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++

mac: the setup test question keeps the engine's own prompt window (a 16-token window gave a wrong first word on Metal)

Niko12213939193dd305
diff --git a/tools/dsv41/smoke.py b/tools/dsv41/smoke.py
index 7a2b94e9c..9a0dcc907 100644
--- a/tools/dsv41/smoke.py
+++ b/tools/dsv41/smoke.py
@@ -77,8 +77,12 @@ def main() -> int:
with tempfile.TemporaryDirectory() as d:
f = Path(d) / "ids.txt"
f.write_text(",".join(str(i) for i in ids), encoding="ascii")
- cmd = [a.exe, "--pack", str(pack), "--ids", "@" + str(f), "--n", str(a.tokens), "--max-context", "1024", "--dspark", "none" if sys.platform == "darwin" else "auto",
- "--window", "16"] # (a Mac runs without the drafter: its Metal engine has no kernels for it)
+ mac = sys.platform == "darwin"
+ cmd = [a.exe, "--pack", str(pack), "--ids", "@" + str(f), "--n", str(a.tokens), "--max-context", "1024", "--dspark", "none" if mac else "auto"]
+ # (a Mac runs without the drafter: its Metal engine has no kernels for it, and it keeps the engine's own prompt window: with a window of
+ # 16 or 17 tokens the Metal engine gave a wrong first word in about half the runs; the engine's default window never did)
+ if not mac:
+ cmd += ["--window", "16"]
if a.gpus:
cmd += ["--gpus", a.gpus]
try: