[Bugfix] In LongRoPE, decide short vs long based on max_model_len (#27431)

Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
2025-10-28 08:00:56 -04:00
parent 7a865f2325
commit 44b5ce956d
3 changed files with 39 additions and 11 deletions
--- a/tests/entrypoints/openai/test_default_mm_loras.py
+++ b/tests/entrypoints/openai/test_default_mm_loras.py
@ -29,7 +29,7 @@ def multimodal_server():  # noqa: F811
        "--dtype",
        "half",
        "--max-model-len",
-        "12800",
+        "4096",
        "--enforce-eager",
        # lora config below
        "--enable-lora",