Merge branch 'master' of https://github.com/datawhalechina/self-llm

xueyou2000 · Feb 19, 2024 · c2e6ed0 · c2e6ed0
2 parents da4ea1e + f8a89a3
commit c2e6ed0
Show file tree

Hide file tree

Showing 3 changed files with 11 additions and 11 deletions.
diff --git a/ChatGLM/06-ChatGLM3-6B-Lora微调.ipynb b/ChatGLM/06-ChatGLM3-6B-Lora微调.ipynb
@@ -58,7 +58,7 @@
    "outputs": [],
    "source": [
     "# 将JSON文件转换为CSV文件\n",
-    "df = pd.read_json('../dataset/huanhuan.jsonl')\n",
+    "df = pd.read_json('../dataset/huanhuan.json')\n",
     "ds = Dataset.from_pandas(df)"
    ]
   },
@@ -112,7 +112,7 @@
     }
    ],
    "source": [
-    "tokenizer = AutoTokenizer.from_pretrained(\"/CV/xhr/xhr_project/LLM_learn/transformers-code-master/model/chatglm3-6b\", trust_remote_code=True)\n",
+    "tokenizer = AutoTokenizer.from_pretrained(\"/root/autodl-tmp/ZhipuAI/chatglm3-6b\", trust_remote_code=True)\n",
     "tokenizer"
    ]
   },
@@ -313,7 +313,7 @@
     }
    ],
    "source": [
-    "model = AutoModelForCausalLM.from_pretrained(\"/CV/xhr/xhr_project/LLM_learn/transformers-code-master/model/chatglm3-6b\", trust_remote_code=True, low_cpu_mem_usage=True)"
+    "model = AutoModelForCausalLM.from_pretrained(\"/root/autodl-tmp/ZhipuAI/chatglm3-6b\", trust_remote_code=True, low_cpu_mem_usage=True)"
    ]
   },
   {
@@ -357,7 +357,7 @@
    "source": [
     "from peft import LoraConfig, TaskType, get_peft_model\n",
     "\n",
-    "# model = AutoModelForCausalLM.from_pretrained(\"/CV/xhr/xhr_project/LLM_learn/transformers-code-master/model/chatglm3-6b\", low_cpu_mem_usage=True)\n",
+    "# model = AutoModelForCausalLM.from_pretrained(\"/root/autodl-tmp/ZhipuAI/chatglm3-6b\", low_cpu_mem_usage=True)\n",
     "config = LoraConfig(task_type=TaskType.CAUSAL_LM, target_modules={\"query_key_value\"}, r=8, lora_alpha=32)\n",
     "config"
    ]

diff --git a/ChatGLM/06-ChatGLM3-6B-Lora微调.py b/ChatGLM/06-ChatGLM3-6B-Lora微调.py
@@ -37,15 +37,15 @@ def process_func(example):
 
 if "__main__" == __name__:
     # 将JSON文件转换为CSV文件,处理数据集
-    df = pd.read_json('../dataset/huanhuan.jsonl')
+    df = pd.read_json('../dataset/huanhuan.json')
     ds = Dataset.from_pandas(df)
     # 加载tokenizer
-    tokenizer = AutoTokenizer.from_pretrained("./model/chatglm3-6b", trust_remote_code=True)
+    tokenizer = AutoTokenizer.from_pretrained("/root/autodl-tmp/ZhipuAI/chatglm3-6b", trust_remote_code=True)
     # 将数据集变化为token形式
     tokenized_ds = ds.map(process_func, remove_columns=ds.column_names)
 
     # 创建模型
-    model = AutoModelForCausalLM.from_pretrained("./model/chatglm3-6b",torch_dtype=torch.half, trust_remote_code=True, low_cpu_mem_usage=True)
+    model = AutoModelForCausalLM.from_pretrained("/root/autodl-tmp/ZhipuAI/chatglm3-6b",torch_dtype=torch.half, trust_remote_code=True, low_cpu_mem_usage=True)
 
     # 创建loRA参数
     config = LoraConfig(task_type=TaskType.CAUSAL_LM, target_modules={"query_key_value"}, r=8, lora_alpha=32)

diff --git a/Qwen/05-Qwen-7B-Chat Ptuning 微调.py b/Qwen/05-Qwen-7B-Chat Ptuning 微调.py
@@ -47,19 +47,19 @@ def process_func(example):
 
 
 if "__main__" == __name__:
-    os.chdir('/CV/xhr/xhr_project/LLM_learn/transformers-code-master/self-llm')
+    os.chdir('/root/self-llm')  # /root/self-llm 需改成自己 self-llm 项目的绝对路径
     # 处理数据集
     # 将JSON文件转换为CSV文件
-    df = pd.read_json('./huanhuan.json')
+    df = pd.read_json('./dataset/huanhuan.json')
     ds = Dataset.from_pandas(df)
     # 加载tokenizer
-    tokenizer = AutoTokenizer.from_pretrained('/CV/xhr/xhr_project/LLM_learn/transformers-code-master/model/Qwen-7B-Chat', use_fast=False, trust_remote_code=True)
+    tokenizer = AutoTokenizer.from_pretrained('/root/autodl-tmp/qwen/Qwen-7B-Chat', use_fast=False, trust_remote_code=True)
     tokenizer.pad_token_id = tokenizer.eod_id
     # 将数据集变化为token形式
     tokenized_id = ds.map(process_func, remove_columns=ds.column_names)
 
     # 创建模型并以半精度形式加载
-    model = AutoModelForCausalLM.from_pretrained('/CV/xhr/xhr_project/LLM_learn/transformers-code-master/model/Qwen-7B-Chat', trust_remote_code=True, torch_dtype=torch.half, device_map="auto")
+    model = AutoModelForCausalLM.from_pretrained('/root/autodl-tmp/qwen/Qwen-7B-Chat', trust_remote_code=True, torch_dtype=torch.half, device_map="auto")
     # model.enable_input_require_grads()  # 开启梯度检查点时，要执行该方法
     # 加载lora参数
     model = get_peft_model(model, config)