I didn't know this! I've always thought speculative decoding was "if p(draft_token) > threshold, use it". You made me go read how it actually works and it's pretty neat!
That said, I still think some providers are cheating. Please correct me if the test below is flawed.
I generated texts at temperature = 0 vs temperature = 2. At high temperature, the distributions effectively become flatter, meaning the difference between real and draft effective distributions (the D_LK used in theorem 3.5 of 2211.17192) becomes smaller. When T=2, the model speaks complete gibberish, so the effective distribution must be pretty flat. This should mean fewer rejections --> a lot faster speculative decoding. Yet, I see no increase in throughput at all...
That said, I still think some providers are cheating. Please correct me if the test below is flawed.
I generated texts at temperature = 0 vs temperature = 2. At high temperature, the distributions effectively become flatter, meaning the difference between real and draft effective distributions (the D_LK used in theorem 3.5 of 2211.17192) becomes smaller. When T=2, the model speaks complete gibberish, so the effective distribution must be pretty flat. This should mean fewer rejections --> a lot faster speculative decoding. Yet, I see no increase in throughput at all...