[Speculative Decoding] EAGLE Implementation with Top-1 proposer (#6830)
This commit is contained in:
@@ -1,5 +1,6 @@
|
||||
from vllm.transformers_utils.configs.chatglm import ChatGLMConfig
|
||||
from vllm.transformers_utils.configs.dbrx import DbrxConfig
|
||||
from vllm.transformers_utils.configs.eagle import EAGLEConfig
|
||||
# RWConfig is for the original tiiuae/falcon-40b(-instruct) and
|
||||
# tiiuae/falcon-7b(-instruct) models. Newer Falcon models will use the
|
||||
# `FalconConfig` class from the official HuggingFace transformers library.
|
||||
@@ -20,6 +21,7 @@ __all__ = [
|
||||
"InternVLChatConfig",
|
||||
"JAISConfig",
|
||||
"MedusaConfig",
|
||||
"EAGLEConfig",
|
||||
"MLPSpeculatorConfig",
|
||||
"NemotronConfig",
|
||||
"UltravoxConfig",
|
||||
|
||||
Reference in New Issue
Block a user