Merge pull request #17499 from ShirasawaSama/patch-5

feat: Dynamically load @huggingface/transformers to improve first-screen loading speed (-1.9MB)

src/lib/components/admin/Evaluations/Leaderboard.svelte

@@ -1,9 +1,4 @@
 <script lang="ts">
-	import * as ort from 'onnxruntime-web';
-	import { env, AutoModel, AutoTokenizer } from '@huggingface/transformers';
-
-	env.backends.onnx.wasm.wasmPaths = '/wasm/';
-
 	import { onMount, getContext } from 'svelte';
 	import { models } from '$lib/stores';
 
@@ -237,6 +232,11 @@
 	//////////////////////
 
 	const loadEmbeddingModel = async () => {
+		const { env, AutoModel, AutoTokenizer } = await import('@huggingface/transformers');
+		if (env.backends.onnx.wasm) {
+			env.backends.onnx.wasm.wasmPaths = '/wasm/';
+		}
+
 		// Check if the tokenizer and model are already loaded and stored in the window object
 		if (!window.tokenizer) {
 			window.tokenizer = await AutoTokenizer.from_pretrained(EMBEDDING_MODEL);

src/lib/components/chat/Settings/Audio.svelte

@@ -6,7 +6,6 @@
 	import { getVoices as _getVoices } from '$lib/apis/audio';
 
 	import Switch from '$lib/components/common/Switch.svelte';
-	import { round } from '@huggingface/transformers';
 	import Spinner from '$lib/components/common/Spinner.svelte';
 	import Tooltip from '$lib/components/common/Tooltip.svelte';
 	const dispatch = createEventDispatcher();