diff --git a/tests/data/test_template.py b/tests/data/test_template.py
index 1ce1b0d8..4a0dfaa8 100644
--- a/tests/data/test_template.py
+++ b/tests/data/test_template.py
@@ -26,6 +26,7 @@ if TYPE_CHECKING:
 
 
 HF_TOKEN = os.environ.get("HF_TOKEN", None)
+IN_GITHUB_ACTIONS = os.getenv("GITHUB_ACTIONS") == "true"
 
 TINY_LLAMA = os.environ.get("TINY_LLAMA", "llamafactory/tiny-random-Llama-3")
 
@@ -121,7 +122,7 @@ def test_jinja_template(use_fast: bool):
     assert tokenizer.apply_chat_template(MESSAGES) == ref_tokenizer.apply_chat_template(MESSAGES)
 
 
-@pytest.mark.skipif(HF_TOKEN is None, reason="Gated model.")
+@pytest.mark.skipif(IN_GITHUB_ACTIONS or HF_TOKEN is None, reason="Gated model.")
 def test_gemma_template():
     prompt_str = (
         "<bos><start_of_turn>user\nHow are you<end_of_turn>\n"
@@ -133,7 +134,7 @@ def test_gemma_template():
     _check_template("google/gemma-2-9b-it", "gemma", prompt_str, answer_str, extra_str="<end_of_turn>\n")
 
 
-@pytest.mark.skipif(HF_TOKEN is None, reason="Gated model.")
+@pytest.mark.skipif(IN_GITHUB_ACTIONS or HF_TOKEN is None, reason="Gated model.")
 def test_llama3_template():
     prompt_str = (
         "<|begin_of_text|><|start_header_id|>user<|end_header_id|>\n\nHow are you<|eot_id|>"