bamba/BambaModel/config.json

{
  "architectures": [
    "BambaModel"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "attn_layer_indices": [
    1,
    3
  ],
  "attn_rotary_emb": 8,
  "bos_token_id": 1,
  "dtype": "float32",
  "eos_token_id": 2,
  "hidden_act": "silu",
  "hidden_size": 32,
  "initializer_range": 0.02,
  "intermediate_size": 64,
  "mamba_chunk_size": 16,
  "mamba_conv_bias": true,
  "mamba_d_conv": 4,
  "mamba_d_head": 4,
  "mamba_d_state": 16,
  "mamba_expand": 2,
  "mamba_n_groups": 1,
  "mamba_n_heads": 16,
  "mamba_proj_bias": false,
  "max_position_embeddings": 512,
  "mlp_bias": false,
  "model_type": "bamba",
  "num_attention_heads": 4,
  "num_hidden_layers": 4,
  "num_key_value_heads": 2,
  "num_logits_to_keep": 1,
  "pad_token_id": 0,
  "partial_rotary_factor": 0.5,
  "rms_norm_eps": 1e-05,
  "rope_parameters": {
    "partial_rotary_factor": 0.5,
    "rope_theta": 10000.0,
    "rope_type": "default"
  },
  "tie_word_embeddings": false,
  "time_step_limit": [
    0.0,
    {
      "__float__": "Infinity"
    }
  ],
  "transformers_version": "5.15.0.dev0",
  "use_cache": true,
  "vocab_size": 1024,
  "z_loss_coefficient": 0.0
}