Merge branch 'v0.5.0-dtk24.04.1_wangmin' into 'v0.5.0-dtk24.04.1'

fix error in test_long_context.py::test_batched_rope_kernel

See merge request dcutoolkit/deeplearing/vllm!16

vllm/worker/model_runner.py

@@ -830,6 +830,7 @@ class ModelRunner:
                     multi_modal_data=dummy_multi_modal_data,
                 )
                 seqs.append(seq)
+                max_num_batched_tokens -= seq_len
                 
         for group_id in range(max_num_seqs):
             seq_len = (max_num_batched_tokens // max_num_seqs +