Open Pattaro opened 1 year ago
Any solutions? I encountered the same issue with bloomz model, so far I just removed --enable_hybrid_engine to bypass the issue, and the program runs. However, guess the efficiency would reduce.
It seems enable_hybrid_engine only support for OPT models.
try adjusting the --inference_tp_size to a lower number, it may be you don't have enough GPUs across your nodes.
[bug]AttributeError: 'DeepSpeedHybridEngine' object has no attribute 'mp_group' #525