configs/mamba/mamba-130M.yml

{
  # Parallelism is not yet supported for Mamba
  "pipe_parallel_size": 0,
  "model_parallel_size": 1,

  "num_layers": 24,
  "hidden_size": 768,
  "num_attention_heads": 12, # ignored when using mamba
  "seq_length": 2048,
  "max_position_embeddings": 2048,
  "output_layer_parallelism": "column",
  "norm": "rmsnorm",
  "rms_norm_epsilon": 1.0e-5,

  "attention_config": [[["mamba"], 24]],

  "mamba_selective_scan_fusion": true,
  "mamba_causal_conv_fusion": true,
  "mamba_inner_func_fusion": true, # supersedes scan or conv fusion
  "activation": "silu",

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


   # optimizer settings
   "optimizer": {
     "type": "Adam",
     "params": {
       "lr": 0.0006,
       "betas": [0.9, 0.95],
       "eps": 1.0e-8,
     }
   },
   "min_lr": 0.00006,

   # 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
   "train_micro_batch_size_per_gpu": 4,
   "data_impl": "mmap",

   # 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": 0.0,
   "attention_dropout": 0.0,

   # precision settings
   "fp16": {
     "enabled": true,
     "loss_scale": 0,
     "loss_scale_window": 1000,
     "hysteresis": 2,
     "min_loss_scale": 1
   },

   # misc. training settings
   "train_iters": 320000,
   "lr_decay_iters": 320000,
   "distributed_backend": "nccl",
   "lr_decay_style": "cosine",
   "warmup": 0.01,
   "checkpoint_factor": 10000,
   "eval_interval": 1000,
   "eval_iters": 10,

   # logging
   "log_interval": 100,
   "steps_per_print": 10,
   "keep_last_n_checkpoints": 4,
   "wall_clock_breakdown": true,
}