[Speculative decoding 7/9] Speculative decoding end-to-end correctness tests. (#3951)

This commit is contained in:
Cade Daniel
2024-04-23 08:02:36 +00:00
committed by GitHub
parent 050f285ff6
commit 62b8aebc6f
22 changed files with 1164 additions and 175 deletions
+1
View File
@@ -83,6 +83,7 @@ class GPUExecutor(ExecutorBase):
scheduler_config=self.scheduler_config,
device_config=self.device_config,
cache_config=self.cache_config,
# TODO allow draft-model specific load config.
load_config=self.load_config,
local_rank=0,
rank=0,