A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 16, 22:00 UTC
[Bugfix] Fix `compute_logits` in Jamba (#6093)
Roger Wang committed
Jul 3, 2024 at 00:32 UTC
7cd2ebb0251fd1fd0eec5c93dac674603a22eddd
1 file changed
+1
-1
vllm/model_executor/models/jamba.py
+1
-1
@@ -876,7 +876,7 @@ class JambaForCausalLM(nn.Module):
876
877
def compute_logits(self, hidden_states: torch.Tensor,
878
sampling_metadata: SamplingMetadata) -> torch.Tensor:
879
- logits = self.logits_processor(self.lm_head.weight, hidden_states,
879
+ logits = self.logits_processor(self.lm_head, hidden_states,
880
sampling_metadata)
881
return logits
882