|
{ |
|
"adaptive_ratio": 0.05, |
|
"adaptive_ratio_decay": 1.001, |
|
"additive_modeling": false, |
|
"auto_mapping": null, |
|
"base_model_name_or_path": "microsoft/phi-2", |
|
"bias": "none", |
|
"curr_learning": true, |
|
"detached_training": false, |
|
"dynamic_adapter_pool": true, |
|
"enable_lora": null, |
|
"encoder_hidden_size": 2560, |
|
"fan_in_fan_out": false, |
|
"hypernetwork": false, |
|
"inference_mode": true, |
|
"input_based_adapter_selection": true, |
|
"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": false, |
|
"r": 16, |
|
"scale": 64, |
|
"selective_num": 8, |
|
"simple_hidden_matching": false, |
|
"simple_instance_matching": true, |
|
"target_modules": [ |
|
"q_proj", |
|
"k_proj", |
|
"v_proj" |
|
], |
|
"task_type": "CAUSAL_LM", |
|
"token_dim": 2560 |
|
} |