AquilaChat lora微调后，无法加载模型

[yuqin11]

System Info

flagai 1.7.3 torch 2.0.1 cuda 11.0

Information

• [X] The official example scripts
• [ ] My own modified scripts

Tasks

• [ ] An officially supported task in the examples folder (such as T5/AltCLIP, ...)
• [X] My own task or dataset (give details below)

Reproduction

微调关键事项：注释了如下代码

加载模型的代码

错误信息： ******************** lm aquilachat-7b Traceback (most recent call last): File "generate_chat.py", line 11, in loader = AutoLoader( File "/dockerdata/buyllm/anaconda3/envs/falgai/lib/python3.8/site-packages/flagai/auto_model/auto_loader.py", line 216, in init self.model = getattr(LazyImport(self.model_name[0]), File "/dockerdata/buyllm/anaconda3/envs/falgai/lib/python3.8/site-packages/flagai/model/base_model.py", line 184, in from_pretrain return load_local(checkpoint_path, only_download_config=only_download_config) File "/dockerdata/buyllm/anaconda3/envs/falgai/lib/python3.8/site-packages/flagai/model/base_model.py", line 115, in load_local model = cls.init_from_json(config_path, **kwargs) File "/dockerdata/buyllm/anaconda3/envs/falgai/lib/python3.8/site-packages/flagai/model/base_model.py", line 69, in init_from_json model = cls(change_json_to_cls(args), **kwargs) File "/dockerdata/buyllm/anaconda3/envs/falgai/lib/python3.8/site-packages/flagai/model/aquila_model.py", line 120, in init self.layers.append(AQUILABlock(layer_id, config)) File "/dockerdata/buyllm/anaconda3/envs/falgai/lib/python3.8/site-packages/flagai/model/blocks/aquila_block.py", line 37, in init self.feed_forward = AQUILAForward( File "/dockerdata/buyllm/anaconda3/envs/falgai/lib/python3.8/site-packages/flagai/model/layers/feedforward.py", line 84, in init hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of) TypeError: unsupported operand type(s) for +: 'int' and 'NoneType'

Expected behavior

正常加载并chat

[BAAI-OpenPlatform]

lora得用单独的方式加载，先不用lora吧，我们修修，实在是不好意思

[yuqin11]

不用lora，微调以后，将最后一个step（数字结尾的文件夹里面的）这两个文件覆盖下载下来的文件仍然报错 -rw-rw-r-- 1 buyllm buyllm 2 6月 16 01:25 config.json -rw-rw-r-- 1 buyllm buyllm 14G 6月 16 01:24 pytorch_model.bin 用上诉方法加载，后仍然相同的错误 @BAAI-OpenPlatform

[yuqin11]

另外你们也没有文档说明，微调后的模型如何使用

[yuqin11]

一言难尽

[yuqin11]

@BAAI-OpenPlatform 这里有关注吗

[zhangxuan602461121]

按照 Aquila-7B 的 README 执行到 python generate.py 的时候，有跟楼主相同的报错