Published April 1, 2026
| Version v1
Software
Open
Profiling Code for Speculative Decoding: Performance or Illusion?
Authors/Creators
-
1.
University of California, Berkeley
Description
Profiling Code for paper "Speculative Decoding: Performance or Illusion?". Note that this only contains the branch for perf/e2e-v0.10.1.1. Please refer to our GitHub Repo (https://github.com/SpecDecode-Bench/vllm) for the remaining profiling code.
Files
vllm-profiling-v0.10.1.1.zip
Files
(12.2 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:fe561b0e2f8acdb1f3df0c77d2ff4386
|
12.2 MB | Preview Download |
Additional details
Software
- Repository URL
- https://github.com/SpecDecode-Bench/vllm