Fix caching for LFS files from the Hugging Face Hub (#251)

* Fix model caching for LFS files from the HF Hub

* Ignore local model check on demo site

examples/demo-site/src/worker.js

@@ -4,7 +4,8 @@
 // Needed to ensure the UI thread is not blocked when running //
 /////////////////////////////////////////////////////////////////
 
-import { pipeline } from "@xenova/transformers";
+import { pipeline, env } from "@xenova/transformers";
+env.allowLocalModels = false;
 
 // Define task function mapping
 const TASK_FUNCTION_MAPPING = {

src/utils/hub.js

@@ -410,8 +410,8 @@ export async function getModelFile(path_or_repo_id, filename, fatal = true, opti
  let cacheKey;
  let proposedCacheKey = cache instanceof FileCache ? fsCacheKey : remoteURL;
 
- /** @type {Response|undefined} */
- let responseToCache;
+ // Whether to cache the final response in the end.
+ let toCacheResponse = false;
 
  /** @type {Response|FileResponse|undefined} */
  let response;
@@ -475,14 +475,14 @@ export async function getModelFile(path_or_repo_id, filename, fatal = true, opti
  cacheKey = proposedCacheKey;
  }
 
-
- if (cache && response instanceof Response && response.status === 200) {
- // only clone if cache available, and response is valid
- responseToCache = response.clone();
- }
+ // Only cache the response if:
+ toCacheResponse =
+ cache // 1. A caching system is available
+ && typeof Response !== 'undefined' // 2. `Response` is defined (i.e., we are in a browser-like environment)
+ && response instanceof Response // 3. result is a `Response` object (i.e., not a `FileResponse`)
+ && response.status === 200 // 4. request was successful (status code 200)
  }
 
-
  // Start downloading
  dispatchCallback(options.progress_callback, {
  status: 'download',
@@ -499,16 +499,18 @@ export async function getModelFile(path_or_repo_id, filename, fatal = true, opti
  })
  })
 
-
  if (
  // Only cache web responses
  // i.e., do not cache FileResponses (prevents duplication)
- responseToCache && cacheKey
+ toCacheResponse && cacheKey
  &&
  // Check again whether request is in cache. If not, we add the response to the cache
  (await cache.match(cacheKey) === undefined)
  ) {
- await cache.put(cacheKey, responseToCache)
+ // NOTE: We use `new Response(buffer, ...)` instead of `response.clone()` to handle LFS files
+ await cache.put(cacheKey, new Response(buffer, {
+ headers: response.headers
+ }))
  .catch(err => {
  // Do not crash if unable to add to cache (e.g., QuotaExceededError).
  // Rather, log a warning and proceed with execution.