llada_40m_dl/checkpoint-463694/config.json

{
  "_comment_activation": "MLP activation function aligned with ModernBert",
  "_comment_attention": "Attention and RoPE settings for shorter context",
  "_comment_bias": "Bias settings for Linear layers aligned with ModernBert",
  "_comment_core_dims": "\u224840M parameter LLaDA variant for TinyStories-scale training",
  "_comment_dropout": "Dropout rates aligned with ModernBert",
  "_comment_initialization": "Initialization scheme aligned with ModernBert",
  "_comment_misc": "Other settings for decoder-style model",
  "_comment_normalization": "LayerNorm settings aligned with ModernBert",
  "activation_type": "silu",
  "alibi": false,
  "alibi_bias_max": 8.0,
  "architectures": [
    "LLaDAModelLM"
  ],
  "attention_dropout": 0.0,
  "attention_layer_norm": false,
  "attention_layer_norm_with_affine": true,
  "auto_map": {
    "AutoConfig": "configuration_llada.LLaDAConfig",
    "AutoModel": "modeling_llada.LLaDAModelLM"
  },
  "bias_for_layer_norm": false,
  "block_group_size": 1,
  "block_type": "llama",
  "d_model": 384,
  "dtype": "float32",
  "embedding_dropout": 0.0,
  "embedding_size": 50368,
  "eos_token_id": 50256,
  "flash_attention": false,
  "include_bias": false,
  "include_qkv_bias": false,
  "init_cutoff_factor": 2.0,
  "init_device": "meta",
  "init_fn": "full_megatron",
  "init_std": 0.02,
  "input_emb_norm": false,
  "layer_norm_type": "default",
  "layer_norm_with_affine": true,
  "mask_token_id": 50256,
  "max_sequence_length": 2048,
  "mlp_hidden_size": 1024,
  "mlp_ratio": 4,
  "model_type": "llada",
  "multi_query_attention": null,
  "n_heads": 6,
  "n_kv_heads": 6,
  "n_layers": 12,
  "pad_token_id": 50283,
  "precision": null,
  "residual_dropout": 0.0,
  "rms_norm_eps": 1e-05,
  "rope": true,
  "rope_full_precision": true,
  "rope_theta": 10000.0,
  "scale_logits": false,
  "transformers_version": "4.57.1",
  "use_cache": false,
  "vocab_size": 50368,
  "weight_tying": true
}