inference-cache-config/trn1/smollm3.json

{
 "HuggingFaceTB/SmolLM3-3B": [
   {
      "batch_size": 1,
      "sequence_length": 4096,
      "tensor_parallel_size": 2,
      "instance_type" : "trn1"
    },
   {
      "batch_size": 4,
      "sequence_length": 4096,
      "tensor_parallel_size": 2,
      "instance_type" : "trn1"
    },
    {
      "batch_size": 16,
      "sequence_length": 4096,
      "tensor_parallel_size": 2,
      "instance_type" : "trn1"
    }
  ]
}