Safetensors
MineMA-8B / MineMA-3-8b-v4 /untitled.txt
Avansler's picture
Add four folders using Git LFS
c14b4f9
raw
history blame contribute delete
321 Bytes
config = LoraConfig(
task_type=TaskType.CAUSAL_LM,
target_modules=["q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj"],
inference_mode=False, # 训练模式
r=128, # Lora 秩
lora_alpha=256, # Lora alaph,具体作用参见 Lora 原理
lora_dropout=0.1# Dropout 比例
)