Add support for LoRA adapters in Nemotron-H models (#30802)
Signed-off-by: Daniel Serebrenik <daserebrenik@nvidia.com>
This commit is contained in:
@@ -376,6 +376,7 @@ class SupportsLoRA(Protocol):
|
||||
MRO of your model class.
|
||||
"""
|
||||
is_3d_moe_weight: ClassVar[bool] = False
|
||||
is_non_gated_moe: ClassVar[bool] = False
|
||||
# The `embedding_module` and `embedding_padding_modules`
|
||||
# are empty by default.
|
||||
embedding_modules: ClassVar[dict[str, str]] = {}
|
||||
|
||||
Reference in New Issue
Block a user