improve python snippets to add huggingface_hub in more tasks (#1185)

julien-c · Wauplin · web-flow · commit cea35c9de83d · 2025-02-06T13:28:21.000+01:00
(the code for inference snippets is still quite ugly, but we'll improve
later)

---------

Co-authored-by: Wauplin &lt;lucainp@gmail.com&gt;
diff --git a/packages/tasks-gen/scripts/generate-snippets-fixtures.ts b/packages/tasks-gen/scripts/generate-snippets-fixtures.ts
@@ -19,7 +19,7 @@ import { existsSync as pathExists } from "node:fs";
 import * as fs from "node:fs/promises";
 import * as path from "node:path/posix";
 
-import type { InferenceProvider, InferenceSnippet } from "@huggingface/tasks";
+import type { SnippetInferenceProvider, InferenceSnippet } from "@huggingface/tasks";
 import { snippets } from "@huggingface/tasks";
 
 type LANGUAGE = "sh" | "js" | "py";
@@ -28,7 +28,7 @@ const TEST_CASES: {
 	testName: string;
 	model: snippets.ModelDataMinimal;
 	languages: LANGUAGE[];
-	providers: InferenceProvider[];
+	providers: SnippetInferenceProvider[];
 	opts?: Record<string, unknown>;
 }[] = [
 	{
@@ -90,6 +90,17 @@ const TEST_CASES: {
 		providers: ["hf-inference"],
 		languages: ["sh", "js", "py"],
 	},
+	{
+		testName: "text-classification",
+		model: {
+			id: "distilbert/distilbert-base-uncased-finetuned-sst-2-english",
+			pipeline_tag: "text-classification",
+			tags: [],
+			inference: "",
+		},
+		providers: ["hf-inference"],
+		languages: ["sh", "js", "py"],
+	},
 ] as const;
 
 const GET_SNIPPET_FN = {
@@ -119,17 +130,16 @@ function getFixtureFolder(testName: string): string {
 function generateInferenceSnippet(
 	model: snippets.ModelDataMinimal,
 	language: LANGUAGE,
-	provider: InferenceProvider,
+	provider: SnippetInferenceProvider,
 	opts?: Record<string, unknown>
 ): InferenceSnippet[] {
-	const generatedSnippets = GET_SNIPPET_FN[language](model, "api_token", provider, opts);
-	return Array.isArray(generatedSnippets) ? generatedSnippets : [generatedSnippets];
+	return GET_SNIPPET_FN[language](model, "api_token", provider, opts);
 }
 
 async function getExpectedInferenceSnippet(
 	testName: string,
 	language: LANGUAGE,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): Promise<InferenceSnippet[]> {
 	const fixtureFolder = getFixtureFolder(testName);
 	const files = await fs.readdir(fixtureFolder);
@@ -146,7 +156,7 @@ async function getExpectedInferenceSnippet(
 async function saveExpectedInferenceSnippet(
 	testName: string,
 	language: LANGUAGE,
-	provider: InferenceProvider,
+	provider: SnippetInferenceProvider,
 	snippets: InferenceSnippet[]
 ) {
 	const fixtureFolder = getFixtureFolder(testName);
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/0.curl.hf-inference.sh b/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/0.curl.hf-inference.sh
@@ -1,4 +1,4 @@
-curl 'https://api-inference.huggingface.co/models/meta-llama/Llama-3.1-8B-Instruct/v1/chat/completions' \
+curl 'https://router.huggingface.co/hf-inference/models/meta-llama/Llama-3.1-8B-Instruct/v1/chat/completions' \
 -H 'Authorization: Bearer api_token' \
 -H 'Content-Type: application/json' \
 --data '{
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/1.openai.hf-inference.js b/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/1.openai.hf-inference.js
@@ -1,7 +1,7 @@
 import { OpenAI } from "openai";
 
 const client = new OpenAI({
-	baseURL: "https://router.huggingface.co/hf-inference",
+	baseURL: "https://router.huggingface.co/hf-inference/v1",
 	apiKey: "api_token"
 });
 
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/1.openai.hf-inference.py b/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/1.openai.hf-inference.py
@@ -1,7 +1,7 @@
 from openai import OpenAI
 
 client = OpenAI(
-	base_url="https://router.huggingface.co/hf-inference",
+	base_url="https://router.huggingface.co/hf-inference/v1",
 	api_key="api_token"
 )
 
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/0.curl.hf-inference.sh b/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/0.curl.hf-inference.sh
@@ -1,4 +1,4 @@
-curl 'https://api-inference.huggingface.co/models/meta-llama/Llama-3.1-8B-Instruct/v1/chat/completions' \
+curl 'https://router.huggingface.co/hf-inference/models/meta-llama/Llama-3.1-8B-Instruct/v1/chat/completions' \
 -H 'Authorization: Bearer api_token' \
 -H 'Content-Type: application/json' \
 --data '{
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/1.openai.hf-inference.js b/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/1.openai.hf-inference.js
@@ -1,7 +1,7 @@
 import { OpenAI } from "openai";
 
 const client = new OpenAI({
-	baseURL: "https://router.huggingface.co/hf-inference",
+	baseURL: "https://router.huggingface.co/hf-inference/v1",
 	apiKey: "api_token"
 });
 
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/1.openai.hf-inference.py b/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/1.openai.hf-inference.py
@@ -1,7 +1,7 @@
 from openai import OpenAI
 
 client = OpenAI(
-	base_url="https://router.huggingface.co/hf-inference",
+	base_url="https://router.huggingface.co/hf-inference/v1",
 	api_key="api_token"
 )
 
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/0.curl.hf-inference.sh b/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/0.curl.hf-inference.sh
@@ -1,4 +1,4 @@
-curl 'https://api-inference.huggingface.co/models/meta-llama/Llama-3.2-11B-Vision-Instruct/v1/chat/completions' \
+curl 'https://router.huggingface.co/hf-inference/models/meta-llama/Llama-3.2-11B-Vision-Instruct/v1/chat/completions' \
 -H 'Authorization: Bearer api_token' \
 -H 'Content-Type: application/json' \
 --data '{
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/1.openai.hf-inference.js b/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/1.openai.hf-inference.js
@@ -1,7 +1,7 @@
 import { OpenAI } from "openai";
 
 const client = new OpenAI({
-	baseURL: "https://router.huggingface.co/hf-inference",
+	baseURL: "https://router.huggingface.co/hf-inference/v1",
 	apiKey: "api_token"
 });
 
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/1.openai.hf-inference.py b/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/1.openai.hf-inference.py
@@ -1,7 +1,7 @@
 from openai import OpenAI
 
 client = OpenAI(
-	base_url="https://router.huggingface.co/hf-inference",
+	base_url="https://router.huggingface.co/hf-inference/v1",
 	api_key="api_token"
 )
 
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/0.curl.hf-inference.sh b/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/0.curl.hf-inference.sh
@@ -1,4 +1,4 @@
-curl 'https://api-inference.huggingface.co/models/meta-llama/Llama-3.2-11B-Vision-Instruct/v1/chat/completions' \
+curl 'https://router.huggingface.co/hf-inference/models/meta-llama/Llama-3.2-11B-Vision-Instruct/v1/chat/completions' \
 -H 'Authorization: Bearer api_token' \
 -H 'Content-Type: application/json' \
 --data '{
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/1.openai.hf-inference.js b/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/1.openai.hf-inference.js
@@ -1,7 +1,7 @@
 import { OpenAI } from "openai";
 
 const client = new OpenAI({
-	baseURL: "https://router.huggingface.co/hf-inference",
+	baseURL: "https://router.huggingface.co/hf-inference/v1",
 	apiKey: "api_token"
 });
 
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/1.openai.hf-inference.py b/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/1.openai.hf-inference.py
@@ -1,7 +1,7 @@
 from openai import OpenAI
 
 client = OpenAI(
-	base_url="https://router.huggingface.co/hf-inference",
+	base_url="https://router.huggingface.co/hf-inference/v1",
 	api_key="api_token"
 )
 
diff --git a/packages/tasks-gen/snippets-fixtures/text-classification/0.curl.hf-inference.sh b/packages/tasks-gen/snippets-fixtures/text-classification/0.curl.hf-inference.sh
@@ -0,0 +1,5 @@
+curl https://router.huggingface.co/hf-inference/models/distilbert/distilbert-base-uncased-finetuned-sst-2-english \
+	-X POST \
+	-d '{"inputs": "I like you. I love you"}' \
+	-H 'Content-Type: application/json' \
+	-H 'Authorization: Bearer api_token'
diff --git a/packages/tasks-gen/snippets-fixtures/text-classification/0.huggingface.js.hf-inference.js b/packages/tasks-gen/snippets-fixtures/text-classification/0.huggingface.js.hf-inference.js
@@ -0,0 +1,11 @@
+import { HfInference } from "@huggingface/inference";
+
+const client = new HfInference("api_token");
+
+const output = await client.textClassification({
+	model: "distilbert/distilbert-base-uncased-finetuned-sst-2-english",
+	inputs: "I like you. I love you",
+	provider: "hf-inference",
+});
+
+console.log(output);
diff --git a/packages/tasks-gen/snippets-fixtures/text-classification/0.huggingface_hub.hf-inference.py b/packages/tasks-gen/snippets-fixtures/text-classification/0.huggingface_hub.hf-inference.py
@@ -0,0 +1,14 @@
+from huggingface_hub import InferenceClient
+
+client = InferenceClient(
+	provider="hf-inference",
+	api_key="api_token"
+)
+
+result = client.text_classification(
+	model="distilbert/distilbert-base-uncased-finetuned-sst-2-english",
+	inputs="I like you. I love you",
+	provider="hf-inference",
+)
+
+print(result)
diff --git a/packages/tasks-gen/snippets-fixtures/text-classification/1.fetch.hf-inference.js b/packages/tasks-gen/snippets-fixtures/text-classification/1.fetch.hf-inference.js
@@ -0,0 +1,19 @@
+async function query(data) {
+	const response = await fetch(
+		"https://router.huggingface.co/hf-inference/models/distilbert/distilbert-base-uncased-finetuned-sst-2-english",
+		{
+			headers: {
+				Authorization: "Bearer api_token",
+				"Content-Type": "application/json",
+			},
+			method: "POST",
+			body: JSON.stringify(data),
+		}
+	);
+	const result = await response.json();
+	return result;
+}
+
+query({"inputs": "I like you. I love you"}).then((response) => {
+	console.log(JSON.stringify(response));
+});
diff --git a/packages/tasks-gen/snippets-fixtures/text-classification/1.requests.hf-inference.py b/packages/tasks-gen/snippets-fixtures/text-classification/1.requests.hf-inference.py
@@ -0,0 +1,12 @@
+import requests
+
+API_URL = "https://router.huggingface.co/hf-inference/v1"
+headers = {"Authorization": "Bearer api_token"}
+
+def query(payload):
+	response = requests.post(API_URL, headers=headers, json=payload)
+	return response.json()
+	
+output = query({
+	"inputs": "I like you. I love you",
+})
diff --git a/packages/tasks-gen/snippets-fixtures/text-to-image/0.curl.hf-inference.sh b/packages/tasks-gen/snippets-fixtures/text-to-image/0.curl.hf-inference.sh
@@ -1,4 +1,4 @@
-curl https://api-inference.huggingface.co/models/black-forest-labs/FLUX.1-schnell \
+curl https://router.huggingface.co/hf-inference/models/black-forest-labs/FLUX.1-schnell \
 	-X POST \
 	-d '{"inputs": "Astronaut riding a horse"}' \
 	-H 'Content-Type: application/json' \
diff --git a/packages/tasks-gen/snippets-fixtures/text-to-image/1.fetch.hf-inference.js b/packages/tasks-gen/snippets-fixtures/text-to-image/1.fetch.hf-inference.js
@@ -1,6 +1,6 @@
 async function query(data) {
 	const response = await fetch(
-		"https://api-inference.huggingface.co/models/black-forest-labs/FLUX.1-schnell",
+		"https://router.huggingface.co/hf-inference/models/black-forest-labs/FLUX.1-schnell",
 		{
 			headers: {
 				Authorization: "Bearer api_token",
diff --git a/packages/tasks-gen/snippets-fixtures/text-to-image/1.requests.hf-inference.py b/packages/tasks-gen/snippets-fixtures/text-to-image/1.requests.hf-inference.py
@@ -1,6 +1,6 @@
 import requests
 
-API_URL = "https://api-inference.huggingface.co/models/black-forest-labs/FLUX.1-schnell"
+API_URL = "https://router.huggingface.co/hf-inference/v1"
 headers = {"Authorization": "Bearer api_token"}
 
 def query(payload):
diff --git a/packages/tasks/src/inference-providers.ts b/packages/tasks/src/inference-providers.ts
@@ -20,5 +20,6 @@ export const HF_HUB_INFERENCE_PROXY_TEMPLATE = `https://router.huggingface.co/{{
  * TODO(Expose this from HfInference in the future?)
  */
 export function openAIbaseUrl(provider: SnippetInferenceProvider): string {
-	return HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", provider);
+	const url = HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", provider);
+	return provider === "hf-inference" ? `${url}/v1` : url;
 }
diff --git a/packages/tasks/src/library-to-tasks.ts b/packages/tasks/src/library-to-tasks.ts
@@ -3,7 +3,7 @@ import type { PipelineType } from "./pipelines.js";
 
 /**
  * Mapping from library name to its supported tasks.
- * Inference API (serverless) should be disabled for all other (library, task) pairs beyond this mapping.
+ * HF-Inference API (serverless) should be disabled for all other (library, task) pairs beyond this mapping.
  * This mapping is partially generated automatically by "python-api-export-tasks" action in
  * huggingface/api-inference-community repo upon merge. For transformers, the mapping is manually
  * based on api-inference (hf_types.rs).
diff --git a/packages/tasks/src/snippets/curl.ts b/packages/tasks/src/snippets/curl.ts
@@ -17,7 +17,7 @@ export const snippetBasic = (
 		{
 			client: "curl",
 			content: `\
-curl https://api-inference.huggingface.co/models/${model.id} \\
+curl https://router.huggingface.co/hf-inference/models/${model.id} \\
 	-X POST \\
 	-d '{"inputs": ${getModelInputSnippet(model, true)}}' \\
 	-H 'Content-Type: application/json' \\
@@ -41,7 +41,7 @@ export const snippetTextGeneration = (
 	if (model.tags.includes("conversational")) {
 		const baseUrl =
 			provider === "hf-inference"
-				? `https://api-inference.huggingface.co/models/${model.id}/v1/chat/completions`
+				? `https://router.huggingface.co/hf-inference/models/${model.id}/v1/chat/completions`
 				: HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", provider) + "/v1/chat/completions";
 
 		// Conversational model detected, so we display a code snippet that features the Messages API
@@ -92,7 +92,7 @@ export const snippetZeroShotClassification = (
 	return [
 		{
 			client: "curl",
-			content: `curl https://api-inference.huggingface.co/models/${model.id} \\
+			content: `curl https://router.huggingface.co/hf-inference/models/${model.id} \\
 	-X POST \\
 	-d '{"inputs": ${getModelInputSnippet(model, true)}, "parameters": {"candidate_labels": ["refund", "legal", "faq"]}}' \\
 	-H 'Content-Type: application/json' \\
@@ -112,7 +112,7 @@ export const snippetFile = (
 	return [
 		{
 			client: "curl",
-			content: `curl https://api-inference.huggingface.co/models/${model.id} \\
+			content: `curl https://router.huggingface.co/hf-inference/models/${model.id} \\
 	-X POST \\
 	--data-binary '@${getModelInputSnippet(model, true, true)}' \\
 	-H 'Authorization: Bearer ${accessToken || `{API_TOKEN}`}'`,
diff --git a/packages/tasks/src/snippets/js.ts b/packages/tasks/src/snippets/js.ts
diff --git a/packages/tasks/src/snippets/python.ts b/packages/tasks/src/snippets/python.ts
diff --git a/packages/tasks/src/tasks/audio-classification/about.md b/packages/tasks/src/tasks/audio-classification/about.md
diff --git a/packages/tasks/src/tasks/audio-to-audio/about.md b/packages/tasks/src/tasks/audio-to-audio/about.md
diff --git a/packages/tasks/src/tasks/automatic-speech-recognition/about.md b/packages/tasks/src/tasks/automatic-speech-recognition/about.md
diff --git a/packages/tasks/src/tasks/image-text-to-text/about.md b/packages/tasks/src/tasks/image-text-to-text/about.md
diff --git a/packages/tasks/src/tasks/sentence-similarity/about.md b/packages/tasks/src/tasks/sentence-similarity/about.md
diff --git a/packages/tasks/src/tasks/text-to-speech/about.md b/packages/tasks/src/tasks/text-to-speech/about.md

Original file line number	Diff line number	Diff line change
`@@ -1,4 +1,4 @@`
`1`		`-curl 'https://api-inference.huggingface.co/models/meta-llama/Llama-3.1-8B-Instruct/v1/chat/completions' \`
	`1`	`+curl 'https://router.huggingface.co/hf-inference/models/meta-llama/Llama-3.1-8B-Instruct/v1/chat/completions' \`
`2`	`2`	`-H 'Authorization: Bearer api_token' \`
`3`	`3`	`-H 'Content-Type: application/json' \`
`4`	`4`	`--data '{`
Original file line number	Diff line number	Diff line change
`@@ -1,7 +1,7 @@`
`1`	`1`	`from openai import OpenAI`
`2`	`2`
`3`	`3`	`client = OpenAI(`
`4`		`- base_url="https://router.huggingface.co/hf-inference",`
	`4`	`+ base_url="https://router.huggingface.co/hf-inference/v1",`
`5`	`5`	`api_key="api_token"`
`6`	`6`	`)`
`7`	`7`