model_config_spatial_0.yaml

num_gpus: 1
max_num_seqs: 256
overload_threshold: 3
gpu_memory_utilization: 0.7

models:
  -
    name: llm-0
    model: /mnt/afs/share/LLMCKPTs/huggyllama/llama-7b
    tensor_parallel_size: 1
    pipeline_parallel_size: 1
    placement:
      - [0]
    mps_percentage: [100, 40]
    model_dtype: fp16
workloads:
  # workload generation refer to README
  workload_file: