config.json

{
  "_name_or_path": "memoryllm-8b",
  "add_bos_embedding": true,
  "add_decoder_lora": true,
  "architectures": [
    "MemoryLLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "bos_token_id": 128000,
  "drop_memory_per_layer": true,
  "eos_token_id": 128001,
  "hidden_act": "silu",
  "hidden_size": 4096,
  "initializer_range": 0.02,
  "intermediate_size": 14336,
  "lora_config": {
    "inference_mode": false,
    "lora_alpha": 32,
    "lora_dropout": 0.1,
    "r": 16,
    "target_modules": [
      "q_proj",
      "v_proj",
      "k_proj",
      "up_proj",
      "down_proj",
      "gate_proj"
    ]
  },
  "max_length": 512,
  "max_position_embeddings": 8192,
  "max_seq_length": 2048,
  "mlp_bias": false,
  "model_type": "llama",
  "new_memory_embedding_fullset": true,
  "num_attention_heads": 32,
  "num_blocks": 50,
  "num_hidden_layers": 32,
  "num_key_value_heads": 8,
  "num_memory_tokens": 12800,
  "num_tokens": 256,
  "pad_token_id": 128001,
  "pretraining_tp": 1,
  "rms_norm_eps": 1e-05,
  "rope_scaling": {
    "factor": 1.8125,
    "type": "linear"
  },
  "rope_theta": 500000.0,
  "shrink_to_one_embedding": true,
  "tie_word_embeddings": false,
  "torch_dtype": "float32",
  "transformers_version": "4.43.0.dev0",
  "use_cache": true,
  "vocab_size": 128256
}