Skip to content

[https://nvbugs/6337235][test] Fix MX/GMS model loader fixtures #15471

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
chienchunhung merged 1 commit into from
Jun 22, 2026
+40 −1
Merged

[https://nvbugs/6337235][test] Fix MX/GMS model loader fixtures #15471

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 22 additions & 1 deletion tests/unittest/_torch/pyexecutor/test_model_loader_gms.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,6 +15,16 @@
from tensorrt_llm._torch.pyexecutor.model_loader import ModelLoader
from tensorrt_llm.llmapi.llm_args import LoadFormat

_SOURCE_IDENTITY = model_loader_mod.SourceIdentity(
format_version=1,
model_fingerprint="model",
quant_fingerprint="quant",
backend_fingerprint="backend",
parallel_fingerprint="parallel",
rank=0,
shard_fingerprint="shard",
)


class _TinyModel(nn.Module):
def __init__(self, events, *, include_draft=False):
Expand Down Expand Up @@ -71,6 +81,13 @@ def _make_loader(monkeypatch, *, events, spec_config=None):
monkeypatch.setattr(model_loader_mod, "timing", lambda *_args, **_kwargs: nullcontext())
monkeypatch.setattr(model_loader_mod, "maybe_create_moe_load_balancer", _moe_context)
monkeypatch.setattr(model_loader_mod, "MetaInitMode", lambda: nullcontext())
# These tests stub ModelConfig, while SourceIdentity has dedicated
# coverage. Keep this file focused on ModelLoader GMS branch behavior.
monkeypatch.setattr(
model_loader_mod.SourceIdentity,
"from_model_config",
classmethod(lambda cls, *_args, **_kwargs: _SOURCE_IDENTITY),
)
monkeypatch.setattr(
model_loader_mod.AutoModelForCausalLM,
"from_config",
Expand All @@ -93,6 +110,7 @@ def _build_gms_backend(*, is_rw, events):
if is_rw:
backend.mem_pool_scope.side_effect = lambda _device: _pool_scope(events)
else:
backend.get_source_identity.return_value = _SOURCE_IDENTITY

def _materialize(_model):
events.append("materialize")
Expand Down Expand Up @@ -169,7 +187,10 @@ def test_gms_load_branch(monkeypatch, is_rw, expected_events):
# path (see model_loader.py); HF ignores it, MX uses it for direct
# P2P writes when MX+GMS composition eventually lands.
checkpoint_loader.load_weights.assert_called_once_with(
"/ckpt", mapping=loader.mapping, model=model
"/ckpt",
mapping=loader.mapping,
model=model,
source_identity=loader._source_identity,
)
loader._call_load_weights.assert_called_once()
backend.move_untracked_params.assert_called_once_with(model)
Expand Down
18 changes: 18 additions & 0 deletions tests/unittest/_torch/pyexecutor/test_model_loader_mx.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,16 @@
from tensorrt_llm._torch.pyexecutor.model_loader import ModelLoader
from tensorrt_llm.llmapi.llm_args import LoadFormat

_SOURCE_IDENTITY = model_loader_mod.SourceIdentity(
format_version=1,
model_fingerprint="model",
quant_fingerprint="quant",
backend_fingerprint="backend",
parallel_fingerprint="parallel",
rank=0,
shard_fingerprint="shard",
)


class _LinearStub(nn.Module):
def post_load_weights(self):
Expand Down Expand Up @@ -77,6 +87,13 @@ def _make_loader(monkeypatch, *, events, spec_config=None):
monkeypatch.setattr(model_loader_mod, "timing", lambda *_args, **_kwargs: nullcontext())
monkeypatch.setattr(model_loader_mod, "maybe_create_moe_load_balancer", _moe_context)
monkeypatch.setattr(model_loader_mod, "MetaInitMode", lambda: nullcontext())
# These tests stub ModelConfig, while SourceIdentity has dedicated
# coverage. Keep this file focused on ModelLoader MX branch behavior.
monkeypatch.setattr(
model_loader_mod.SourceIdentity,
"from_model_config",
classmethod(lambda cls, *_args, **_kwargs: _SOURCE_IDENTITY),
)
monkeypatch.setattr(
model_loader_mod.AutoModelForCausalLM,
"from_config",
Expand Down Expand Up @@ -105,6 +122,7 @@ def test_mx_success_initializes_mapper_skips_weight_mapping_and_reload_works(mon
_args, kwargs = checkpoint_loader.load_weights.call_args
assert kwargs["mapping"] is loader.mapping
assert kwargs["model"] is model
assert kwargs["source_identity"] is loader._source_identity
assert loader._call_load_weights.call_count == 0
checkpoint_loader.get_initialized_weight_mapper.assert_called_once()
assert loader.weight_mapper is checkpoint_loader.get_initialized_weight_mapper.return_value
Expand Down
Loading