[Qwen4Exp] Support UVA PLE-offload and Engram tensor parallelism (#54371)

Signed-off-by: huanghaoyan.hhy <huanghaoyan.hhy@alibaba-inc.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Harry Huang
2026-09-09 07:32:34 -07:00
committed by GitHub
co-authored by mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent 62f3bf58a5
commit 3116c5d06b
16 changed files with 1712 additions and 495 deletions
+33
View File
@@ -46,6 +46,39 @@ def test_optional_type():
assert optional_type_func("42") == 42
@pytest.mark.parametrize(
"options",
[
["--engram-config", '{"cpu_offload": false, "embedding_across_dp": true}'],
[
"--engram-config.cpu_offload",
"false",
"--engram-config.embedding_across_dp",
"true",
],
],
)
def test_engram_config_cli(options, monkeypatch):
"""CLI settings take precedence over the legacy offload environment."""
monkeypatch.setenv("VLLM_PLE_CPU_OFFLOAD", "1")
parser = EngineArgs.add_cli_args(FlexibleArgumentParser())
args = EngineArgs.from_cli_args(parser.parse_args(options))
assert args.engram_config is not None
assert args.engram_config.cpu_offload is False
assert args.engram_config.embedding_across_dp is True
@pytest.mark.parametrize(
"options, provided",
[([], False), (["--engram-config", "{}"], True)],
)
def test_engram_config_cli_optional(options, provided):
"""An explicit empty config must remain distinct from an omitted config."""
parser = EngineArgs.add_cli_args(FlexibleArgumentParser())
args = EngineArgs.from_cli_args(parser.parse_args(options))
assert (args.engram_config is not None) == provided
@pytest.mark.parametrize(
("type_hint", "type", "expected"),
[