akaashrp opened a new pull request, #18467: URL: https://github.com/apache/tvm/pull/18467
Passing in a string to `ArrayDecodeStorage` via the packed function definition sometimes led to memory issues for larger models (such as `gemma-2-9b-it-q4f32_1-MLC`). Replacing string with TVMFFIByteArray* fixes this issue and also alleviates the stack pollution issue discussed in an earlier PR (https://github.com/apache/tvm/pull/18415). Note that this does not completely fix generation for q0f32 models. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
