inference-cache-config/trn2/qwen3-moe.json

{
  "Qwen/Qwen3-30B-A3B-Instruct-2507": [
    {
      "batch_size": 1,
      "sequence_length": 4096,
      "tensor_parallel_size": 4,
      "instance_type" : "trn2"
    },
    {
      "batch_size": 8,
      "sequence_length": 4096,
      "tensor_parallel_size": 4,
      "instance_type" : "trn2"
    },
    {
      "batch_size": 16,
      "sequence_length": 4096,
      "tensor_parallel_size": 4,
      "instance_type" : "trn2"
    }
  ],
  "Qwen/Qwen3-235B-A22B-Instruct-2507": [
    {
      "batch_size": 1,
      "sequence_length": 4096,
      "tensor_parallel_size": 64,
      "instance_type" : "trn2"
    },
    {
      "batch_size": 4,
      "sequence_length": 4096,
      "tensor_parallel_size": 64,
      "instance_type" : "trn2"
    }
  ]
}