chunked/nltk_chunking/_add_new_model/chunk_258.txt

When you're confident that both implementations yield the same output, verify the outputs with
torch.allclose(original_output, output, atol=1e-3), you're done with the most difficult part!