config.json

{
  "architectures": [
    "GraniteMoeHybridForCausalLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "attention_multiplier": 0.015625,
  "bos_token_id": 100257,
  "torch_dtype": "bfloat16",
  "embedding_multiplier": 12,
  "eos_token_id": 100257,
  "hidden_act": "silu",
  "hidden_size": 2560,
  "initializer_range": 0.1,
  "intermediate_size": 8192,
  "layer_types": [
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention",
    "attention"
  ],
  "logits_scaling": 10,
  "mamba_chunk_size": 256,
  "mamba_conv_bias": true,
  "mamba_d_conv": 4,
  "mamba_d_head": 40,
  "mamba_d_state": 256,
  "mamba_expand": 2,
  "mamba_n_groups": 1,
  "mamba_n_heads": 128,
  "mamba_proj_bias": false,
  "max_position_embeddings": 131072,
  "model_type": "granitemoehybrid",
  "normalization_function": "rmsnorm",
  "num_attention_heads": 40,
  "num_experts_per_tok": 0,
  "num_hidden_layers": 40,
  "num_key_value_heads": 8,
  "num_local_experts": 0,
  "output_router_logits": false,
  "pad_token_id": 100256,
  "position_embedding_type": "rope",
  "quantization_config": {
    "_load_in_4bit": true,
    "_load_in_8bit": false,
    "bnb_4bit_compute_dtype": "bfloat16",
    "bnb_4bit_quant_storage": "uint8",
    "bnb_4bit_quant_type": "nf4",
    "bnb_4bit_use_double_quant": true,
    "llm_int8_enable_fp32_cpu_offload": false,
    "llm_int8_has_fp16_weight": false,
    "llm_int8_skip_modules": [
      "embed_tokens",
      "embedding",
      "lm_head",
      "multi_modal_projector",
      "merger",
      "modality_projection",
      "router",
      "visual",
      "vision_tower",
      "mamba",
      "model.layers.17.self_attn",
      "model.layers.19.self_attn",
      "model.layers.8.self_attn",
      "model.layers.2.shared_mlp",
      "model.layers.2.self_attn",
      "model.layers.39.shared_mlp",
      "model.layers.0.shared_mlp",
      "model.layers.6.shared_mlp",
      "model.layers.1.self_attn"
    ],
    "llm_int8_threshold": 6.0,
    "load_in_4bit": true,
    "load_in_8bit": false,
    "quant_method": "bitsandbytes"
  },
  "residual_multiplier": 0.22,
  "rms_norm_eps": 1e-05,
  "rope_scaling": null,
  "rope_theta": 10000000,
  "router_aux_loss_coef": 0.01,
  "shared_intermediate_size": 8192,
  "tie_word_embeddings": true,
  "transformers_version": "4.57.0",
  "unsloth_fixed": true,
  "use_cache": true,
  "vocab_size": 100352
}