feat(train_page): 添加 LoRA 秩动态输入功能

- 在训练页面新增 LoRA 秩输入框，使用户可以动态设置 LoRA 秩
- 更新训练模型函数，添加 LoRA 秩参数并将其用于模型配置
- 保留原有功能，仅增加 LoRA 秩相关配置

frontend/train_page.py

@@ -9,7 +9,7 @@ from unsloth import FastLanguageModel
 from trl import SFTTrainer  # 用于监督微调的训练器
 from transformers import TrainingArguments,DataCollatorForSeq2Seq  # 用于配置训练参数
 from unsloth import is_bfloat16_supported  # 检查是否支持bfloat16精度训练
-from unsloth.chat_templates import get_chat_template
+from unsloth.chat_templates import get_chat_template, train_on_responses_only
 from tools import formatting_prompts_func
 
 sys.path.append(str(Path(__file__).resolve().parent.parent))
@@ -36,18 +36,20 @@ def train_page():
         per_device_train_batch_size_input = gr.Number(value=1, label="batch size", precision=0)
         epoch_input = gr.Number(value=1, label="epoch", precision=0)
         save_steps_input = gr.Number(value=20, label="保存步数", precision=0)  # 新增保存步数输入框
+        lora_rank_input = gr.Number(value=16, label="LoRA秩", precision=0)  # 新增LoRA秩输入框
         
         train_button = gr.Button("开始微调")
         
         # 训练状态输出
         output = gr.Textbox(label="训练日志", interactive=False)
         
-        def train_model(dataset_name, learning_rate, per_device_train_batch_size, epoch, save_steps):
+        def train_model(dataset_name, learning_rate, per_device_train_batch_size, epoch, save_steps, lora_rank):
             # 使用动态传入的超参数
             learning_rate = float(learning_rate)
             per_device_train_batch_size = int(per_device_train_batch_size)
             epoch = int(epoch)
             save_steps = int(save_steps)  # 新增保存步数参数
+            lora_rank = int(lora_rank)  # 新增LoRA秩参数
             
             # 模型配置参数
             dtype = None  # 数据类型，None表示自动选择
@@ -62,7 +64,7 @@ def train_page():
                 model,
                 # LoRA秩,用于控制低秩矩阵的维度,值越大表示可训练参数越多,模型性能可能更好但训练开销更大
                 # 建议: 8-32之间
-                r=16,
+                r=lora_rank,  # 使用动态传入的LoRA秩
                 # 需要应用LoRA的目标模块列表
                 target_modules=[
                     "q_proj", "k_proj", "v_proj", "o_proj",  # attention相关层
@@ -132,7 +134,13 @@ def train_page():
                     # report_to="tensorboard",  # 将信息输出到tensorboard
                 ),
             )
-            
+
+            trainer = train_on_responses_only(
+                trainer,
+                instruction_part = "<|im_start|>user\n",
+                response_part = "<|im_start|>assistant\n",
+            )
+
             # 开始训练
             trainer_stats = trainer.train(resume_from_checkpoint=False)
 
@@ -155,6 +163,5 @@ if __name__ == "__main__":
     from model_manage_page import model_manage_page
     init_global_var("workdir")
     demo = gr.TabbedInterface([model_manage_page(), train_page()], ["模型管理", "聊天"])
-    # demo = gr.TabbedInterface([ train_page()], ["模型管理", "聊天"])    
     demo.queue()
     demo.launch()