Fix Training Error with AdapterDrop and Prefix Tuning (#673)

Fixes #669 

Changes in this PR:
- Avoid throwing `RuntimeError` due to dimension mismatch occuring when
passing the positional encoding from layers dropped by AdapterDrop to
layers modified by prefix tuning.

src/adapters/models/t5/modeling_t5.py

@@ -77,7 +77,7 @@ def forward(
         if past_key_value is not None:
             assert (
                 len(past_key_value) == 2
-            ), f"past_key_value should have 2 past states: keys and values. Got { len(past_key_value)} past states"
+            ), f"past_key_value should have 2 past states: keys and values. Got {len(past_key_value)} past states"
             real_seq_length += past_key_value[0].shape[2] if query_length is None else query_length
 
         key_length = real_seq_length if key_value_states is None else key_value_states.shape[1]
@@ -144,7 +144,13 @@ def project(hidden_states, proj_layer, key_value_states, past_key_value):
             query_states, key_states.transpose(3, 2)
         )  # equivalent of torch.einsum("bnqd,bnkd->bnqk", query_states, key_states), compatible with onnx op>9
 
-        if position_bias is None:
+        # For Prefix Tuning, when training with AdapterDrop, we must additionally check that the sequence lengths of
+        # both positional encoding and the scores account for the prefix tokens.
+        # This is because the positional encoding is calculated only once in the beginning and then used for all layers.
+        # However, if the encoding was calculated without the prefix tokens due to AdapterDrop having dropped an
+        # adapter layer in the beginning, the positional encoding will be shorter than the scores, resulting in a
+        # dimension mismatch when adding the positional encoding to the scores.
+        if position_bias is None or position_bias.shape[3] != scores.shape[3]:
             if not self.has_relative_attention_bias:
                 position_bias = torch.zeros(
                     (1, self.n_heads, real_seq_length, key_length), device=scores.device, dtype=scores.dtype