Prediction Heads

This section gives an overview of how different prediction heads can be used together with adapter modules and how pre-trained adapters can be distributed side-by-side with matching prediction heads in AdapterHub. We will take a look at the AdapterModel classes (e.g. BertAdapterModel) introduced by adapters, which provide flexible support for prediction heads, as well as models with static heads provided out-of-the-box by Hugging Face Transformers (e.g. BertForSequenceClassification).

Tip

We recommend to use the AdapterModel classes whenever possible. These flexible models have been created specifically for working with adapters.

AdapterModel classes

The AdapterModel classes provided by adapters allow a flexible configuration of prediction heads on top of a pre-trained language model.

First, we load pre-trained model from the Hugging Face Hub via the AutoAdapterModel class:

model = AutoAdapterModel.from_pretrained("bert-base-uncased")

By default, this model doesn’t have any heads yet, so let’s add a new binary sequence classification head on top of our model:

model.add_classification_head("mrpc", num_labels=2)

All heads have a name, we called this new head "mrpc". Since all heads are named, we can add multiple other heads with different names to the same model. To see the head types of a model and how they can get configured, please refer to the class references of the respective model classes, e.g. BertAdapterModel.

A head alone is just one layer with very few parameters. Hence, we want to train our classification head together with an adapter, so let’s add one:

model.add_adapter("mrpc", config="seq_bn")
model.set_active_adapters("mrpc")

Since we gave the task adapter the same name as our head, we can easily identify them as belonging together. The call to set_active_adapters() in the second line tells our model to use the adapter - head configuration we specified by default in a forward pass. At this point, we can start to train our setup.

Note

The set_active_adapters() will search for an adapter and a prediction head with the given name to be activated. Alternatively, prediction heads can also be activated explicitly (i.e. without adapter modules). These three options are possible (in order of priority when multiple are specified):

If head is passed to the forward call, the head with the given name is used.
If the forward call is executed within an AdapterSetup context, the head configuration is read from the context.
If the active_head property is set, the head configuration is read from there.

After training has completed, we can save our whole setup (adapter module and prediction head), with a single call:

model.save_adapter("/path/to/dir", "mrpc", with_head=True)

Now, you just have to share your work with the world. After you published the adapter together with its head in the Hub, anyone else can load both adapter and head by using the same model class.

Alternatively, we can also save and load the prediction head separately from an adapter module:

# save
model.save_head("/path/to/dir", "mrpc")
# load
model.load_head("/path/to/dir")

Lastly, it’s also possible to delete an added head again:

model.delete_head("mrpc")

Model classes with static heads (Hugging Face Transformers)

The transformers library provides strongly typed model classes with heads for various different tasks (e.g. RobertaForSequenceClassification, AutoModelForMultipleChoice …). If an adapter module is trained with one of these out-of-the-box classes, it is encouraged to also distribute the prediction head weights together with the adapter weights. Therefore, we can also easily save the prediction head weights for these models together with an adapter:

model.save_adapter("/path/to/dir", "mrpc", with_head=True)

In the next step, we can provide both the adapter weights and the head weights to the Hub. If someone else then downloads the pre-trained adapter, the resolving method will check if the prediction head matches the class of his model. In case the classes match, the prediction head weights will be automatically loaded too.

Automatic conversion

adapters supports loading static heads, e.g., created with AutoModelForSequenceClassification, into model classes with flexible heads, e.g. AutoAdapterModel.

For this, for a model created with AutoModelForSequenceClassification we first need to enable adapter support by calling the init() method.

from adapters import init, AutoAdapterModel
from transformers import AutoModelForSequenceClassification
import os

static_head_model = AutoModelForSequenceClassification.from_pretrained("bert-base-uncased")
# Enable adapter support
init(static_head_model) 

Now we can add an adapter and save it together with the head as usual:

static_head_model.add_adapter("test")

temp_dir = os.path.join(os.getcwd(), "temp_dir")
static_head_model.save_adapter(temp_dir, "test", with_head=True)

When now loading the adapter and head into a new AdapterModel, the conversion of weights happens automatically during the call of load_adapter(), so no additional steps are needed:

flex_head_model = AutoAdapterModel.from_pretrained("bert-base-uncased")
flex_head_model.load_adapter(temp_dir)

assert "test" in flex_head_model.adapters_config
assert "test" in flex_head_model.heads

Note

The conversion in the opposite direction is not supported, i.e. you cannot load a head created with AutoAdapterModel into a model of type AutoModelForSequenceClassification.

Custom Heads

If none of the available prediction heads fit your requirements, you can define and add a custom head.

First, we need to define the new head class. For that, the initialization and the forward pass need to be implemented. The initialization of the head gets a reference to the model, the name of the head, and additionally defined kwargs. You can use the following template as a guideline.

class CustomHead(PredictionHead):
    def __init__(
        self,
        model,
        head_name,
        **kwargs,
    ):
        # innitialization of the custom head

    def forward(self, outputs, cls_output=None, attention_mask=None, return_dict=False, **kwargs):
        # implementation of the forward pass

Next, we can register the new custom head and give the new head type a name. This only notifies the model that there is a new head type. Then, we can add an instance of the new head to the model by calling add_custom_head with the name of the new head type, the name of the head instance we are creating, and additional arguments required by the head.

model.register_custom_head("my_custom_head", CustomHead)
model.add_custom_head(head_type="my_custom_head", head_name="custom_head", **kwargs)

After adding the custom head you can treat it like any other build-in head type.