Inline the trivial _build_model_config wrapper (#25448)
This commit is contained in:
@@ -419,7 +419,7 @@ class ModelRunner(ModelRunnerKVCacheMixin):
|
||||
self.dflash_draft_num_layers = None
|
||||
if self.spec_algorithm.is_eagle3() and not self.is_draft_worker:
|
||||
# load draft config
|
||||
draft_model_config = self._build_model_config(
|
||||
draft_model_config = ModelConfig.from_server_args(
|
||||
server_args,
|
||||
model_path=(server_args.speculative_draft_model_path),
|
||||
model_revision=server_args.speculative_draft_model_revision,
|
||||
@@ -448,7 +448,7 @@ class ModelRunner(ModelRunnerKVCacheMixin):
|
||||
)
|
||||
|
||||
# Select target layers to capture for building DFlash context features.
|
||||
draft_model_config = self._build_model_config(
|
||||
draft_model_config = ModelConfig.from_server_args(
|
||||
server_args,
|
||||
model_path=(server_args.speculative_draft_model_path),
|
||||
model_revision=server_args.speculative_draft_model_revision,
|
||||
@@ -568,16 +568,6 @@ class ModelRunner(ModelRunnerKVCacheMixin):
|
||||
self._model_update_group = {}
|
||||
self._weights_send_group = {}
|
||||
|
||||
def _build_model_config(
|
||||
self, server_args, model_path=None, model_revision=None, is_draft_model=False
|
||||
):
|
||||
return ModelConfig.from_server_args(
|
||||
server_args,
|
||||
model_path=model_path,
|
||||
model_revision=model_revision,
|
||||
is_draft_model=is_draft_model,
|
||||
)
|
||||
|
||||
def init_msprobe(self):
|
||||
# Init the msprobe
|
||||
try:
|
||||
|
||||
Reference in New Issue
Block a user