Merge branch 'main' into patch-4

Author: krampstudio

.github/workflows/lint.yml

@@ -42,7 +42,11 @@ jobs:
 
       - name: "Checking format errors"
         run: |
-          pnpm --filter ...[${{ steps.since.outputs.SINCE }}] format:check
+          # Could also use pnpm --filter [${{ steps.since.outputs.SINCE }}] format:check
+          # But this way we can see the diff
+          pnpm --filter [${{ steps.since.outputs.SINCE }}] format
+          git diff | head -n 100
+          git diff --name-only --exit-code
 
       - name: Check types
         run: pnpm --filter ...[${{ steps.since.outputs.SINCE }}] check

.github/workflows/space-header-publish.yml

@@ -0,0 +1,67 @@
+name: Space-header - Version and Release
+
+on:
+  workflow_dispatch:
+    inputs:
+      newversion:
+        type: choice
+        description: "Semantic Version Bump Type"
+        default: patch
+        options:
+          - patch
+          - minor
+          - major
+
+concurrency:
+  group: "push-to-main"
+
+defaults:
+  run:
+    working-directory: packages/space-header
+
+jobs:
+  version_and_release:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v3
+        with:
+          # Needed to push the tag and the commit on the main branch, otherwise we get:
+          # > Run git push --follow-tags
+          # remote: error: GH006: Protected branch update failed for refs/heads/main.
+          # remote: error: Changes must be made through a pull request. Required status check "lint" is expected.
+          token: ${{ secrets.BOT_ACCESS_TOKEN }}
+      - run: corepack enable
+      - uses: actions/setup-node@v3
+        with:
+          node-version: "20"
+          cache: "pnpm"
+          cache-dependency-path: |
+            packages/space-header/pnpm-lock.yaml
+          # setting a registry enables the NODE_AUTH_TOKEN env variable where we can set an npm token.  REQUIRED
+          registry-url: "https://registry.npmjs.org"
+      - run: pnpm install
+      - run: git config --global user.name machineuser
+      - run: git config --global user.email [email protected]
+      - run: |
+          PACKAGE_VERSION=$(node -p "require('./package.json').version")
+          BUMPED_VERSION=$(node -p "require('semver').inc('$PACKAGE_VERSION', '${{ github.event.inputs.newversion }}')")
+          # Update package.json with the new version
+          node -e "const fs = require('fs'); const package = JSON.parse(fs.readFileSync('./package.json')); package.version = '$BUMPED_VERSION'; fs.writeFileSync('./package.json', JSON.stringify(package, null, '\t') + '\n');"
+          git commit . -m "🔖 @hugginface/space-header $BUMPED_VERSION"
+          git tag "space-header-v$BUMPED_VERSION"
+      - run: pnpm publish --no-git-checks .
+        env:
+          NODE_AUTH_TOKEN: ${{ secrets.NPM_TOKEN }}
+      - run: git pull --rebase && git push --follow-tags
+      # hack - reuse actions/setup-node@v3 just to set a new registry
+      - uses: actions/setup-node@v3
+        with:
+          node-version: "20"
+          registry-url: "https://npm.pkg.github.com"
+      - run: pnpm publish --no-git-checks .
+        env:
+          NODE_AUTH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+      - name: "Update Doc"
+        uses: peter-evans/repository-dispatch@v2
+        with:
+          event-type: doc-build

.github/workflows/sync-widgets-demo.yml

@@ -1,15 +1,15 @@
 name: Sync Widgets Demo
 on:
   workflow_dispatch:
-  push:
-    branches:
-      - main
-    paths:
-      - "packages/widgets/**"
-      - "packages/tasks/**"
-      - "package.json"
-      - "pnpm-lock.yaml"
-      - ".github/workflows/sync-widgets-demo.yml"
+  # push:
+  #   branches:
+  #     - main
+  #   paths:
+  #     - "packages/widgets/**"
+  #     - "packages/tasks/**"
+  #     - "package.json"
+  #     - "pnpm-lock.yaml"
+  #     - ".github/workflows/sync-widgets-demo.yml"
 
 jobs:
   sync:

.github/workflows/trufflehog.yml

@@ -0,0 +1,18 @@
+name: Secret Leaks
+
+on:
+  push:
+
+permissions:
+  contents: read
+
+jobs:
+  trufflehog:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+      - name: Secret Scanning
+        uses: trufflesecurity/trufflehog@main

.vscode/settings.json

@@ -13,5 +13,6 @@
 		"editor.defaultFormatter": "esbenp.prettier-vscode"
 	},
 	"cSpell.words": ["huggingface"],
-	"deno.enablePaths": ["./e2e/deno"]
+	"deno.enablePaths": ["./e2e/deno"],
+	"typescript.tsdk": "node_modules/typescript/lib"
 }

CODEOWNERS

@@ -4,11 +4,7 @@
 
 # Ownership for the Tasks Package
 
-/packages/tasks/ @osanseviero @SBrandeis @gary149 @Wauplin @julien-c
-
-# Ownership for the Widgets Package
-
-/packages/widgets/ @mishig25 @SBrandeis
+/packages/tasks/ @osanseviero @SBrandeis @gary149 @Wauplin @julien-c @pcuenca
 
 # Ownership for the Hub Package
 
@@ -25,3 +21,6 @@
 # Ownership for the gguf Package
 
 /packages/gguf @mishig25 @julien-c
+
+# Ownership for the space-header Package
+/packages/space-header @enzostvs

README.md

@@ -54,6 +54,7 @@ This is a collection of JS libraries to interact with the Hugging Face API, with
 - [@huggingface/agents](packages/agents/README.md): Interact with HF models through a natural language interface
 - [@huggingface/gguf](packages/gguf/README.md): A GGUF parser that works on remotely hosted files.
 - [@huggingface/tasks](packages/tasks/README.md): The definition files and source-of-truth for the Hub's main primitives like pipeline tasks, model libraries, etc.
+- [@huggingface/space-header](packages/space-header/README.md): Use the Space `mini_header` outside Hugging Face
 
 
 We use modern features to avoid polyfills and dependencies, so the libraries will only work on modern browsers / Node.js >= 18 / Bun / Deno.
@@ -87,7 +88,7 @@ You can run our packages with vanilla JS, without any bundler, by using a CDN or
 
 ```html
 <script type="module">
-    import { HfInference } from 'https://cdn.jsdelivr.net/npm/@huggingface/inference@2.7.0/+esm';
+    import { HfInference } from 'https://cdn.jsdelivr.net/npm/@huggingface/inference@2.8.0/+esm';
     import { createRepo, commit, deleteRepo, listFiles } from "https://cdn.jsdelivr.net/npm/@huggingface/[email protected]/+esm";
 </script>
 ```

docs/_toctree.yml

@@ -21,3 +21,13 @@
       title: Use Agents to run multi-modal workflows from a natural language API
     - local: agents/modules
       title: API Reference
+- title: "@huggingface/space-header"
+  isExpanded: true
+  sections:
+    - local: space-header/README
+      title: Use Space mini_header in your app
+- title: "@huggingface/gguf"
+  isExpanded: true
+  sections:
+    - local: gguf/README
+      title: Parse local and remote GGUF files

package.json

@@ -17,6 +17,7 @@
 		"eslint-config-prettier": "^9.0.0",
 		"eslint-plugin-prettier": "^4.2.1",
 		"eslint-plugin-svelte": "^2.30.0",
+		"npm-run-all": "^4.1.5",
 		"prettier": "^3.1.0",
 		"prettier-plugin-svelte": "^3.1.2",
 		"semver": "^7.5.0",

packages/doc-internal/package.json

@@ -5,14 +5,16 @@
 	"description": "Package to generate doc for other @huggingface packages",
 	"private": true,
 	"scripts": {
-		"start": "pnpm run fix-cdn-versions && pnpm run doc-hub && pnpm run doc-inference && pnpm run doc-agents && cp ../../README.md ../../docs/index.md && pnpm run update-toc && pnpm run fix-md-links && pnpm run fix-md-headinghashlinks",
+		"start": "pnpm run fix-cdn-versions && pnpm run doc-hub && pnpm run doc-inference && pnpm run doc-agents && pnpm run doc-space-header && pnpm run doc-gguf && cp ../../README.md ../../docs/index.md && pnpm run update-toc && pnpm run fix-md-links && pnpm run fix-md-headinghashlinks",
 		"lint": "eslint --quiet --fix --ext .cjs,.ts .",
 		"lint:check": "eslint --ext .cjs,.ts .",
 		"format": "prettier --write .",
 		"format:check": "prettier --check .",
 		"doc-hub": "typedoc --tsconfig ../hub/tsconfig.json --githubPages false --plugin typedoc-plugin-markdown --out ../../docs/hub --hideBreadcrumbs --hideInPageTOC --sourceLinkTemplate https://github.com/huggingface/huggingface.js/blob/main/{path}#L{line} ../hub/index.ts",
 		"doc-inference": "typedoc --tsconfig ../inference/tsconfig.json --githubPages false --plugin typedoc-plugin-markdown --out ../../docs/inference --hideBreadcrumbs --hideInPageTOC --sourceLinkTemplate https://github.com/huggingface/huggingface.js/blob/main/{path}#L{line} ../inference/src/index.ts",
 		"doc-agents": "typedoc --tsconfig ../agents/tsconfig.json --githubPages false --plugin typedoc-plugin-markdown --out ../../docs/agents --hideBreadcrumbs --hideInPageTOC --sourceLinkTemplate https://github.com/huggingface/huggingface.js/blob/main/{path}#L{line} ../agents/src/index.ts",
+		"doc-gguf": "mkdir -p ../../docs/gguf && cp ../../packages/gguf/README.md ../../docs/gguf/README.md",
+		"doc-space-header": "mkdir -p ../../docs/space-header && cp ../../packages/space-header/README.md ../../docs/space-header/README.md",
 		"update-toc": "tsx update-toc.ts",
 		"fix-cdn-versions": "tsx fix-cdn-versions.ts",
 		"fix-md-links": "tsx fix-md-links.ts",

packages/gguf/package.json

@@ -1,7 +1,7 @@
 {
 	"name": "@huggingface/gguf",
 	"packageManager": "[email protected]",
-	"version": "0.1.5",
+	"version": "0.1.10",
 	"description": "a GGUF parser that works on remotely hosted files",
 	"repository": "https://github.com/huggingface/huggingface.js.git",
 	"publishConfig": {

packages/gguf/scripts/generate-llm.ts

@@ -5,7 +5,7 @@
 
 import { writeFileSync } from "node:fs";
 
-const SOURCE_CPP_URL = "https://raw.githubusercontent.com/ggerganov/llama.cpp/master/llama.cpp";
+const SOURCE_CPP_URL = "https://raw.githubusercontent.com/ggerganov/llama.cpp/master/src/llama.cpp";
 const DEST_FILE_PATH = "./src/transformer-llm.ts";
 const DEST_COMMON_SOURCE = `
 /** This file is auto-generated by generate-llm.ts */
@@ -80,6 +80,20 @@ const KV_TYPE = {
 	LLM_KV_SSM_STATE_SIZE: "number",
 	LLM_KV_SSM_TIME_STEP_RANK: "number",
 	LLM_KV_LOGIT_SCALE: "number",
+	LLM_KV_EXPERT_FEED_FORWARD_LENGTH: "number",
+	LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH: "number",
+	LLM_KV_ATTENTION_SLIDING_WINDOW: "number",
+	LLM_KV_ATTN_LOGIT_SOFTCAPPING: "number",
+	LLM_KV_FINAL_LOGIT_SOFTCAPPING: "number",
+	LLM_KV_LEADING_DENSE_BLOCK_COUNT: "number",
+	LLM_KV_ATTENTION_KV_LORA_RANK: "number",
+	LLM_KV_EXPERT_SHARED_COUNT: "number",
+	LLM_KV_EXPERT_WEIGHTS_SCALE: "number",
+	LLM_KV_ROPE_SCALING_YARN_LOG_MUL: "number",
+	LLM_KV_ATTENTION_Q_LORA_RANK: "number",
+	LLM_KV_ATTENTION_RELATIVE_BUCKETS_COUNT: "number",
+	LLM_KV_DECODER_START_TOKEN_ID: "number",
+	LLM_KV_USE_PARALLEL_RESIDUAL: "boolean",
 };
 
 interface Arch {
@@ -199,7 +213,12 @@ async function main() {
 			if (a.hparams.length) {
 				code += [
 					" & {",
-					...a.hparams.map((k) => `\t${JSON.stringify(constToKVName[k].replace("%s", a.name))}: ${KV_TYPE[k]},`),
+					...a.hparams.map((k) => {
+						if (!KV_TYPE[k]) {
+							throw new Error(`Cannot find type definition of ${k}`);
+						}
+						return `\t${JSON.stringify(constToKVName[k].replace("%s", a.name))}: ${KV_TYPE[k]},`;
+					}),
 					"};",
 				].join("\n");
 			} else {

packages/gguf/src/gguf.spec.ts

@@ -1,5 +1,6 @@
-import { describe, expect, it } from "vitest";
-import { GGMLQuantizationType, gguf, ggufAllShards, parseGgufShardFilename } from "./gguf";
+import { beforeAll, describe, expect, it } from "vitest";
+import type { GGUFParseOutput } from "./gguf";
+import { GGMLFileQuantizationType, GGMLQuantizationType, gguf, ggufAllShards, parseGgufShardFilename } from "./gguf";
 import fs from "node:fs";
 
 const URL_LLAMA = "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGUF/resolve/191239b/llama-2-7b-chat.Q2_K.gguf";
@@ -12,8 +13,21 @@ const URL_V1 =
 	"https://huggingface.co/tmadge/testing/resolve/66c078028d1ff92d7a9264a1590bc61ba6437933/tinyllamas-stories-260k-f32.gguf";
 const URL_SHARDED_GROK =
 	"https://huggingface.co/Arki05/Grok-1-GGUF/resolve/ecafa8d8eca9b8cd75d11a0d08d3a6199dc5a068/grok-1-IQ3_XS-split-00001-of-00009.gguf";
+const URL_BIG_METADATA = "https://huggingface.co/ngxson/test_gguf_models/resolve/main/gguf_test_big_metadata.gguf";
 
 describe("gguf", () => {
+	beforeAll(async () => {
+		// download the gguf for "load file" test, save to .cache directory
+		if (!fs.existsSync(".cache")) {
+			fs.mkdirSync(".cache");
+		}
+		if (!fs.existsSync(".cache/model.gguf")) {
+			const res = await fetch(URL_BIG_METADATA);
+			const arrayBuf = await res.arrayBuffer();
+			fs.writeFileSync(".cache/model.gguf", Buffer.from(arrayBuf));
+		}
+	});
+
 	it("should parse a llama2 7b", async () => {
 		const { metadata, tensorInfos } = await gguf(URL_LLAMA);
 
@@ -25,7 +39,7 @@ describe("gguf", () => {
 			tensor_count: 291n,
 			kv_count: 19n,
 			"general.architecture": "llama",
-			"general.file_type": 10,
+			"general.file_type": GGMLFileQuantizationType.MOSTLY_Q2_K,
 			"general.name": "LLaMA v2",
 			"general.quantization_version": 2,
 			"llama.attention.head_count": 32,
@@ -84,7 +98,7 @@ describe("gguf", () => {
 			tensor_count: 291n,
 			kv_count: 24n,
 			"general.architecture": "llama",
-			"general.file_type": 17,
+			"general.file_type": GGMLFileQuantizationType.MOSTLY_Q5_K_M,
 			"general.name": "mistralai_mistral-7b-instruct-v0.2",
 			"general.quantization_version": 2,
 			"llama.attention.head_count": 32,
@@ -122,7 +136,7 @@ describe("gguf", () => {
 			tensor_count: 164n,
 			kv_count: 21n,
 			"general.architecture": "gemma",
-			"general.file_type": GGMLQuantizationType.Q8_K, // 15
+			"general.file_type": GGMLFileQuantizationType.MOSTLY_Q4_K_M,
 			"general.name": "gemma-2b-it",
 			"general.quantization_version": 2,
 			"gemma.attention.head_count": 8,
@@ -159,7 +173,7 @@ describe("gguf", () => {
 			tensor_count: 197n,
 			kv_count: 23n,
 			"general.architecture": "bert",
-			"general.file_type": GGMLQuantizationType.F16,
+			"general.file_type": GGMLFileQuantizationType.MOSTLY_F16,
 			"general.name": "bge-small-en-v1.5",
 			"bert.attention.causal": false,
 			"bert.attention.head_count": 12,
@@ -228,16 +242,10 @@ describe("gguf", () => {
 	});
 
 	it("should parse a local file", async () => {
-		// download the file and save to .cache folder
-		if (!fs.existsSync(".cache")) {
-			fs.mkdirSync(".cache");
-		}
-		const res = await fetch(URL_V1);
-		const arrayBuf = await res.arrayBuffer();
-		fs.writeFileSync(".cache/model.gguf", Buffer.from(arrayBuf));
-
-		const { metadata } = await gguf(".cache/model.gguf", { allowLocalFile: true });
-		expect(metadata).toMatchObject({ "general.name": "tinyllamas-stories-260k" });
+		const parsedGguf = await gguf(".cache/model.gguf", { allowLocalFile: true });
+		const { metadata } = parsedGguf as GGUFParseOutput<{ strict: false }>; // custom metadata arch, no need for typing
+		expect(metadata["dummy.1"]).toBeDefined(); // first metadata in the list
+		expect(metadata["dummy.32767"]).toBeDefined(); // last metadata in the list
 	});
 
 	it("should detect sharded gguf filename", async () => {

packages/gguf/src/gguf.ts

@@ -4,11 +4,12 @@ import { isBackend } from "./utils/isBackend";
 import { promisesQueue } from "./utils/promisesQueue";
 
 export type { MetadataBaseValue, MetadataValue, Version, GGUFMetadata, GGUFTensorInfo, GGUFParseOutput } from "./types";
-export { GGUFValueType, GGMLQuantizationType, Architecture } from "./types";
+export { GGUFValueType, GGMLFileQuantizationType, GGMLQuantizationType, Architecture } from "./types";
 export { GGUF_QUANT_DESCRIPTIONS } from "./quant-descriptions";
 
 export const RE_GGUF_FILE = /\.gguf$/;
 export const RE_GGUF_SHARD_FILE = /^(?<prefix>.*?)-(?<shard>\d{5})-of-(?<total>\d{5})\.gguf$/;
+const PARALLEL_DOWNLOADS = 20;
 
 export interface GgufShardFileInfo {
 	prefix: string;
@@ -143,6 +144,7 @@ class RangeViewLocalFile extends RangeView {
 		const range = [this.chunk * HTTP_CHUNK_SIZE, (this.chunk + 1) * HTTP_CHUNK_SIZE - 1];
 		const buffer = await blob.slice(range[0], range[1]).arrayBuffer();
 		this.appendBuffer(new Uint8Array(buffer));
+		this.chunk += 1;
 	}
 }
 
@@ -400,8 +402,13 @@ export async function ggufAllShards(
 		 */
 		fetch?: typeof fetch;
 		additionalFetchHeaders?: Record<string, string>;
+		parallelDownloads?: number;
 	}
 ): Promise<{ shards: GGUFParseOutput[]; parameterCount: number }> {
+	const parallelDownloads = params?.parallelDownloads ?? PARALLEL_DOWNLOADS;
+	if (parallelDownloads < 1) {
+		throw new TypeError("parallelDownloads must be greater than 0");
+	}
 	const ggufShardFileInfo = parseGgufShardFilename(url);
 	if (ggufShardFileInfo) {
 		const total = parseInt(ggufShardFileInfo.total);
@@ -412,10 +419,9 @@ export async function ggufAllShards(
 			urls.push(`${prefix}-${shardIdx.toString().padStart(5, "0")}-of-${total.toString().padStart(5, "0")}.gguf`);
 		}
 
-		const PARALLEL_DOWNLOADS = 20;
 		const shards = await promisesQueue(
 			urls.map((shardUrl) => () => gguf(shardUrl, { ...params, computeParametersCount: true })),
-			PARALLEL_DOWNLOADS
+			parallelDownloads
 		);
 		return {
 			shards,