home | feeds |donate  
dspark: speculative decoding accelerates llm inference [pdf]
- Jun 27