nemo_automodel.components.models.common.bidirectional#

Module Contents#

Classes#

EncoderStateDictAdapter

Adapter for encoder model state dicts.

Data#

__all__

API#

class nemo_automodel.components.models.common.bidirectional.EncoderStateDictAdapter#

Bases: nemo_automodel.components.checkpoint.state_dict_adapter.StateDictAdapter

Adapter for encoder model state dicts.

Internal format uses a model. prefix on all keys. HF format does not. This adapter strips or adds the model. prefix as needed, including for PEFT-wrapped keys (base_model.model.model.X <-> base_model.model.X).

Initialization

_PEFT_PREFIX#

‘base_model.model.’

_MODEL_PREFIX#

‘model.’

_PEFT_MODEL_PREFIX#

None

_strip_model_prefix(key)#
_add_model_prefix(key)#
to_hf(state_dict, **kwargs)#
from_hf(hf_state_dict, device_mesh=None, **kwargs)#
convert_single_tensor_to_hf(fqn, tensor, **kwargs)#
nemo_automodel.components.models.common.bidirectional.__all__#

[‘EncoderStateDictAdapter’]