Skip to content

Qwen3.8-Flash-Next on any consumer hardware: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.

C++

tests: resident-mode detection

Niko12215f054d2a2c12
diff --git a/tests/tuning/test_session.py b/tests/tuning/test_session.py
index 9d6a526ec..3474b801a 100644
--- a/tests/tuning/test_session.py
+++ b/tests/tuning/test_session.py
@@ -104,6 +104,10 @@ class ParsingTest(unittest.TestCase):
self.assertEqual(failure_kind("", -9)[0], "OOM")
self.assertEqual(failure_kind("something else broke", 1)[0], "FAIL_BOOT")
+ def test_gpu_cache_wording_is_not_the_low_ram_mode(self):
+ self.assertFalse(parse_facts("R4 hit path ON - resident experts are computed on the GPU")["resident_ram"])
+ self.assertTrue(parse_facts("WARNING: the whole resident RAM mode does not fit (FileExpertSource: resident complement 34.47 GiB")["resident_ram"])
+
def test_facts(self):
f = parse_facts("strata generate: 6 CPU threads\nhybrid experts: 37% on the GPU\n")
self.assertEqual(f["workers"], 6)