Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦
C
#!/usr/bin/env python3
import unittest
from pathlib import Path
ROOT = Path(__file__).resolve().parents[2]
SRC = ROOT / "c" / "glm53.c"
class Glm53MetalStatsSourceTest(unittest.TestCase):
@classmethod
def setUpClass(cls):
cls.src = SRC.read_text()
def test_counters_declared(self):
for needle in (
"g_metal_moe_attempt",
"g_metal_moe_ok",
"g_metal_moe_fallback",
"g_metal_moe_rows",
):
self.assertIn(needle, self.src)
def test_attempt_counted_at_routed_dispatch(self):
self.assertIn("g_metal_moe_attempt++;", self.src)
self.assertIn("metal_done = coli_metal_moe_block_clamped(", self.src)
def test_success_and_fallback_counted(self):
self.assertIn("g_metal_moe_ok++;", self.src)
self.assertIn("g_metal_moe_rows += (uint64_t)R;", self.src)
self.assertIn("if (!metal_done) g_metal_moe_fallback++;", self.src)
def test_verbose_summary_emitted(self):
self.assertIn(
'printf("metal moe attempts %llu ok %llu fallback %llu rows %llu\\n",',
self.src,
)
if __name__ == "__main__":
unittest.main()