Llama3.1-Ar-ce-AdaLoRA / adapter_config.json
arshiakarimian1's picture
Training in progress, epoch 1
79a56b9 verified
raw
history blame
855 Bytes
{
"alpha_pattern": {},
"auto_mapping": null,
"base_model_name_or_path": "meta-llama/Meta-Llama-3.1-8B-Instruct",
"beta1": 0.85,
"beta2": 0.85,
"bias": "none",
"deltaT": 1,
"fan_in_fan_out": false,
"inference_mode": true,
"init_lora_weights": true,
"init_r": 12,
"layer_replication": null,
"layers_pattern": null,
"layers_to_transform": null,
"loftq_config": {},
"lora_alpha": 2,
"lora_dropout": 0.01,
"megatron_config": null,
"megatron_core": "megatron.core",
"modules_to_save": [
"classifier",
"score"
],
"orth_reg_weight": 0.5,
"peft_type": "ADALORA",
"r": 64,
"rank_pattern": null,
"revision": null,
"target_modules": [
"k_proj",
"v_proj"
],
"target_r": 8,
"task_type": "SEQ_CLS",
"tfinal": 0,
"tinit": 0,
"total_step": null,
"use_dora": false,
"use_rslora": false
}