chunked/content_aware_chunking/_multilingual/chunk_3.txt

The language embedding is a tensor filled with 0 since that is the language id for English. This tensor should be the same size as input_ids.