diff --git a/python/sglang/srt/batch_overlap/single_batch_overlap.py b/python/sglang/srt/batch_overlap/single_batch_overlap.py index 815ba8715..42cdec660 100644 --- a/python/sglang/srt/batch_overlap/single_batch_overlap.py +++ b/python/sglang/srt/batch_overlap/single_batch_overlap.py @@ -87,7 +87,7 @@ def compute_overlap_args(dispatch_output, alt_stream): hidden_states = dispatch_output.hidden_states - num_local_experts, num_tokens_static, hidden_dim = hidden_states.shape + num_local_experts, num_tokens_static = hidden_states.shape total_num_sms = torch.cuda.get_device_properties( device="cuda"