search

bytedance / lightseq

LightSeq: A High Performance Library for Sequence Processing and Generation
Other
3.22k stars 329 forks source link

[CUDA][ERROR]: misaligned address #329

Closed fc20567 closed 2 years ago

fc20567 commented 2 years ago

hi, I have 1 question. When a large amount of text requests the model, the model starts to run properly. After the model runs for a period of time, the program reports an error : [CUDA][ERROR] /tmp/build-via-sdist-uagdfpbf/lightseq-2.2.1/lightseq/inference/pywrapper/gpt.cc.cu(160): misaligned address.

hexisyztem commented 2 years ago

What is the run command you executed.

hexisyztem commented 2 years ago

I guess that your input request is too large, in this case it is recommended to use tritonserver to schedule requests.

fc20567 commented 2 years ago

What is the run command you executed. cuda_error

fc20567 commented 2 years ago

I guess that your input request is too large, in this case it is recommended to use tritonserver to schedule requests.

The test code has been posted. Is this the error caused by large requests? thank you.

hexisyztem commented 2 years ago

Sorry, I missed your email message over the weekend. You can try to run the input statement that reports the error once alone for further judgment.

fc20567 commented 2 years ago

Sorry, I missed your email message over the weekend. You can try to run the input statement that reports the error once alone for further judgment.

Thx，problem has solved. input data have "-" symbol, cause model.sample() throw an exception.