{
  "instance_0": {
    "model_name": "Qwen/Qwen2.5-72B",
    "gpu_type": "A100_40GB",
    "gpu_count": 4,
    "tensor_parallel": 4,
    "inference_backend": "vllm",
    "inference_backend_version": "v1_dev11708",
    "engine_version": "v1",
    "max_model_len": 8192,
    "gpu_memory_utilization": 0.95,
    "enforce_eager": true,
    "attention_backend": "FLASH_ATTN"
  },
  "instance_1": {
    "model_name": "Qwen/Qwen2.5-72B",
    "gpu_type": "A100_40GB",
    "gpu_count": 4,
    "tensor_parallel": 4,
    "inference_backend": "vllm",
    "inference_backend_version": "v1_dev11708",
    "engine_version": "v1",
    "max_model_len": 8192,
    "gpu_memory_utilization": 0.95,
    "enforce_eager": true,
    "attention_backend": "FLASH_ATTN"
  },
  "instance_2": {
    "model_name": "Qwen/Qwen2.5-14B",
    "gpu_type": "V100_16GB",
    "gpu_count": 4,
    "tensor_parallel": 4,
    "inference_backend": "vllm",
    "inference_backend_version": "v1_dev11708",
    "engine_version": "v1",
    "max_model_len": 8192,
    "gpu_memory_utilization": 0.85,
    "enforce_eager": false,
    "attention_backend": "FLASH_ATTN"
  },
  "instance_3": {
    "model_name": "Qwen/Qwen2.5-14B",
    "gpu_type": "V100_16GB",
    "gpu_count": 4,
    "tensor_parallel": 4,
    "inference_backend": "vllm",
    "inference_backend_version": "v1_dev11708",
    "engine_version": "v1",
    "max_model_len": 8192,
    "gpu_memory_utilization": 0.85,
    "enforce_eager": false,
    "attention_backend": "FLASH_ATTN"
  },
  "instance_4": {
    "model_name": "Qwen/Qwen2.5-14B",
    "gpu_type": "V100_16GB",
    "gpu_count": 4,
    "tensor_parallel": 4,
    "inference_backend": "vllm",
    "inference_backend_version": "v1_dev11708",
    "engine_version": "v1",
    "max_model_len": 8192,
    "gpu_memory_utilization": 0.85,
    "enforce_eager": false,
    "attention_backend": "FLASH_ATTN"
  },
  "instance_5": {
    "model_name": "Qwen/Qwen2.5-7B",
    "gpu_type": "A30_24GB",
    "gpu_count": 1,
    "tensor_parallel": 1,
    "inference_backend": "vllm",
    "inference_backend_version": "v1_dev11708",
    "engine_version": "v1",
    "max_model_len": 8192,
    "gpu_memory_utilization": 0.9,
    "enforce_eager": false,
    "attention_backend": "FLASH_ATTN"
  },
  "instance_6": {
    "model_name": "Qwen/Qwen2.5-7B",
    "gpu_type": "A30_24GB",
    "gpu_count": 1,
    "tensor_parallel": 1,
    "inference_backend": "vllm",
    "inference_backend_version": "v1_dev11708",
    "engine_version": "v1",
    "max_model_len": 8192,
    "gpu_memory_utilization": 0.9,
    "enforce_eager": false,
    "attention_backend": "FLASH_ATTN"
  },
  "instance_7": {
    "model_name": "Qwen/Qwen2.5-7B",
    "gpu_type": "A30_24GB",
    "gpu_count": 1,
    "tensor_parallel": 1,
    "inference_backend": "vllm",
    "inference_backend_version": "v1_dev11708",
    "engine_version": "v1",
    "max_model_len": 8192,
    "gpu_memory_utilization": 0.9,
    "enforce_eager": false,
    "attention_backend": "FLASH_ATTN"
  },
  "instance_8": {
    "model_name": "Qwen/Qwen2.5-7B",
    "gpu_type": "A30_24GB",
    "gpu_count": 1,
    "tensor_parallel": 1,
    "inference_backend": "vllm",
    "inference_backend_version": "v1_dev11708",
    "engine_version": "v1",
    "max_model_len": 8192,
    "gpu_memory_utilization": 0.9,
    "enforce_eager": false,
    "attention_backend": "FLASH_ATTN"
  },
  "instance_9": {
    "model_name": "Qwen/Qwen2.5-7B",
    "gpu_type": "A30_24GB",
    "gpu_count": 1,
    "tensor_parallel": 1,
    "inference_backend": "vllm",
    "inference_backend_version": "v1_dev11708",
    "engine_version": "v1",
    "max_model_len": 8192,
    "gpu_memory_utilization": 0.9,
    "enforce_eager": false,
    "attention_backend": "FLASH_ATTN"
  },
  "instance_10": {
    "model_name": "Qwen/Qwen2.5-3B",
    "gpu_type": "A30_24GB",
    "gpu_count": 1,
    "tensor_parallel": 1,
    "inference_backend": "vllm",
    "inference_backend_version": "v1_dev11708",
    "engine_version": "v1",
    "max_model_len": 8192,
    "gpu_memory_utilization": 0.9,
    "enforce_eager": false,
    "attention_backend": "FLASH_ATTN"
  },
  "instance_11": {
    "model_name": "Qwen/Qwen2.5-3B",
    "gpu_type": "A30_24GB",
    "gpu_count": 1,
    "tensor_parallel": 1,
    "inference_backend": "vllm",
    "inference_backend_version": "v1_dev11708",
    "engine_version": "v1",
    "max_model_len": 8192,
    "gpu_memory_utilization": 0.9,
    "enforce_eager": false,
    "attention_backend": "FLASH_ATTN"
  },
  "instance_12": {
    "model_name": "Qwen/Qwen2.5-3B",
    "gpu_type": "A30_24GB",
    "gpu_count": 1,
    "tensor_parallel": 1,
    "inference_backend": "vllm",
    "inference_backend_version": "v1_dev11708",
    "engine_version": "v1",
    "max_model_len": 8192,
    "gpu_memory_utilization": 0.9,
    "enforce_eager": false,
    "attention_backend": "FLASH_ATTN"
  },
  "instance_13": {
    "model_name": "Qwen/Qwen2.5-3B",
    "gpu_type": "P100_12GB",
    "gpu_count": 1,
    "tensor_parallel": 1,
    "inference_backend": "vllm",
    "inference_backend_version": "v0_dev5353",
    "engine_version": "v0",
    "max_model_len": 4096,
    "gpu_memory_utilization": 0.9,
    "enforce_eager": false,
    "attention_backend": "XFORMERS"
  },
  "instance_14": {
    "model_name": "Qwen/Qwen2.5-3B",
    "gpu_type": "P100_12GB",
    "gpu_count": 1,
    "tensor_parallel": 1,
    "inference_backend": "vllm",
    "inference_backend_version": "v0_dev5353",
    "engine_version": "v0",
    "max_model_len": 4096,
    "gpu_memory_utilization": 0.9,
    "enforce_eager": false,
    "attention_backend": "XFORMERS"
  },
  "instance_15": {
    "model_name": "Qwen/Qwen2.5-3B",
    "gpu_type": "P100_12GB",
    "gpu_count": 1,
    "tensor_parallel": 1,
    "inference_backend": "vllm",
    "inference_backend_version": "v0_dev5353",
    "engine_version": "v0",
    "max_model_len": 4096,
    "gpu_memory_utilization": 0.9,
    "enforce_eager": false,
    "attention_backend": "XFORMERS"
  },
  "instance_16": {
    "model_name": "Qwen/Qwen2.5-3B",
    "gpu_type": "P100_12GB",
    "gpu_count": 1,
    "tensor_parallel": 1,
    "inference_backend": "vllm",
    "inference_backend_version": "v0_dev5353",
    "engine_version": "v0",
    "max_model_len": 4096,
    "gpu_memory_utilization": 0.9,
    "enforce_eager": false,
    "attention_backend": "XFORMERS"
  },
  "instance_17": {
    "model_name": "Qwen/Qwen2.5-3B",
    "gpu_type": "P100_12GB",
    "gpu_count": 1,
    "tensor_parallel": 1,
    "inference_backend": "vllm",
    "inference_backend_version": "v0_dev5353",
    "engine_version": "v0",
    "max_model_len": 4096,
    "gpu_memory_utilization": 0.9,
    "enforce_eager": false,
    "attention_backend": "XFORMERS"
  }
}