-
### Description
### Alternatives
_No response_
-
这里支持CPU部署模型吗？商用的话，一般生产环境缺少GPU机器
当前cpu加载模型的时候会出错
/data/liu/CPM-Bee/src/test_cpm.py:180 in …
-
### System Info
flagai 1.7.3
centos
### Information
- [X] The official example scripts
- [ ] My own modified scripts
### Tasks
- [X] An officially supported task in the `e…
-
### System Info
跑预训练，bmtrain 有这个报错
bash dist_trigger_docker.sh hostfile Aquila-pretrain.yaml aquila-7b test0
![9201686728893_ pic](https://github.com/FlagAI-Open/FlagAI/assets/8345745/68d38f2e-ff4f…
-
您好，目前应用部署在V100s上，1s约生成10个字左右，请问下是否可以提供模型推理加速的一些方案，谢谢
-
### Description
我通过git clone https://github.com/FlagAI-Open/FlagAI.git,下载源码。并按照README.md按照好依赖了。
我在./FlagAI目录下，是可以成功执行 import flagai.model.tools.peft的
当我进入./FlagAi/examples/Aquila/Aquila-chat后，执…
-
### Description
1、拉取commit id为e1ab88a的最新代码
2、调用examples/Aquila/Aquila-chat/local_trigger_docker.sh
Aquila-chat-lora.yaml配置
batch_size: 1
gradient_accumulation_steps: 1
lr: 5.0e-5
warm_up: 0…
-
**_!pip install -r requirements.txt_**
Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Collecting torch=1.10
Using cached torch-1.13.1-cp310-c…
-
When NaN happens during forward step of one rank, how to skip this global step for all ranks? Thanks.
-
```
if args.use_delta:
delta_model = LoraModel(
backbone_model=model, modified_modules=["project_q", "project_v"], backend="bmt"
)
lora_ckpt_path = 'results/cp…