Merge 6ef9f77789 into bc1fca39f3

tests: use urllib and skip on network failure

nanochat/gpt.py

@@ -8,7 +8,7 @@ Notable features:
 - norm after token embedding
 - no learnable params in rmsnorm
 - no bias in linear layers
-- Multi-Query Attention (MQA) support for more efficient inference
+- Group-Query Attention (GQA) support for more efficient inference
 """
 
 import math
@@ -29,7 +29,7 @@ class GPTConfig:
     vocab_size: int = 50304
     n_layer: int = 12
     n_head: int = 6 # number of query heads
-    n_kv_head: int = 6 # number of key/value heads (MQA)
+    n_kv_head: int = 6 # number of key/value heads (GQA)
     n_embd: int = 768
 
 

tests/test_rustbpe.py

@@ -438,10 +438,12 @@ def enwik8_path():
     enwik8_local_path_zip = os.path.join(base_dir, "enwik8.zip")
     if not os.path.exists(enwik8_local_path):
         print(f"Downloading enwik8 to {enwik8_local_path_zip}")
-        import requests
-        response = requests.get(enwik8_url)
-        with open(enwik8_local_path_zip, "wb") as f:
-            f.write(response.content)
+        import urllib.request, urllib.error
+        try:
+            with urllib.request.urlopen(enwik8_url, timeout=30) as resp, open(enwik8_local_path_zip, "wb") as f:
+                f.write(resp.read())
+        except (urllib.error.URLError, urllib.error.HTTPError) as e:
+            pytest.skip(f"Network unavailable or download failed: {e}")
         with zipfile.ZipFile(enwik8_local_path_zip, "r") as zip_ref:
             zip_ref.extractall(base_dir)
         print(f"Unzipped enwik8 to {enwik8_local_path}")