I paid to deploy a Dedicated Endpoints for the chemdfm-13b-v1-0-ycy model, but the request reported an error
{'error ':'Request failed during generation: Server error: Unexpected<class 'RuntimeError '>: captures_underground==0 INTERNATIONAL ASSERT FAILED at “/opt/conda/conda bld/Pytorch_1699449201336/work/c10/cuda/CUDAachingallocator. cpp”: 2939, please report a bug to PyTorch. ‘,’ error type ‘:’ generation '}
May I ask where I can provide feedback on abnormalities? Can the fees paid for service errors be refunded?
1 Like
Same error using GOAT-AI/GOAT-70B-Storytelling with a dedicated endpoint on AWS of 2x A100s per the recommended config.