Speeding up LLM Inference with parallel decoding

1 points| pgspaintbrush | 2 years ago |twitter.com