Dynamic Speculation Lookahead Accelerates Speculative Decoding of Large Language Models
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866913573405458432 |
|---|---|
| author | Mamou, Jonathan Pereg, Oren Korat, Daniel Berchansky, Moshe Timor, Nadav Wasserblat, Moshe Schwartz, Roy |
| author_facet | Mamou, Jonathan Pereg, Oren Korat, Daniel Berchansky, Moshe Timor, Nadav Wasserblat, Moshe Schwartz, Roy |
| contents | Speculative decoding is commonly used for reducing the inference latency of large language models. Its effectiveness depends highly on the speculation lookahead (SL)-the number of tokens generated by the draft model at each iteration. In this work we show that the common practice of using the same SL for all iterations (static SL) is suboptimal. We introduce DISCO (DynamIc SpeCulation lookahead Optimization), a novel method for dynamically selecting the SL. Our experiments with four datasets show that DISCO reaches an average speedup of 10% compared to the best static SL baseline, while generating the exact same text. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_04304 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Dynamic Speculation Lookahead Accelerates Speculative Decoding of Large Language Models Mamou, Jonathan Pereg, Oren Korat, Daniel Berchansky, Moshe Timor, Nadav Wasserblat, Moshe Schwartz, Roy Computation and Language Speculative decoding is commonly used for reducing the inference latency of large language models. Its effectiveness depends highly on the speculation lookahead (SL)-the number of tokens generated by the draft model at each iteration. In this work we show that the common practice of using the same SL for all iterations (static SL) is suboptimal. We introduce DISCO (DynamIc SpeCulation lookahead Optimization), a novel method for dynamically selecting the SL. Our experiments with four datasets show that DISCO reaches an average speedup of 10% compared to the best static SL baseline, while generating the exact same text. |
| title | Dynamic Speculation Lookahead Accelerates Speculative Decoding of Large Language Models |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2405.04304 |