34b can not use gpu

[Lv Qin]

When I run the 34b demo, I add ”model.tie_weights()“。also ”The model weights are not tied. Please use the tie_weights method before using the infer_auto_device function. OrderedDict([('', 'cpu')])“