MOSS

MOSS copied to clipboard

请教 RuntimeError: `<class 'models.quantization.QuantLinear'>' was not properly set up for sharding by zero.Init(). A subclass of torch.nn.Module must be defined before zero.Init() where an instance of the class is created.

13

请问量化的模型怎么训练呀，可以单卡吗，只要把路径改了就行吗 RuntimeError: `' was not properly set up for sharding by zero.Init(). A subclass of torch.nn.Module must be defined before zero.Init() where an instance of the class is created. ERROR:torch.distributed.elastic.multiprocessing.api:failed...

starplatinum3

``` AttributeError: type object 'QuantLinear' has no attribute '_old_init' To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`. library_or_version 0.6.5 INFO:torch.distributed.distributed_c10d:Added key:...

starplatinum3

Fix a function naming error

jotline

多卡finetune launch_agent run报错

2

hello，使用accelerate多卡finetune时launch_agent报错 报错信息如下： ```` INFO:torch.distributed.distributed_c10d:Added key: store_based_barrier_key:2 to store for rank: 2 INFO:torch.distributed.distributed_c10d:Added key: store_based_barrier_key:2 to store for rank: 3 INFO:torch.distributed.distributed_c10d:Added key: store_based_barrier_key:2 to store for rank: 1 INFO:torch.distributed.distributed_c10d:Added key: store_based_barrier_key:2...

yueshiyi

在8张40g A100上运行微调代码，bsz=1，报显存不够错误，请问最低训练硬件条件是什么？

11

KUANWB

python moss_cli_demo.py Traceback (most recent call last): File "moss_cli_demo.py", line 6, in import torch File "D:\workspace\install\miniconda3\lib\site-packages\torch\__init__.py", line 128, in raise err OSError: [WinError 127] 找不到指定的程序。 Error loading "D:\workspace\install\miniconda3\lib\site-packages\torch\lib\caffe2_detectron_ops.dll" or one...

xiaohe-person

This PR adds a reproducible conda environment that works on both Linux/Windows. ```bash conda env create --file conda-recipe.yaml # or `mamba env create --file conda-recipe.yaml` conda activate moss python3 moss_cli_demo.py...

XuehaiPan

Respect user-defined `CUDA_VISIBLE_DEVICES` in demos if it is set.

XuehaiPan

如何修改.cache的路径

2

本人单机测试时，由于C盘较小，下载的huggingface文件夹较大，快放不下了，请问如何在pip的时候修改下载路径

xiaohe-person

采用微调代码训练后得到的pytorch_model.bin达到了62g，请问有什么办法拆分成多个文件并且满足推理代码的调用格式？

4

KUANWB