Remove resolve/main from model record input

mlc-ai · May 30, 2024 · c5adc8c · c5adc8c
1 parent 45f41e6
commit c5adc8c
Show file tree

Hide file tree

Showing 7 changed files with 114 additions and 90 deletions.
diff --git a/examples/function-calling/src/function_calling.ts b/examples/function-calling/src/function_calling.ts
@@ -13,7 +13,7 @@ async function main() {
     model_list: [
       {
         model:
-          "https://huggingface.co/mlc-ai/gorilla-openfunctions-v2-q4f16_1-MLC/resolve/main/",
+          "https://huggingface.co/mlc-ai/gorilla-openfunctions-v2-q4f16_1-MLC",
         model_id: "gorilla-openfunctions-v2-q4f16_1",
         model_lib:
           "https://raw.githubusercontent.com/mlc-ai/binary-mlc-llm-libs/main/gorilla-openfunctions-v2/gorilla-openfunctions-v2-q4f16_1.wasm",

diff --git a/examples/get-started/src/get_started.ts b/examples/get-started/src/get_started.ts
@@ -23,7 +23,7 @@ async function main() {
   // const appConfig: webllm.AppConfig = {
   //   model_list: [
   //     {
-  //       "model": "https://huggingface.co/mlc-ai/Llama-3-8B-Instruct-q4f32_1-MLC/resolve/main/",
+  //       "model": "https://huggingface.co/mlc-ai/Llama-3-8B-Instruct-q4f32_1-MLC",
   //       "model_id": "Llama-3-8B-Instruct-q4f32_1-MLC",
   //       "model_lib": webllm.modelLibURLPrefix + webllm.modelVersion + "/Llama-3-8B-Instruct-q4f32_1-ctx4k_cs1k-webgpu.wasm",
   //     },

diff --git a/src/cache_util.ts b/src/cache_util.ts
@@ -1,5 +1,6 @@
 import * as tvmjs from "tvmjs";
 import { AppConfig, ModelRecord, prebuiltAppConfig } from "./config";
+import { cleanModelUrl } from "./support";
 
 function findModelRecord(modelId: string, appConfig?: AppConfig): ModelRecord {
   const matchedItem = appConfig?.model_list.find(
@@ -19,7 +20,7 @@ export async function hasModelInCache(
     appConfig = prebuiltAppConfig;
   }
   const modelRecord = findModelRecord(modelId, appConfig);
-  const modelUrl = modelRecord.model;
+  const modelUrl = cleanModelUrl(modelRecord.model);
   const cacheType = appConfig.useIndexedDBCache ? "indexeddb" : "cache";
   return tvmjs.hasNDArrayInCache(modelUrl, "webllm/model", cacheType);
 }
@@ -49,20 +50,17 @@ export async function deleteModelInCache(
     appConfig = prebuiltAppConfig;
   }
   const modelRecord = findModelRecord(modelId, appConfig);
+  const modelUrl = cleanModelUrl(modelRecord.model);
   let modelCache: tvmjs.ArtifactCacheTemplate;
   if (appConfig.useIndexedDBCache) {
-    tvmjs.deleteNDArrayCache(modelRecord.model, "webllm/model", "indexeddb");
+    tvmjs.deleteNDArrayCache(modelUrl, "webllm/model", "indexeddb");
     modelCache = new tvmjs.ArtifactIndexedDBCache("webllm/model");
   } else {
-    tvmjs.deleteNDArrayCache(modelRecord.model, "webllm/model", "cache");
+    tvmjs.deleteNDArrayCache(modelUrl, "webllm/model", "cache");
     modelCache = new tvmjs.ArtifactCache("webllm/model");
   }
-  await modelCache.deleteInCache(
-    new URL("tokenizer.model", modelRecord.model).href,
-  );
-  await modelCache.deleteInCache(
-    new URL("tokenizer.json", modelRecord.model).href,
-  );
+  await modelCache.deleteInCache(new URL("tokenizer.model", modelUrl).href);
+  await modelCache.deleteInCache(new URL("tokenizer.json", modelUrl).href);
 }
 
 export async function deleteChatConfigInCache(
@@ -80,7 +78,8 @@ export async function deleteChatConfigInCache(
   } else {
     configCache = new tvmjs.ArtifactCache("webllm/config");
   }
-  const configUrl = new URL("mlc-chat-config.json", modelRecord.model).href;
+  const modelUrl = cleanModelUrl(modelRecord.model);
+  const configUrl = new URL("mlc-chat-config.json", modelUrl).href;
   await configCache.deleteInCache(configUrl);
 }