Analyze Qwen235 fixed PD simulator state
This commit is contained in:
@@ -187,6 +187,42 @@ class FidelityEnvelopeTest(unittest.TestCase):
|
||||
base, metrics_root=Path("/new/metrics"), run_id="state-run"
|
||||
)
|
||||
|
||||
def test_qwen235_real_state_proxy_parser(self) -> None:
|
||||
module = load("analyze_qwen235_fixed_pd_state.py")
|
||||
line = (
|
||||
"Engine 000: Avg prompt throughput: 3276.8 tokens/s, "
|
||||
"Avg generation throughput: 204.8 tokens/s, Running: 5 reqs, "
|
||||
"Waiting: 0 reqs, GPU KV cache usage: 5.6%, Prefix cache hit rate: 0.0%\n"
|
||||
)
|
||||
with tempfile.TemporaryDirectory() as temporary:
|
||||
path = Path(temporary) / "server.log"
|
||||
path.write_text(line)
|
||||
self.assertEqual(
|
||||
module.parse_real_log(path),
|
||||
[
|
||||
{
|
||||
"prompt_tokens_per_s": 3276.8,
|
||||
"generation_tokens_per_s": 204.8,
|
||||
"running": 5,
|
||||
"waiting": 0,
|
||||
"kv_percent": 5.6,
|
||||
}
|
||||
],
|
||||
)
|
||||
|
||||
def test_qwen235_component_categories_are_additive(self) -> None:
|
||||
module = load("analyze_qwen235_fixed_pd_state.py")
|
||||
components = {
|
||||
name: 1.0
|
||||
for names in module.CATEGORIES.values()
|
||||
for name in names
|
||||
}
|
||||
categorized = module.categorized_components(components)
|
||||
self.assertEqual(sum(categorized.values()), float(len(components)))
|
||||
components["unknown_graph_overhead"] = 1.0
|
||||
with self.assertRaisesRegex(ValueError, "component schema drift"):
|
||||
module.categorized_components(components)
|
||||
|
||||
def test_materialize_qwen235_allreduce_requires_serving_contract(self) -> None:
|
||||
module = load("materialize_qwen235_v020_allreduce.py")
|
||||
with tempfile.TemporaryDirectory() as temporary:
|
||||
|
||||
Reference in New Issue
Block a user