Icemist opened a new pull request #9637: URL: https://github.com/apache/tvm/pull/9637
Below are some results of comparing the old implementation(int32) with the suggested(fp32): <html xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns="http://www.w3.org/TR/REC-html40";> <head> <meta name=ProgId content=Excel.Sheet> <meta name=Generator content="Microsoft Excel 12"> <link id=Main-File rel=Main-File href="file:///C:\Users\icemist\AppData\Local\Temp\msohtmlclip1\01\clip.htm"> <link rel=File-List href="file:///C:\Users\icemist\AppData\Local\Temp\msohtmlclip1\01\clip_filelist.xml"> <!--table {mso-displayed-decimal-separator:"\."; mso-displayed-thousand-separator:"\,";} @page {margin:.75in .7in .75in .7in; mso-header-margin:.3in; mso-footer-margin:.3in;} tr {mso-height-source:auto;} col {mso-width-source:auto;} br {mso-data-placement:same-cell;} td {padding-top:1px; padding-right:1px; padding-left:1px; mso-ignore:padding; color:black; font-size:11.0pt; font-weight:400; font-style:normal; text-decoration:none; font-family:Calibri, sans-serif; mso-font-charset:204; mso-number-format:General; text-align:general; vertical-align:bottom; border:none; mso-background-source:auto; mso-pattern:auto; mso-protection:locked visible; white-space:nowrap; mso-rotate:0;} .xl65 {mso-number-format:Percent;} --> </head> <body link=blue vlink=purple> resnet50_INT8bit_quantized 50000 images Top1 accuracy | llvm | llvm avx2 -- | -- | -- int32 TONEAREST | 75.38% | 75.39% int32 UPWARD | 75.38% | 75.39% float32 TONEAREST | 75.30% | 75.32% float32 UPWARD | 75.30% | 75.32% </body> </html> <html xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns="http://www.w3.org/TR/REC-html40";> <head> <meta name=ProgId content=Excel.Sheet> <meta name=Generator content="Microsoft Excel 12"> <link id=Main-File rel=Main-File href="file:///C:\Users\icemist\AppData\Local\Temp\msohtmlclip1\01\clip.htm"> <link rel=File-List href="file:///C:\Users\icemist\AppData\Local\Temp\msohtmlclip1\01\clip_filelist.xml"> <!--table {mso-displayed-decimal-separator:"\."; mso-displayed-thousand-separator:"\,";} @page {margin:.75in .7in .75in .7in; mso-header-margin:.3in; mso-footer-margin:.3in;} tr {mso-height-source:auto;} col {mso-width-source:auto;} br {mso-data-placement:same-cell;} td {padding-top:1px; padding-right:1px; padding-left:1px; mso-ignore:padding; color:black; font-size:11.0pt; font-weight:400; font-style:normal; text-decoration:none; font-family:Calibri, sans-serif; mso-font-charset:204; mso-number-format:General; text-align:general; vertical-align:bottom; border:none; mso-background-source:auto; mso-pattern:auto; mso-protection:locked visible; white-space:nowrap; mso-rotate:0;} .xl65 {mso-number-format:"0\.000";} --> </head> <body link=blue vlink=purple> relay.qnn.op.requantize median (ms) | llvm | llvm avx2 -- | -- | -- int32 TONEAREST | 2.203 | 1.510 int32 UPWARD | 1.490 | 1.020 float32 TONEAREST | 1.289 | 0.675 float32 UPWARD | 1.140 | 0.666 </body> </html> The performance measurements were made via graph_executor.benchmark on AMD Ryzen 7 5800H with TVM_NUM_THREADS=1. -- 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]
