adapter_config.json

{
  "adaptive_ratio": 0.01,
  "adaptive_ratio_decay": 1.005,
  "additive_modeling": false,
  "allow_empty_lora": false,
  "auto_mapping": null,
  "base_model_name_or_path": "microsoft/Phi-3-mini-4k-instruct",
  "bias": "none",
  "curr_learning": true,
  "detached_training": true,
  "dynamic_adapter_pool": true,
  "enable_lora": null,
  "encoder_hidden_size": 3072,
  "fan_in_fan_out": false,
  "hypernetwork": true,
  "inference_mode": true,
  "input_based_adapter_selection": true,
  "insert_zero_lora": false,
  "layer_to_lora": [],
  "lora_alpha": 16,
  "lora_dropout": 0.05,
  "merge_weights": false,
  "modules_to_save": null,
  "num_attention_heads": 32,
  "num_layers": 32,
  "num_prefix_set": 3,
  "num_transformer_submodules": 1,
  "num_virtual_tokens": 30,
  "number_of_adapter_pre_layer": 8,
  "ot_diversified_dispatcher": false,
  "ot_diversified_prefix": false,
  "peft_type": "PREFIX_MA_LORA",
  "pool_selective_inference": true,
  "pool_selective_training": true,
  "prefix_projection": true,
  "r": 8,
  "random_routing": false,
  "random_routing_inference": false,
  "scale": 64,
  "selective_num": 8,
  "simple_hidden_matching": true,
  "simple_instance_matching": true,
  "target_modules": [
    "qkv_proj"
  ],
  "task_type": "CAUSAL_LM",
  "token_dim": 3072
}