "Bearer ${accessToken || `{API_TOKEN}`}",
 				"Content-Type": "application/json",
 			},
 			method: "POST",
@@ -175,7 +175,7 @@ export const snippetTextToImage = (model: ModelDataMinimal, accessToken: string)
 		"https://api-inference.huggingface.co/models/${model.id}",
 		{
 			headers: {
-				Authorization: "Bearer ${accessToken || `{API_TOKEN}`}"
+				Authorization: "Bearer ${accessToken || `{API_TOKEN}`}",
 				"Content-Type": "application/json",
 			},
 			method: "POST",
@@ -196,7 +196,7 @@ export const snippetTextToAudio = (model: ModelDataMinimal, accessToken: string)
 			"https://api-inference.huggingface.co/models/${model.id}",
 			{
 				headers: {
-					Authorization: "Bearer ${accessToken || `{API_TOKEN}`}"
+					Authorization: "Bearer ${accessToken || `{API_TOKEN}`}",
 					"Content-Type": "application/json",
 				},
 				method: "POST",
@@ -238,7 +238,7 @@ export const snippetFile = (model: ModelDataMinimal, accessToken: string): Infer
 		"https://api-inference.huggingface.co/models/${model.id}",
 		{
 			headers: {
-				Authorization: "Bearer ${accessToken || `{API_TOKEN}`}"
+				Authorization: "Bearer ${accessToken || `{API_TOKEN}`}",
 				"Content-Type": "application/json",
 			},
 			method: "POST",

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

@@ -13,17 +13,18 @@ export interface AudioClassificationInput {
 	 */
 	inputs: string;
 	/**
-	 * Additional inference parameters
+	 * Additional inference parameters for Audio Classification
 	 */
 	parameters?: AudioClassificationParameters;
 	[property: string]: unknown;
 }
 /**
- * Additional inference parameters
- *
  * Additional inference parameters for Audio Classification
  */
 export interface AudioClassificationParameters {
+	/**
+	 * The function to apply to the model outputs in order to retrieve the scores.
+	 */
 	function_to_apply?: ClassificationOutputTransform;
 	/**
 	 * When specified, limits the output to the top K most probable classes.

package/src/tasks/audio-classification/spec/input.json

@@ -10,19 +10,19 @@
 			"type": "string"
 		},
 		"parameters": {
-			"description": "Additional inference parameters",
+			"description": "Additional inference parameters for Audio Classification",
 			"$ref": "#/$defs/AudioClassificationParameters"
 		}
 	},
 	"$defs": {
 		"AudioClassificationParameters": {
 			"title": "AudioClassificationParameters",
-			"description": "Additional inference parameters for Audio Classification",
 			"type": "object",
 			"properties": {
 				"function_to_apply": {
 					"title": "AudioClassificationOutputTransform",
-					"$ref": "/inference/schemas/common-definitions.json#/definitions/ClassificationOutputTransform"
+					"$ref": "/inference/schemas/common-definitions.json#/definitions/ClassificationOutputTransform",
+					"description": "The function to apply to the model outputs in order to retrieve the scores."
 				},
 				"top_k": {
 					"type": "integer",

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

@@ -14,15 +14,13 @@ export interface AutomaticSpeechRecognitionInput {
 	 */
 	inputs: string;
 	/**
-	 * Additional inference parameters
+	 * Additional inference parameters for Automatic Speech Recognition
 	 */
 	parameters?: AutomaticSpeechRecognitionParameters;
 	[property: string]: unknown;
 }
 
 /**
- * Additional inference parameters
- *
  * Additional inference parameters for Automatic Speech Recognition
  */
 export interface AutomaticSpeechRecognitionParameters {
@@ -39,8 +37,6 @@ export interface AutomaticSpeechRecognitionParameters {
 
 /**
  * Parametrization of the text generation process
- *
- * Ad-hoc parametrization of the text generation process
  */
 export interface GenerationParameters {
 	/**

package/src/tasks/automatic-speech-recognition/spec/input.json

@@ -10,14 +10,13 @@
 			"type": "string"
 		},
 		"parameters": {
-			"description": "Additional inference parameters",
+			"description": "Additional inference parameters for Automatic Speech Recognition",
 			"$ref": "#/$defs/AutomaticSpeechRecognitionParameters"
 		}
 	},
 	"$defs": {
 		"AutomaticSpeechRecognitionParameters": {
 			"title": "AutomaticSpeechRecognitionParameters",
-			"description": "Additional inference parameters for Automatic Speech Recognition",
 			"type": "object",
 			"properties": {
 				"return_timestamps": {

package/src/tasks/chat-completion/inference.ts

@@ -79,7 +79,7 @@ export interface ChatCompletionInput {
 	 * We generally recommend altering this or `top_p` but not both.
 	 */
 	temperature?: number;
-	tool_choice?: ChatCompletionInputTool;
+	tool_choice?: ChatCompletionInputToolChoice;
 	/**
 	 * A prompt to be appended before the tools
 	 */
@@ -89,7 +89,7 @@ export interface ChatCompletionInput {
 	 * Use this to provide a list of
 	 * functions the model may generate JSON inputs for.
 	 */
-	tools?: ToolElement[];
+	tools?: ChatCompletionInputTool[];
 	/**
 	 * An integer between 0 and 5 specifying the number of most likely tokens to return at each
 	 * token position, each with
@@ -154,10 +154,23 @@ export interface ChatCompletionInputStreamOptions {
 	[property: string]: unknown;
 }
 
-export type ChatCompletionInputTool = ChatCompletionInputToolType | string;
+/**
+ *
+ * <https://platform.openai.com/docs/guides/function-calling/configuring-function-calling-behavior-using-the-tool_choice-parameter>
+ */
+export type ChatCompletionInputToolChoice = ChatCompletionInputToolChoiceEnum | ChatCompletionInputToolChoiceObject;
+
+/**
+ * Means the model can pick between generating a message or calling one or more tools.
+ *
+ * Means the model will not call any tool and instead generates a message.
+ *
+ * Means the model must call one or more tools.
+ */
+export type ChatCompletionInputToolChoiceEnum = "auto" | "none" | "required";
 
-export interface ChatCompletionInputToolType {
-	function?: ChatCompletionInputFunctionName;
+export interface ChatCompletionInputToolChoiceObject {
+	function: ChatCompletionInputFunctionName;
 	[property: string]: unknown;
 }
 
@@ -166,7 +179,7 @@ export interface ChatCompletionInputFunctionName {
 	[property: string]: unknown;
 }
 
-export interface ToolElement {
+export interface ChatCompletionInputTool {
 	function: ChatCompletionInputFunctionDefinition;
 	type: string;
 	[property: string]: unknown;