Check with the quant tag

Signed-off-by: Adrien Gallouët <angt@huggingface.co>
2026-03-20 17:25:18 +00:00 · 2026-03-20 17:25:18 +00:00 · e915644c7b
parent 6fd16ba05c
commit e915644c7b
1 changed files with 2 additions and 2 deletions
--- a/tools/server/tests/unit/test_router.py
+++ b/tools/server/tests/unit/test_router.py
@ -103,8 +103,8 @@ def test_router_models_max_evicts_lru():

    candidate_models = [
        "ggml-org/tinygemma3-GGUF:Q8_0",
-        "ggml-org/test-model-stories260K",
-        "ggml-org/test-model-stories260K-infill",
+        "ggml-org/test-model-stories260K:F32",
+        "ggml-org/test-model-stories260K-infill:F32",
    ]

    # Load only the first 2 models to fill the cache