Does opt_13b model support tensor parallelism vias inferflow? about inferflow

[shumingshi]

Thank you for raising this issue. Tensor parallelsim is not supported so far for this model. We will try to fix this issue. As a mitigation, you can either apply pipeline parallelsim or serve the model on one GPU device (together with quantization if the VRAM of each device is less than 32GB).

[LHQUer]

Thank you for raising this issue. Tensor parallelsim is not supported so far for this model. We will try to fix this issue. As a mitigation, you can either apply pipeline parallelsim or serve the model on one GPU device (together with quantization if the VRAM of each device is less than 32GB).

Thanks for your general answer！