def create_bindings(self):
return {
"mergeable_model":
diag_execs.diagonal_mergeable_model_from_checkpoint,
"model_merger":
self.model_merger,
#
"checkpoint_to_fisher_matrix_uuid":
self.get_checkpoint_to_fisher_matrix_uuid(),
"weightings":
create_pairwise_weightings(self.num_weightings),
#
"checkpoints":
[m.model_checkpoint_uuid for m in self.models_to_merge],
"checkpoint_tasks": [m.task for m in self.models_to_merge],
"additional_model_bindings":
[m.additional_model_bindings for m in self.models_to_merge],
#
"tasks": [m.task for m in self.models_to_merge],
#
"pretrained_model":
self.pretrained_model,
#
"num_examples":
self.validation_examples,
"sequence_length":
self.sequence_length,
"batch_size":
self.batch_size,
#
"normalize_fishers":
self.normalize_fishers,
}
def create_bindings(self):
weightings = create_pairwise_weightings(self.num_weightings,
self.min_target_weighting)
return {
"mergeable_model":
diag_execs.diagonal_mergeable_model_from_checkpoint_or_pretrained,
"model_merger":
diag_execs.diagonal_model_merger,
#
"checkpoint_to_fisher_matrix_uuid":
self.get_checkpoint_to_fisher_matrix_uuid(),
"weightings":
weightings,
#
"checkpoints":
[m.model_checkpoint_uuid for m in self.models_to_merge],
"checkpoint_tasks": [m.task for m in self.models_to_merge],
#
"tasks": [m.task for m in self.models_to_merge],
#
"pretrained_model":
self.pretrained_model,
#
"num_examples":
self.validation_examples,
"image_size":
self.image_size,
"batch_size":
self.batch_size,
#
"normalize_fishers":
self.normalize_fishers,
}
def create_bindings(self):
return {
"mergeable_model":
diag_execs.diagonal_mergeable_model_from_checkpoint_or_pretrained,
"model_merger":
self.model_merger,
#
"checkpoint_to_fisher_matrix_uuid":
self.get_checkpoint_to_fisher_matrix_uuid(),
"weightings":
create_pairwise_weightings(self.num_weightings),
#
"checkpoints":
[m.model_checkpoint_uuid for m in self.models_to_merge],
"checkpoint_tasks": [m.task for m in self.models_to_merge],
#
"tasks": [m.task for m in self.models_to_merge],
#
"pretrained_model":
self.pretrained_model,
#
"num_examples":
self.validation_examples,
"sequence_length":
self.sequence_length,
"batch_size":
self.batch_size,
#
"normalize_fishers":
self.normalize_fishers,
#
#
"hf_back_compat":
False,
"glue_label_map_overrides":
defs.LABEL_MAP_OVERRIDES,
}
def create_bindings(self):
return {
"mergeable_model":
diag_execs.diagonal_mergeable_model_from_checkpoint_or_pretrained,
"model_merger":
diag_execs.diagonal_model_merger,
#
"checkpoint_to_fisher_matrix_uuid":
self.get_checkpoint_to_fisher_matrix_uuid(),
"weightings":
create_pairwise_weightings(self.num_weightings),
#
"normalize_fishers":
self.normalize_fishers,
#
"checkpoints":
[m.model_checkpoint_uuid for m in self.models_to_merge],
"checkpoint_tasks": [m.task for m in self.models_to_merge],
#
"task":
self.models_to_merge[0].task,
"tasks": [m.task for m in self.models_to_merge],
#
# NOTE: model_checkpoint_uuid is actually the name of the pretrained_model.
# Note that this won't be the pretrained model for all models, but let's
# hope that it isn't really used.
"pretrained_model":
self.models_to_merge[0].model_checkpoint_uuid,
#
"num_examples":
self.validation_examples,
"sequence_length":
self.sequence_length,
"batch_size":
self.batch_size,
}