{
  "models": [
    {
      "model_label": "Llama-3.1-8B",
      "model_id": "meta-llama/Llama-3.1-8B-Instruct",
      "gpu_model": "H100",
      "task": "lm-arena-chat",
      "gpus_per_replica": 1,
      "tensor_parallel": 1,
      "itl_deadline_s": 0.08,
      "batch_sizes": [8, 16, 32, 64, 96, 128, 192, 256, 384, 512, 768, 1024],
      "feasible_batch_sizes": [8, 16, 32, 64, 128, 256, 512],
      "num_replicas": 720,
      "initial_batch_size": 128
    },
    {
      "model_label": "Llama-3.1-70B",
      "model_id": "meta-llama/Llama-3.1-70B-Instruct",
      "gpu_model": "H100",
      "task": "lm-arena-chat",
      "gpus_per_replica": 4,
      "tensor_parallel": 4,
      "itl_deadline_s": 0.10,
      "batch_sizes": [8, 16, 32, 64, 96, 128, 192, 256, 384, 512, 768, 1024, 1536, 2048],
      "feasible_batch_sizes": [8, 16, 32, 64, 128, 256, 512],
      "num_replicas": 180,
      "initial_batch_size": 128
    },
    {
      "model_label": "Llama-3.1-405B",
      "model_id": "meta-llama/Llama-3.1-405B-Instruct-FP8",
      "gpu_model": "H100",
      "task": "lm-arena-chat",
      "gpus_per_replica": 8,
      "tensor_parallel": 8,
      "precision": "fp8",
      "itl_deadline_s": 0.12,
      "batch_sizes": [8, 16, 32, 64, 96, 128, 192, 256, 384, 512],
      "feasible_batch_sizes": [8, 16, 32, 64, 128, 256, 512],
      "num_replicas": 90,
      "initial_batch_size": 128
    },
    {
      "model_label": "Qwen3-30B-A3B",
      "model_id": "Qwen/Qwen3-30B-A3B-Thinking-2507",
      "gpu_model": "H100",
      "task": "gpqa",
      "gpus_per_replica": 2,
      "tensor_parallel": 2,
      "itl_deadline_s": 0.06,
      "batch_sizes": [8, 16, 32, 64, 96, 128, 192, 256, 384, 512],
      "feasible_batch_sizes": [8, 16, 32, 64, 128, 256, 512],
      "num_replicas": 480,
      "initial_batch_size": 128
    },
    {
      "model_label": "Qwen3-235B-A22B",
      "model_id": "Qwen/Qwen3-235B-A22B-Thinking-2507",
      "gpu_model": "H100",
      "task": "gpqa",
      "gpus_per_replica": 8,
      "tensor_parallel": 8,
      "itl_deadline_s": 0.14,
      "batch_sizes": [8, 16, 32, 64, 96, 128, 192, 256, 384, 512],
      "feasible_batch_sizes": [8, 16, 32, 64, 128, 256, 512],
      "num_replicas": 210,
      "initial_batch_size": 128
    }
  ],
  "training_trace_params": {},
  "data_dir": null,
  "ieee_case_dir": "examples/ieee13",
  "mlenergy_data_dir": null
}