default.yaml

# SuperBench Config
superbench:
  enable: null
  benchmarks:
    matmul:
      enable: true
      modes:
        - name: local
          proc_num: 8
          prefix: CUDA_VISIBLE_DEVICES={proc_rank}
          parallel: no
      frameworks:
        - pytorch
      parameters:
        num_steps: 2048
    bert_models:
      enable: true
      modes:
        - name: torch.distributed
          proc_num: 8
          node_num: all
      frameworks:
        - pytorch
      models:
        - bert-base
        - bert-large
      parameters:
        duration: 0
        num_warmup: 64
        num_steps: 2048
        batch_size: 16
        precision:
          - float32
          - float16
        model_action:
          - train
          - inference