From: Georgi Gerganov <redacted>
Date: Sat, 10 Jan 2026 15:51:56 +0000 (+0200)
Subject: server : adjust unified KV cache tests (#18716)
X-Git-Tag: upstream/0.0.7721~25
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=f307926482a465a6c7af5f212f365d50232d9cfb;p=pkg%2Fggml%2Fsources%2Fllama.cpp

server : adjust unified KV cache tests (#18716)
---

diff --git a/tools/server/tests/unit/test_completion.py b/tools/server/tests/unit/test_completion.py
index ef1757db..2a980601 100644
--- a/tools/server/tests/unit/test_completion.py
+++ b/tools/server/tests/unit/test_completion.py
@@ -393,12 +393,12 @@ def test_completion_unified(n_ctx, n_slots, n_predict_vals, expected_success):
     for res, n_predict, expect_ok in zip(results, n_predict_vals, expected_success):
         if expect_ok:
             assert res.status_code == 200
+
+        # note: https://github.com/ggml-org/llama.cpp/pull/18700#issuecomment-3728695581
+        if res.status_code == 200:
             assert "content" in res.body
             if "timings" in res.body:
                 assert res.body["timings"]["predicted_n"] == n_predict
-        else:
-            assert res.status_code == 500
-            assert "content" not in res.body
 
 
 @pytest.mark.parametrize(