Cannot Compile AlexNet, Inception_V1(GoogleNet), VGG-16, VGG-19, SqueezeNet Caffe Models

[Hassan313]

@prasshantg Hi Prashant these are the errors for the below Caffe models:

All the Caffe models are from the below website:

https://github.com/microsoft/MMdnn/tree/master/mmdnn/models

All the caffe models are compiled with the below command (the below is an exmaple):

./nvdla_compiler --prototxt ~/CaffeModels/AlexNet/Alexnet.prototxt --caffemodel ~/CaffeModels/AlexNet/bvlc_alexnet.caffemodel -o . --profile fast-math --cprecision int8 --configtarget nv_small --quantizationMode per-filter --batch 1 --informat nhwc

---

AlexNet:

---

creating new wisdom context... opening wisdom context... parsing caffe network... libnvdla<3> mark prob Marking total 1 outputs initialize all tensors with const scaling factors of 127... attaching parsed network to the wisdom... compiling profile "fast-math"... config "nv_small"... libnvdla<2> Overriding input multibatch size from 10 to 1 libnvdla<2> Prototxt #chnls (C = 3) != Profile #chnls for input (NVDLA_IMG_A8B8G8R8: C = 4). Preferring #chnls from Profile for compiling. (DLA) Error 0x0000000b: Couldn't narrow down to suitable SF for e-2/tsd-27 (propagating from engine-ast/EngineEdge.cpp, function determineSurfaceFormat(), line 278) (DLA) Error 0x0000000b: (propagating from engine-ast/EngineGraph.cpp, function registerAllSurfaces(), line 966) (DLA) Error 0x0000000b: (propagating from Compiler.cpp, function registerBuffers(), line 1586) (DLA) Error 0x00000008: failed compilation phase: registerBuffers (propagating from Compiler.cpp, function compileInternal(), line 506) (DLA) Error 0x00000008: (propagating from Compiler.cpp, function compileInternal(), line 423) (DLA) Error 0x00000008: (propagating from Compiler.cpp, function compile(), line 372) (DLA) Error 0x00000008: (propagating from CompileTest.cpp, function compileProfile(), line 66) (DLA) Error 0x00000008: (propagating from ParseTest.cpp, function parseAndCompile(), line 252) (DLA) Error 0x00000008: (propagating from main.cpp, function launchTest(), line 111)

---

Inception V1

---

creating new wisdom context... opening wisdom context... parsing caffe network... libnvdla<3> mark prob Marking total 1 outputs initialize all tensors with const scaling factors of 127... attaching parsed network to the wisdom... compiling profile "fast-math"... config "nv_small"... libnvdla<2> Overriding input multibatch size from 10 to 1 libnvdla<2> Prototxt #chnls (C = 3) != Profile #chnls for input (NVDLA_IMG_A8B8G8R8: C = 4). Preferring #chnls from Profile for compiling. (DLA) Error 0x0000000b: Couldn't narrow down to suitable SF for e-3/tsd-178 (propagating from engine-ast/EngineEdge.cpp, function determineSurfaceFormat(), line 278) (DLA) Error 0x0000000b: (propagating from engine-ast/EngineGraph.cpp, function registerAllSurfaces(), line 966) (DLA) Error 0x0000000b: (propagating from Compiler.cpp, function registerBuffers(), line 1586) (DLA) Error 0x00000008: failed compilation phase: registerBuffers (propagating from Compiler.cpp, function compileInternal(), line 506) (DLA) Error 0x00000008: (propagating from Compiler.cpp, function compileInternal(), line 423) (DLA) Error 0x00000008: (propagating from Compiler.cpp, function compile(), line 372) (DLA) Error 0x00000008: (propagating from CompileTest.cpp, function compileProfile(), line 66) (DLA) Error 0x00000008: (propagating from ParseTest.cpp, function parseAndCompile(), line 252) (DLA) Error 0x00000008: (propagating from main.cpp, function launchTest(), line 111)

---

SquueezNet

---

creating new wisdom context... opening wisdom context... parsing caffe network... libnvdla<3> mark prob Marking total 1 outputs initialize all tensors with const scaling factors of 127... attaching parsed network to the wisdom... compiling profile "fast-math"... config "nv_small"... libnvdla<2> Overriding input multibatch size from 10 to 1 libnvdla<2> Prototxt #chnls (C = 3) != Profile #chnls for input (NVDLA_IMG_A8B8G8R8: C = 4). Preferring #chnls from Profile for compiling.

It will stick here ...

---

VGG16

---

./nvdla_compiler --prototxt /home/home2/students/afzalikusha/CaffeModels/VGG16/deploy.prototxt --caffemodel /home/home2/students/afzalikusha/CaffeModels/VGG16/VGG_ILSVRC_16_layers.caffemodel -o . --profile fast-math --cprecision int8 --configtarget nv_small --quantizationMode per-filter --batch 1 --informat nhwc creating new wisdom context... opening wisdom context... parsing caffe network... Marking total 0 outputs (DLA) Error 0x0000000b: Unable to identify outputs for the network: 0 (in ParseTest.cpp, function parseCaffeNetwork(), line 166) (DLA) Error 0x0000000b: (propagating from ParseTest.cpp, function parseAndCompile(), line 247) (DLA) Error 0x0000000b: (propagating from main.cpp, function launchTest(), line 111)

---

VGG19

---

creating new wisdom context... opening wisdom context... parsing caffe network... Marking total 0 outputs (DLA) Error 0x0000000b: Unable to identify outputs for the network: 0 (in ParseTest.cpp, function parseCaffeNetwork(), line 166) (DLA) Error 0x0000000b: (propagating from ParseTest.cpp, function parseAndCompile(), line 247) (DLA) Error 0x0000000b: (propagating from main.cpp, function launchTest(), line 111)

So can you kindly help me to have more benchmarks for the NVDLA accelerator for nv_small (int8)?

Thank you very much.

[Uspark-dev]

this DLA error is due to missing drm.ko and open_dla .ko please install them and then you results

[Hassan313]

@Uspark-dev Hi, these are compilation errors.

Your below statement is related to the virtual platform! _this DLA error is due to missing drm.ko and open_dla .ko

[Uspark-dev]

@Hassan313 you are working on FPGA platfom on AMI?

[Hassan313]

@Uspark-dev I tried to work, however my experience was unsuccessful. I am working with CPUs right now.

[Uspark-dev]

@Hassan313 can i have your email id so that we can discuss in detail

[Hassan313]

@Uspark-dev Sure. afzaliku@usc.edu.

[Jade-Hsu]

@Uspark-dev @Hassan313 I met the same problem, haven't been solved until now. Hope we can discuss together ,and this is my email : xyuic@mail.ustc.edu.cn

[Uspark-dev]

hi @Jade-Hsu i have sent you email regarding your issues please check your inbox.

[Jade-Hsu]

hi @Jade-Hsu i have sent you email regarding your issues please check your inbox.

your email is usman@chungbuk.ac.kr? I have recieved that few days ago.

[Uspark-dev]

yes

[Jade-Hsu]

yes

please check your email inbox, i have sent u my problem.

[Uspark-dev]

i do not receive your email can you please check again you sent it correct address or not

[Jade-Hsu]

i do not receive your email can you please check again you sent it correct address or not

i reply your email so there is no fault. Please sent me email again and i try again. my email id : xyuic@mail.ustc.edu.cn thanks!

[MrHaiCoder]

@ Uspark-dev好的。afzaliku@usc.edu。 I use nv_small vp env,Run the mnist, the core dumped appears, what is the reason?

[Hassan313]

@MrHaiCoder Hi, I am not sure what is the problem. Sorry.

Maybe, @prasshantg can help you.

[MrHaiCoder]

@MrHaiCoder嗨，我不确定是什么问题。抱歉。

也许@prasshantg可以为您提供帮助。

Thank you for your reply, but it seems that prassantg has not been updated for a long time. I use the following URL to build VP(nv_small) https://github.com/prasshantg/personal

[applefishsky009]

i do not receive your email can you please check again you sent it correct address or not

i reply your email so there is no fault. Please sent me email again and i try again. my email id : xyuic@mail.ustc.edu.cn thanks!

你好，我在编译darknet53的时候遇到了同样的问题，已经邮件发送给你咨询了。目前你有关于解决方案的最新进展吗~

[DressInWhite]

@Hassan313 Hi, Did you fix the problem now, I met the same problem with AlexNet, Could you help me with it? Command: ./nvdla_compiler --prototxt deploy.prototxt --caffemodel bvlc_alexnet.caffemodel -o basic.nvdla error info: compiling profile "fast-math"... config "nv_full"... libnvdla<2> Prototxt #chnls (C = 3) != Profile #chnls for input (NVDLA_IMG_A8B8G8R8: C = 4). Preferring #chnls from Profile for compiling. libnvdla<2> Overriding output multibatch size from 1 to 10 (DLA) Error 0x0000000b: Couldn't narrow down to suitable SF for e-2/tsd-27 (propagating from engine-ast/EngineEdge.cpp, function determineSurfaceFormat(), line 278) (DLA) Error 0x0000000b: (propagating from engine-ast/EngineGraph.cpp, function registerAllSurfaces(), line 966) (DLA) Error 0x0000000b: (propagating from Compiler.cpp, function registerBuffers(), line 1586) (DLA) Error 0x00000008: failed compilation phase: registerBuffers (propagating from Compiler.cpp, function compileInternal(), line 506) (DLA) Error 0x00000008: (propagating from Compiler.cpp, function compileInternal(), line 423) (DLA) Error 0x00000008: (propagating from Compiler.cpp, function compile(), line 372) (DLA) Error 0x00000008: (propagating from CompileTest.cpp, function compileProfile(), line 66) (DLA) Error 0x00000008: (propagating from ParseTest.cpp, function parseAndCompile(), line 252) (DLA) Error 0x00000008: (propagating from main.cpp, function launchTest(), line 111)

[long771]

这是来自QQ邮箱的自动回复邮件。 您好，我是龙欣荣，很高兴收到您的邮件，我将尽快处理。