config.json

{
  "model_type": "custom_transformer",
  "architectures": [
    "CustomTransformerForCausalLM"
  ],
  "vocab_size": 2564,
  "hidden_size": 1,
  "num_hidden_layers": 1,
  "num_attention_heads": 1,
  "num_key_value_heads": 1,
  "ffn_hidden_size": 1,
  "max_position_embeddings": 64,
  "rope_theta": 768.0,
  "norm_eps": 1e-06,
  "dropout": 0.0,
  "qk_bias": true,
  "use_head_gating": true,
  "attn_res_mode": "full",
  "attn_res_block_size": 4,
  "tie_word_embeddings": true,
  "use_cache": false,
  "pad_token_id": null,
  "bos_token_id": null,
  "eos_token_id": null,
  "torch_dtype": "float32",
  "auto_map": {
    "AutoConfig": "configuration.CustomTransformerConfig",
    "AutoModelForCausalLM": "modeling.CustomTransformerForCausalLM"
  }
}