LanShanPi

Results 8 comments of LanShanPi

[BUG]KeyError: 'attention_mask'

@janglichao 我也遇到了这个问题，请问你解决了吗？怎么解决的。

when I execute the command "python train.py --actor-model facebook/opt-1.3b --reward-model facebook/opt-350m --deployment-type single_gpu" I got the follow result, and I try many solutions but all failure.

I execute the follow command to configuration anaconda environment: pip install deepspeed>=0.9.0 git clone https://github.com/microsoft/DeepSpeedExamples.git cd DeepSpeedExamples/applications/DeepSpeed-Chat/ pip install -r requirements.txt

when I execute the command "python train.py --actor-model facebook/opt-1.3b --reward-model facebook/opt-350m --deployment-type single_gpu" I got the follow result, and I try many solutions but all failure.

Running with CUDA 11.7 on Ubuntu 20/04.

when I execute the command "python train.py --actor-model facebook/opt-1.3b --reward-model facebook/opt-350m --deployment-type single_gpu" I got the follow result, and I try many solutions but all failure.

@Flywolfs I am trying to match your version.

nvcc fatal : Unsupported gpu architecture 'compute_native'

在fastllm/CMakeLIsts.txt文件中将set(CMAKE_CUDA_ARCHITECTURES "native") 中的native改成显卡对应的算力，11.7应该对应80.

请教：flm模型或者是llm.model支持指定GPU吗？默认都是GPU:0

@fushengwuyu 这样在用llm.from_hf()加速的时候不会重复加载模型到gpu吗

请教：flm模型或者是llm.model支持指定GPU吗？默认都是GPU:0

@xycjscs 对的

AssertionError: CUDA_HOME does not exist, unable to compile CUDA op(s)

Have you solved this problem yet.