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

[SpecDec][Misc] Cleanup, remove bonus token logic. (#8701)

Browse Source
This commit is contained in:
Lily Liu
2024-09-22 12:34:14 -07:00
committed by GitHub
parent 5b59532760
commit c6bd70d772
7 changed files with 33 additions and 115 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
tests/spec_decode/e2e/test_medusa_correctness.py
View File

@@ -31,7 +31,7 @@ MAIN_MODEL = "JackFram/llama-68m"
# speculative model
SPEC_MODEL = "abhigoyal/vllm-medusa-llama-68m-random"
# max. number of speculative tokens: this corresponds to
# max number of speculative tokens: this corresponds to
# num_heads in the config.json of the speculator model.
MAX_SPEC_TOKENS = 5