TI中文支持网
TI专业的中文技术问题搜集分享网站

TDA4VM: calibration=13在08_02 tidl_import上转换成功 升级换成08_06 tidl_import上转换不成功

Part Number:TDA4VM

mobinetv2+ssd 的目标检测网络模型设置8bit calibration=13 在tidl_j721e_08_02_00_11上可以
用tidl_model_import.out成功量化转换并且部署到TDA4VM芯片上成功运行,到了tidl_j721e_08_06_00_10 上用同样的配置参数,报错如下:
~~~~~Running TIDL in PC emulation mode to collect Activations range for each layer~~~~~

Processing config file #0 : /home/zet/byd/zhangtiansheng/8_6/8bit_cali13/det_test_mix.txt.qunat_stats_config.txt
Freeing memory for user provided Net
———————– TIDL Process with REF_ONLY FLOW ————————

# 0 . .. T 12279.78 …. ….. … …. …..
# 1 . .. T 12010.71 …. ….. … …. …..
~~~~~Running TIDL in PC emulation mode to collect Activations range for each layer~~~~~

Processing config file #0 : /home/zet/byd/zhangtiansheng/8_6/8bit_cali13/det_test_mix.txt.qunat_stats_config.txt
Freeing memory for user provided Net
———————– TIDL Process with REF_ONLY FLOW ————————

# 0 . .. T 6885.71 Error at line: 1313 : in file /opt/Jenkins_OUT/workspace/J7_TIDL_Parent/c7x-mma-tidl/ti_dl/release/tidl_j721e_08_06_00_10/ti_dl/test/src/pc_linux/../tidl_rt.c, of function : TIDLRT_invoke
Invalid Error Type!
…. …..Error at line: 705 : in file /opt/Jenkins_OUT/workspace/J7_TIDL_Parent/c7x-mma-tidl/ti_dl/release/tidl_j721e_08_06_00_10/ti_dl/test/src/pc_linux/../tidl_tb.c, of function : tidlMultiInstanceTest
Invalid Error Type!
Error at line: 711 : in file /opt/Jenkins_OUT/workspace/J7_TIDL_Parent/c7x-mma-tidl/ti_dl/release/tidl_j721e_08_06_00_10/ti_dl/test/src/pc_linux/../tidl_tb.c, of function : tidlMultiInstanceTest
Invalid Error Type!

Cherry Zhou:

您好,我们已收到您的问题并升级到英文论坛寻求帮助,链接如下,如有答复将尽快回复您:

e2e.ti.com/…/tda4vm-calibration-13-convert-properlyvon-08_02-tidl_import-and-fails-after-replacing-it-with-08_06-tidl_import

,

Cherry Zhou:

您好,

方便分享下您使用的导入配置文件吗?

,

xingye zheng:

,

Cherry Zhou:

好的感谢您提供的信息,我们跟进给工程师看下哈。

,

Cherry Zhou:

配置文件没有什么错误。 您试下使能debugTraceLevel = 2然后分享下log?

请问下您为什么使用 calibrationOption=13? 

,

xingye zheng:

 calibrationOption=13效果相对较好

,

xingye zheng:

tidl_net.bin_netLog.txt

Num of Layer Detected :89--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------Num|TIDL Layer Name|Out Data Name|Group |#Ins|#Outs |Inbuf Ids|Outbuf Id |In NCHW|Out NCHW|MACS|
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------0|TIDL_DataLayer|input_image_3_640_640_original|0|-1|1|xxxxxxxx |0|0000 |13640640 |0 |1|TIDL_ConvolutionLayer|411|0|1|1|0xxxxxxx |1|13640640 |132320320 |91750400 |2|TIDL_ConvolutionLayer|414|0|1|1|1xxxxxxx |2|132320320 |132320320 | 108134400 |3|TIDL_ConvolutionLayer|417|0|1|1|2xxxxxxx |3|132320320 |132320320 |32768000 |4|TIDL_ConvolutionLayer|615|0|1|1|3xxxxxxx |4|132320320 |124320320 |78643200 |5|TIDL_ConvolutionLayer|422|0|1|1|4xxxxxxx |5|124320320 |1144320320 | 368640000 |6|TIDL_ConvolutionLayer|425|0|1|1|5xxxxxxx |6|1144320320 |1144160160 |36864000 |7|TIDL_ConvolutionLayer|624|0|1|1|6xxxxxxx |7|1144160160 |136160160 | 132710400 |8|TIDL_ConvolutionLayer|430|0|1|1|7xxxxxxx |8|136160160 |1216160160 | 204595200 |9|TIDL_ConvolutionLayer|433|0|1|1|8xxxxxxx |9|1216160160 |1216160160 |55296000 |10|TIDL_ConvolutionLayer|633|0|1|1|9xxxxxxx | 10|1216160160 |136160160 | 199065600 |11|TIDL_EltWiseLayer|436|0|2|1|710xxxxxx | 11|136160160 |136160160 |921600 |12|TIDL_ConvolutionLayer|439|0|1|1| 11xxxxxxx | 12|136160160 |1216160160 | 204595200 |13|TIDL_ConvolutionLayer|442|0|1|1| 12xxxxxxx | 13|1216160160 |12168080 |13824000 |14|TIDL_ConvolutionLayer|642|0|1|1| 13xxxxxxx | 14|12168080 |1488080 |66355200 |15|TIDL_ConvolutionLayer|447|0|1|1| 14xxxxxxx | 15|1488080 |12888080 |90316800 |16|TIDL_ConvolutionLayer|450|0|1|1| 15xxxxxxx | 16|12888080 |12888080 |18432000 |17|TIDL_ConvolutionLayer|651|0|1|1| 16xxxxxxx | 17|12888080 |1488080 |88473600 |18|TIDL_EltWiseLayer|453|0|2|1| 1417xxxxxx | 18|1488080 |1488080 |307200 |19|TIDL_ConvolutionLayer|456|0|1|1| 18xxxxxxx | 19|1488080 |12888080 |90316800 |20|TIDL_ConvolutionLayer|459|0|1|1| 19xxxxxxx | 20|12888080 |12888080 |18432000 |21|TIDL_ConvolutionLayer|660|0|1|1| 20xxxxxxx | 21|12888080 |1488080 |88473600 |22|TIDL_EltWiseLayer|462|0|2|1| 1821xxxxxx | 22|1488080 |1488080 |307200 |23|TIDL_ConvolutionLayer|465|0|1|1| 22xxxxxxx | 23|1488080 |12888080 |90316800 |24|TIDL_ConvolutionLayer|468|0|1|1| 23xxxxxxx | 24|12888080 |12884040 |4608000 |25|TIDL_ConvolutionLayer|669|0|1|1| 24xxxxxxx | 25|12884040 |1964040 |44236800 |26|TIDL_ConvolutionLayer|473|0|1|1| 25xxxxxxx | 26|1964040 |15764040 |89395200 |27|TIDL_ConvolutionLayer|476|0|1|1| 26xxxxxxx | 27|15764040 |15764040 |9216000 |28|TIDL_ConvolutionLayer|678|0|1|1| 27xxxxxxx | 28|15764040 |1964040 |88473600 |29|TIDL_EltWiseLayer|479|0|2|1| 2528xxxxxx | 29|1964040 |1964040 |153600 |30|TIDL_ConvolutionLayer|482|0|1|1| 29xxxxxxx | 30|1964040 |15764040 |89395200 |31|TIDL_ConvolutionLayer|485|0|1|1| 30xxxxxxx | 31|15764040 |15764040 |9216000 |32|TIDL_ConvolutionLayer|687|0|1|1| 31xxxxxxx | 32|15764040 |1964040 |88473600 |33|TIDL_EltWiseLayer|488|0|2|1| 2932xxxxxx | 33|1964040 |1964040 |153600 |34|TIDL_ConvolutionLayer|491|0|1|1| 33xxxxxxx | 34|1964040 |15764040 |89395200 |35|TIDL_ConvolutionLayer|494|0|1|1| 34xxxxxxx | 35|15764040 |15764040 |9216000 |36|TIDL_ConvolutionLayer|696|0|1|1| 35xxxxxxx | 36|15764040 |1964040 |88473600 |37|TIDL_EltWiseLayer|497|0|2|1| 3336xxxxxx | 37|1964040 |1964040 |153600 |38|TIDL_ConvolutionLayer|500|0|1|1| 37xxxxxxx | 38|1964040 |15764040 |89395200 |39|TIDL_ConvolutionLayer|503|0|1|1| 38xxxxxxx | 39|15764040 |15764040 |9216000 |40|TIDL_ConvolutionLayer|705|0|1|1| 39xxxxxxx | 40|15764040 |11604040 | 147456000 |41|TIDL_ConvolutionLayer|508|0|1|1| 40xxxxxxx | 41|11604040 |19604040 | 247296000 |42|TIDL_ConvolutionLayer|511|0|1|1| 41xxxxxxx | 42|19604040 |19604040 |15360000 |43|TIDL_ConvolutionLayer|714|0|1|1| 42xxxxxxx | 43|19604040 |11604040 | 245760000 |44|TIDL_EltWiseLayer|514|0|2|1| 4043xxxxxx | 44|11604040 |11604040 |256000 |45|TIDL_ConvolutionLayer|517|0|1|1| 44xxxxxxx | 45|11604040 |19604040 | 247296000 |46|TIDL_ConvolutionLayer|520|0|1|1| 45xxxxxxx | 46|19604040 |19604040 |15360000 |47|TIDL_ConvolutionLayer|723|0|1|1| 46xxxxxxx | 47|19604040 |11604040 | 245760000 |48|TIDL_EltWiseLayer|523|0|2|1| 4447xxxxxx | 48|11604040 |11604040 |256000 |49|TIDL_ConvolutionLayer|526|0|1|1| 48xxxxxxx | 49|11604040 |19604040 | 247296000 |50|TIDL_ConvolutionLayer|loc_0|0|1|1| 48xxxxxxx | 50|11604040 |1484040 |12288000 |51|TIDL_ConvolutionLayer|conf_0|0|1|1| 48xxxxxxx | 51|11604040 |1124040 |3072000 |52|TIDL_ConvolutionLayer|529|0|1|1| 49xxxxxxx | 52|19604040 |19602020 |3840000 |53|TIDL_ConvolutionLayer|732|0|1|1| 52xxxxxxx | 53|19602020 |12402020 |92160000 |54|TIDL_ConvolutionLayer|534|0|1|1| 53xxxxxxx | 54|12402020 |114402020 | 138816000 |55|TIDL_ConvolutionLayer|537|0|1|1| 54xxxxxxx | 55|114402020 |114402020 |5760000 |56|TIDL_ConvolutionLayer|741|0|1|1| 55xxxxxxx | 56|114402020 |12402020 | 138240000 |57|TIDL_EltWiseLayer|540|0|2|1| 5356xxxxxx | 57|12402020 |12402020 |96000 |58|TIDL_ConvolutionLayer|543|0|1|1| 57xxxxxxx | 58|12402020 |114402020 | 138816000 |59|TIDL_ConvolutionLayer|546|0|1|1| 58xxxxxxx | 59|114402020 |114402020 |5760000 |60|TIDL_ConvolutionLayer|750|0|1|1| 59xxxxxxx | 60|114402020 |12402020 | 138240000 |61|TIDL_EltWiseLayer|549|0|2|1| 5760xxxxxx | 61|12402020 |12402020 |96000 |62|TIDL_ConvolutionLayer|552|0|1|1| 61xxxxxxx | 62|12402020 |114402020 | 138816000 |63|TIDL_ConvolutionLayer|555|0|1|1| 62xxxxxxx | 63|114402020 |114402020 |5760000 |64|TIDL_ConvolutionLayer|759|0|1|1| 63xxxxxxx | 64|114402020 |14802020 | 276480000 |65|TIDL_ConvolutionLayer|560|0|1|1| 64xxxxxxx | 65|14802020 |15122020 |98508800 |66|TIDL_ConvolutionLayer|loc_1|0|1|1| 64xxxxxxx | 66|14802020 |1722020 |13824000 |67|TIDL_ConvolutionLayer|conf_1|0|1|1| 64xxxxxxx | 67|14802020 |1182020 |3456000 |68|TIDL_ConvolutionLayer|563|0|1|1| 65xxxxxxx | 68|15122020 |15121010 |512000 |69|TIDL_ConvolutionLayer|566|0|1|1| 68xxxxxxx | 69|15121010 |15121010 |26265600 |70|TIDL_ConvolutionLayer|569|0|1|1| 69xxxxxxx | 70|15121010 |15121010 |26265600 |71|TIDL_ConvolutionLayer|loc_2|0|1|1| 69xxxxxxx | 71|15121010 |1721010 |3686400 |72|TIDL_ConvolutionLayer|conf_2|0|1|1| 69xxxxxxx | 72|15121010 |1181010 |921600 |73|TIDL_ConvolutionLayer|572|0|1|1| 70xxxxxxx | 73|15121010 |151255 |128000 |74|TIDL_ConvolutionLayer|575|0|1|1| 73xxxxxxx | 74|151255 |151255 |6566400 |75|TIDL_ConvolutionLayer|578|0|1|1| 74xxxxxxx | 75|151255 |1102455 |13132800 |76|TIDL_ConvolutionLayer|loc_3|0|1|1| 74xxxxxxx | 76|151255 |17255 |921600 |77|TIDL_ConvolutionLayer|conf_3|0|1|1| 74xxxxxxx | 77|151255 |11855 |230400 |78|TIDL_ConvolutionLayer|581|0|1|1| 75xxxxxxx | 78|1102455 |1102433 |92160 |79|TIDL_ConvolutionLayer|584|0|1|1| 78xxxxxxx | 79|1102433 |1102433 |9446400 |80|TIDL_ConvolutionLayer|587|0|1|1| 79xxxxxxx | 80|1102433 |1102433 |9446400 |81|TIDL_ConvolutionLayer|loc_4|0|1|1| 79xxxxxxx | 81|1102433 |17233 |663552 |82|TIDL_ConvolutionLayer|conf_4|0|1|1| 79xxxxxxx | 82|1102433 |11833 |165888 |83|TIDL_ConvolutionLayer|590|0|1|1| 80xxxxxxx | 83|1102433 |1102422 |40960 |84|TIDL_ConvolutionLayer|593|0|1|1| 83xxxxxxx | 84|1102422 |1102422 |4198400 |85|TIDL_ConvolutionLayer|loc_5|0|1|1| 84xxxxxxx | 85|1102422 |17222 |294912 |86|TIDL_ConvolutionLayer|conf_5|0|1|1| 84xxxxxxx | 86|1102422 |11822 |73728 |87|TIDL_DetectionOutputLayer|output_det|0|12|1| 506671768185516772778286 | 87|1484040 |1111905 |0 |88|TIDL_DataLayer|output|0|1|-1| 87xxxxxxx |150|1111905 |1111905 |0 |
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Total Giga Macs : 5.5179
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

,

xingye zheng:

tidl_net.bin.txt

,

xingye zheng:

两个文件够吗

,

Cherry Zhou:

您好,

能不能再分享下出现错误前的log?

Processing config file #0 : /home/zet/byd/zhangtiansheng/8_6/8bit_cali13/det_test_mix.txt.qunat_stats_config.txtFreeing memory for user provided Net———————– TIDL Process with REF_ONLY FLOW ————————

# 0 . .. T 6885.71 Error at line: 1313 : in file /opt/Jenkins_OUT/workspace/J7_TIDL_Parent/c7x-mma-tidl/ti_dl/release/tidl_j721e_08_06_00_10/ti_dl/test/src/pc_linux/../tidl_rt.c, of function : TIDLRT_invokeInvalid Error Type!

,

?? ?:

import_debuglog.txt

,

Cherry Zhou:

您试下使 postProType=0,看下是否还会有这个错误。

,

xingye zheng:

试了,还是报这个错误

,

xingye zheng:

git.ti.com/cgit/jacinto-ai/pytorch-ssds-keypoints/  这个是官方提供的项目,和我们这个模型是一样的,同样有这个问题,该项目在8.2上是可以正常转化的,在8.6上面是有这个错误的,你们自己也可以用我给的配置文件尝试一下

,

Cherry Zhou:

就是完全相同的项目,然后在8.2可以正常用,8.6 有问题是吗?我们向工程师反馈下。

,

xingye zheng:

是的,而且还有别的目标检测的 regnet+ssd 也有这个问题

,

Cherry Zhou:

您刚刚说,只有在calibration=13时才有这个问题对吗?其他数值时都没有问题?

,

xingye zheng:

是的,但是别的目标检测在升级到8.6时calibration=13也有问题, calibration=13这种模式的效果是我们所需要的,关于目标检测调优,你们有推荐吗,哪些配置参数精度是比较高的

,

xingye zheng:

我还试过calibration=13有问题是从8.5开始才有的,8.4(包括)之前是没有的,8.5 tidl工程开始有了较大的变化

,

Cherry Zhou:

好的我们跟进给工程师了。

,

xingye zheng:

为什么calibration=13在以前的版本可以,在8.5以后的版本就不行了呢,能给个说法吗

,

xingye zheng:

calibration=7效果不行,满足不了我们的精度需求

,

xingye zheng:

  通过设置debugtracelevel=2所定位到的这个问题该怎么解决呢,我们项目挺紧张的,calibration=13这种模式是我们经过很多尝试才得到的,对我们很重要的,请务必支持,给出解决方案,谢谢。

,

Cherry Zhou:

了解,我们跟进给工程师看下。

,

xingye zheng:

今天你们上班吗

,

Cherry Zhou:

看起来可能是对象检测后处理层中的某些内存分配失败了。工程师正在和团队检查这个问题,会尽快给您答复。

,

xingye zheng:

可以催促一下吗,我们这边挺赶的,thanks

,

xingye zheng:

git.ti.com/cgit/jacinto-ai/pytorch-ssds-keypoints/ 的onnx和prototext 这里上传失败了,请给个邮箱地址

,

xingye zheng:

,

Cherry Zhou:

您好,不太行,我没有办法使用网盘提取您的文件,这样,我通过邮件联系您,您把文件再发我。

,

xingye zheng:

邮件已经回复

,

Cherry Zhou:

还是有问题,您给的第一个文件我可以打开,第二个文件您能不能试一下换个文件格式发我?

,

xingye zheng:

onnx需要专业的工具啊,你在浏览器输入netron.app 然后再打开试试

,

xingye zheng:

能不能加个微信联系沟通比较方便

,

Cherry Zhou:

您好,已经转发给工程师了。

,

xingye zheng:

import的config文件,onnx, prototext、还有开启debugTraceLevel = 2的log都上传了,请看本帖子上面的信息

,

xingye zheng:

import的配置文件就是最上面的那张图啊,import就报错了呀,那个工程师不应该看不出来啊

,

xingye zheng:

你要把onnx和prototext 一起发,只发onnx不行的

,

xingye zheng:

这个问题一直还没消息,请代为沟通,谢谢

,

Cherry Zhou:

您所有的文件我们已经通过邮件发给工程师了,我看到您在英文论坛也跟进了,如果之后还有需要支持的地方您可以随时reopen该贴或者发布一个新的帖子。

赞(0)
未经允许不得转载:TI中文支持网 » TDA4VM: calibration=13在08_02 tidl_import上转换成功 升级换成08_06 tidl_import上转换不成功
分享到: 更多 (0)