()
| 58 | BART_decoder.set_return_device("cuda" if use_cuda else "cpu") |
| 59 | |
| 60 | def decoder_stmt(): |
| 61 | BART_decoder( |
| 62 | input_ids=input_ids, encoder_hidden_states=encoder_last_hidden_state, use_cache=use_cache, |
| 63 | past_key_values=past_key_values |
| 64 | ) |
| 65 | |
| 66 | decoder_e2e_time = measure_python_inference_code(decoder_stmt, timing_profile) |
| 67 |