configs/rwkv/170M.yml

{
  # Parallelism is not yet supported for rwkv
  "pipe_parallel_size": 1,
  "model_parallel_size": 1,

  "num_layers": 12,
  "hidden_size": 768,
  "num_attention_heads": 12, # head_size = dim_att / num_attention_heads.
                             # head_size is 64 for all rwkv models
  "seq_length": 512,
  "max_position_embeddings": 2048,
  "output_layer_parallelism": "column",
  "norm": "rmsnorm",
  "rms_norm_epsilon": 1.0e-5,
  "train_micro_batch_size_per_gpu": 32,

  "attention_config": [[["rwkv"], 12]],

  "activation": "silu",

  # model settings

  #"pos_emb": "rotary",
  "rotary_pct": 0.25,
  "no_weight_tying": true,
  "gpt_j_residual": true,

  # these should provide some speedup but takes a while to build, set to true if desired
  "scaled_upper_triang_masked_softmax_fusion": false,
  "bias_gelu_fusion": false,
  "rope_fusion": false,
  "layernorm_fusion": false,


  # init methods
  "init_method": "small_init",
  "output_layer_init_method": "wang_init",

  # optimizer settings
  "optimizer": {
    "type": "Adam",
    "params": {
      "lr": 0.0008,
      "betas": [0.9, 0.95],
      "eps": 1.0e-8,
    }
  },
  "min_lr": 0.00008,

  # for all zero_optimization options, see https://www.deepspeed.ai/docs/config-json/#zero-optimizations-for-fp16-training
  "zero_optimization": {
    "stage": 1,
    "allgather_partitions": True,
    "allgather_bucket_size": 500000000,
    "overlap_comm": True,
    "reduce_scatter": True,
    "reduce_bucket_size": 500000000,
    "contiguous_gradients": True,
  },

  # batch / data settings
  "data_impl": "mmap",
  "num_workers": 1,

  # activation checkpointing
  "checkpoint_activations": true,
  "checkpoint_num_layers": 1,
  "partition_activations": true,
  "synchronize_each_layer": true,

  # regularization
  "gradient_clipping": 1.0,
  "weight_decay": 0.1,
  "hidden_dropout": 0,
  "attention_dropout": 0,

  # precision settings
  "bf16": {
    "bf16": true,
    "enabled": true,
    "loss_scale": 0,
    "loss_scale_window": 1000,
    "initial_scale_power": 12,
    "hysteresis": 2,
    "min_loss_scale": 1,
  },

  # misc. training settings
  "train_iters": 500,
  "lr_decay_iters": 500,
  "distributed_backend": "nccl",
  "lr_decay_style": "constant",
  "warmup": 0.01,
  "checkpoint_factor": 100,
  "eval_interval": 100000,
  "eval_iters": 10,

  # logging
  "log_interval": 10,
  "steps_per_print": 10,
  "wall_clock_breakdown": true,
}