From: Georgi Gerganov <redacted>
Date: Wed, 28 Jan 2026 07:11:40 +0000 (+0200)
Subject: server : adjust spec tests to generate up to 16 tokens (#19093)
X-Git-Tag: upstream/0.0.8067~213
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=b931f81b5a3bc3e16bd74cebc8fee8cbd69f8d4d;p=pkg%2Fggml%2Fsources%2Fllama.cpp

server : adjust spec tests to generate up to 16 tokens (#19093)
---

diff --git a/tools/server/tests/unit/test_speculative.py b/tools/server/tests/unit/test_speculative.py
index d2f3fba5f..eebd3cc8f 100644
--- a/tools/server/tests/unit/test_speculative.py
+++ b/tools/server/tests/unit/test_speculative.py
@@ -30,6 +30,7 @@ def test_with_and_without_draft():
         "prompt": "I believe the meaning of life is",
         "temperature": 0.0,
         "top_k": 1,
+        "n_predict": 16,
     })
     assert res.status_code == 200
     content_no_draft = res.body["content"]
@@ -42,6 +43,7 @@ def test_with_and_without_draft():
         "prompt": "I believe the meaning of life is",
         "temperature": 0.0,
         "top_k": 1,
+        "n_predict": 16,
     })
     assert res.status_code == 200
     content_draft = res.body["content"]
@@ -68,6 +70,7 @@ def test_different_draft_min_draft_max():
             "prompt": "I believe the meaning of life is",
             "temperature": 0.0,
             "top_k": 1,
+            "n_predict": 16,
         })
         assert res.status_code == 200
         if last_content is not None: