inference-cache-config/mixtral.json

{
  "mistralai/Mixtral-8x7B-Instruct-v0.1": [
    {
      "batch_size": 1,
      "sequence_length": 4096,
      "num_cores": 24,
      "auto_cast_type": "bf16"
    },
    {
      "batch_size": 4,
      "sequence_length": 4096,
      "num_cores": 24,
      "auto_cast_type": "bf16"
    }
  ],
  "mistralai/Mixtral-8x22B-Instruct-v0.1": [
    {
      "batch_size": 1,
      "sequence_length": 4096,
      "num_cores": 24,
      "auto_cast_type": "bf16"
    },
    {
      "batch_size": 4,
      "sequence_length": 4096,
      "num_cores": 24,
      "auto_cast_type": "bf16"
    }
  ]
}