max_model_len: 131072
max_num_seqs: 256
tensor_parallel_size: 2
kvcache_storage_backend: "mooncake"
enable_output_caching: True