diff --git a/.github/workflows/checks.yml b/.github/workflows/checks.yml
index e5e595052..ef554b1b4 100644
--- a/.github/workflows/checks.yml
+++ b/.github/workflows/checks.yml
@@ -42,6 +42,13 @@ permissions:
   actions: write
   contents: write
 
+# Cancel in-progress runs on the same PR when a new push arrives.
+# Push-to-main, tag, and workflow_call events are not cancelled so that
+# release and deploy jobs always run to completion.
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.event_name == 'pull_request' }}
+
 jobs:
   compatibility-checks:
     name: Compatibility Checks
diff --git a/transformer_lens/utilities/hf_utils.py b/transformer_lens/utilities/hf_utils.py
index 580255b1a..b39bb08c7 100644
--- a/transformer_lens/utilities/hf_utils.py
+++ b/transformer_lens/utilities/hf_utils.py
@@ -11,6 +11,7 @@
 import os
 import shutil
 import stat
+import time
 from typing import Any, Callable, Dict
 
 import torch
@@ -75,13 +76,27 @@ def download_file_from_hf(
 
     If it's a Torch file without the ".pth" extension, set force_is_torch=True to load it as a Torch object.
     """
-    file_path = hf_hub_download(
-        repo_id=repo_name,
-        filename=file_name,
-        subfolder=subfolder,
-        cache_dir=cache_dir,
-        **select_compatible_kwargs(kwargs, hf_hub_download),
-    )
+    max_retries = 3
+    for attempt in range(max_retries + 1):
+        try:
+            file_path = hf_hub_download(
+                repo_id=repo_name,
+                filename=file_name,
+                subfolder=subfolder,
+                cache_dir=cache_dir,
+                **select_compatible_kwargs(kwargs, hf_hub_download),
+            )
+            break
+        except Exception as exc:
+            if "429" in str(exc) and attempt < max_retries:
+                wait = 10 * (attempt + 1)
+                print(
+                    f"HuggingFace rate limited (429). Retrying in {wait}s "
+                    f"({attempt + 1}/{max_retries})..."
+                )
+                time.sleep(wait)
+            else:
+                raise
 
     if file_path.endswith(".pth") or force_is_torch:
         return torch.load(file_path, map_location="cpu", weights_only=False)