Closed PeiMu closed 1 year ago
I reopen this since #632 only adds the basic infrastructure and not the complete functionality. Maybe we can rename this issue as "Auto Quantization - Basic infrastructure" and close this, or keep this open until all the functionality is implemented.
In general, I'm not sure what is preferred, having a general issue that will be addressed by a very large PR or smaller specific ones, addressed by smaller PRs? @KomaGR @btsouts
I would suggest renaming and closing this as you suggest. If there are specific next steps with regards to this, please open a new issue to describe/list them and backlink to this one.
Also, I would greatly appreciate the addition of some documentation on how to use the existing functionality. Perhaps a README in a subfolder?
Also, I would greatly appreciate the addition of some documentation on how to use the existing functionality. Perhaps a README in a subfolder?
Yes, me and @PeiMu will handle this! Closing this.
The compilers can do quantization automatically through
accuracy
in.nt
file.Maybe we can also use the Quantization in MLIR.