L2L_fineweb-100b_N3.6e08_D4.6e09_C1.0e19/config.json

{
  "activation_type": "gelu",
  "alibi": false,
  "alibi_bias_max": 8.0,
  "attention_dropout": 0.0,
  "attention_layer_norm": true,
  "attention_layer_norm_with_affine": true,
  "bias_for_layer_norm": false,
  "block_group_size": 1,
  "block_type": "sequential",
  "clip_qkv": null,
  "d_model": 1152,
  "embedding_dropout": 0.0,
  "embedding_size": 32000,
  "eos_token_id": 2,
  "flash_attention": false,
  "include_bias": false,
  "init_cutoff_factor": null,
  "init_device": "cpu",
  "init_fn": "mitchell",
  "init_std": 0.02,
  "layer_norm_type": "default",
  "layer_norm_with_affine": true,
  "max_sequence_length": 512,
  "mlp_hidden_size": null,
  "mlp_ratio": 4,
  "multi_query_attention": false,
  "n_heads": 18,
  "n_kv_heads": null,
  "n_layers": 18,
  "pad_token_id": 2,
  "precision": "amp_bf16",
  "residual_dropout": 0.0,
  "rope": true,
  "rope_full_precision": true,
  "scale_logits": false,
  "vocab_size": 32000,
  "weight_tying": false
}