From c728db5f35fc5588b228bffcbdf6d336623dc413 Mon Sep 17 00:00:00 2001
From: sora <2075279110@qq.com>
Date: Thu, 10 Sep 2026 10:51:09 +0000
Subject: [PATCH] HF hub downloads: reachability probe fails fast with the
 mirror fix hint (was minutes of silent retries looking like a hang); status
 lines around hub downloads (start + records + elapsed)

Co-Authored-By: Claude <noreply@anthropic.com>
---
 evalharness/data/loader.py | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)

diff --git a/evalharness/data/loader.py b/evalharness/data/loader.py
index 066ca38..98dfa5d 100644
--- a/evalharness/data/loader.py
+++ b/evalharness/data/loader.py
@@ -14,6 +14,7 @@ import csv
 import json
 import os
 import sys
+import time
 import re
 import shutil
 import urllib.request
@@ -449,14 +450,36 @@ def _load_from_hub(spec: DatasetSpec) -> List[Dict[str, Any]]:
             f'dataset {spec.name!r} lives on a hub ({spec.source!r}); '
             'install it first: pip install evalharness (light deps are default)'
         )
+    _probe_hub_reachable(spec)
     kwargs = {k: v for k, v in spec.params.items() if k not in _RESERVED_PARAMS}
     # filter_column subsets select rows by column value post-load, so the hub
     # load itself must always use the default config.
     subset = None if (spec.subset == 'default' or spec.params.get('filter_column')) else spec.subset
+    print(f'· downloading {spec.source} from HF hub ({_hf_base()}) ...', flush=True)
+    t0 = time.monotonic()
     ds = datasets.load_dataset(spec.source, subset, split=spec.split, revision=spec.version, **kwargs)
+    print(f'· downloaded + parsed {len(ds)} records in {time.monotonic() - t0:.0f}s', flush=True)
     return [dict(r) for r in ds]
 
 
+def _probe_hub_reachable(spec: DatasetSpec) -> None:
+    """Fail fast when the HF endpoint is unreachable -- otherwise the hub
+    client retries silently for minutes and looks like a hang."""
+    base = _hf_base()
+    try:
+        req = urllib.request.Request(base, method='HEAD',
+                                     headers={'User-Agent': 'evalharness/0.1'})
+        urllib.request.urlopen(req, timeout=8)
+        return
+    except Exception as e:
+        raise RuntimeError(
+            f'HuggingFace endpoint unreachable: {base}\n'
+            f'  reason: {type(e).__name__}: {str(e)[:80]}\n'
+            f'  dataset {spec.source!r} cannot download. Fix:\n'
+            f'  evalharness eval run ... --hf-endpoint https://hf-mirror.com\n'
+            f'  (or export HF_ENDPOINT=https://hf-mirror.com)') from e
+
+
 def field_spec_to_record_fn(fields: FieldSpec) -> Callable[[Dict[str, Any]], Sample]:
     """Build the default record->Sample converter from a FieldSpec."""