Fix Transformers backend FP8 MoE and remove some boilerplate (#46820)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
@@ -134,7 +134,6 @@ class GlmOcrMTP(nn.Module, SupportsPP):
|
||||
vllm_config=vllm_config, prefix=maybe_prefix(prefix, "model")
|
||||
)
|
||||
|
||||
self.expert_weights = []
|
||||
self.num_layers = self.config.num_nextn_predict_layers
|
||||
for layer in self.model.layers.values():
|
||||
assert isinstance(layer, GlmOcrMultiTokenPredictorLayer)
|
||||
|
||||
Reference in New Issue
Block a user