Bug fix for modeling utilities function: apply_chunking_to_forward, c…

…hunking should be in the chunking dimension, an exception was raised if the complete shape of the inputs was not the same rather than only the chunking dimension (huggingface#8391)

Co-authored-by: pedro <pe25171@mit.edu>

src/transformers/modeling_utils.py

@@ -1668,9 +1668,9 @@ def forward(self, hidden_states):
     """
 
     assert len(input_tensors) > 0, "{} has to be a tuple/list of tensors".format(input_tensors)
-    tensor_shape = input_tensors[0].shape
+    tensor_shape = input_tensors[0].shape[chunk_dim]
     assert all(
-        input_tensor.shape == tensor_shape for input_tensor in input_tensors
+        input_tensor.shape[chunk_dim] == tensor_shape for input_tensor in input_tensors
     ), "All input tenors have to be of the same shape"
 
     # inspect.signature exist since python 3.5 and is a python method -> no problem with backward compatibility