biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Core] Deprecating block manager v1 and make block manager v2 default (#8704)

Browse Source 
Removing the block manager v1. This is the initial piece of prefix-caching-centric design. In order to achieve prefix-caching-centric design, we need to simplify the code path so that we only use v2 block manager (which has much higher performance on prefix caching).
This commit is contained in:
Kuntai Du
2024-10-17 11:38:15 -05:00
committed by GitHub
parent 5eda21e773
commit 81ede99ca4
45 changed files with 206 additions and 2109 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
tests/spec_decode/e2e/test_seed.py
View File

@@ -17,9 +17,6 @@ SPEC_MODEL = "JackFram/llama-160m"
# Skip cuda graph recording for fast test.
"enforce_eager": True,
# Required for spec decode.
"use_v2_block_manager": True,
# speculative model
"speculative_model": "JackFram/llama-160m",