TDA4VM: WARNING: [TIDL_E_DATAFLOW_INFO_NULL] Network compiler returned with error or didn't executed, this model can only be used on PC/Host emulation mode, it is not expected to work on target/EVM.

Azer Tiop

Part Number: TDA4VM

Tool/software:

Hi,

I am working on TDA4VM EVM for the first time. I have to work offline on my host PC (Ubuntu 22.04) with Python 3.10 and I'm using the latest version of SDK 10_00_00_05, latest version of EDGEAI_TIDL_TOOLS 10_00_08_00.

I followed the steps from the Vision Apps tutorial:

https://software-dl.ti.com/jacinto7/esd/processor-sdk-rtos-jacinto7/10_00_00_05/exports/docs/vision_apps/docs/user_guide/index.html

I want to compile a model : ssd_mobilenet_v2_300

However, I have the following error:

 Preliminary number of subgraphs:1 , 31 nodes delegated out of 31 nodes 
 

 ************** Frame index 1 : Running float import ************* 
****************************************************
**                ALL MODEL CHECK PASSED          **
****************************************************

The soft limit is 2048
The hard limit is 2048
MEM: Init ... !!!
MEM: Init ... Done !!!
 0.0s:  VX_ZONE_INIT:Enabled
 0.34s:  VX_ZONE_ERROR:Enabled
 0.36s:  VX_ZONE_WARNING:Enabled
 0.2159s:  VX_ZONE_INIT:[tivxInit:185] Initialization Done !!!
 25%|██▌       | 1/4 [00:00<00:02,  1.34it/s]

 ************ Frame index 1 : Running float inference **************** 
 50%|█████     | 2/4 [00:01<00:01,  1.34it/s]

 ************ Frame index 2 : Running float inference **************** 
 75%|███████▌  | 3/4 [00:02<00:00,  1.35it/s]

 ************ Frame index 3 : Running float inference **************** 

 ************ Frame index 4 : Running fixed point mode for calibration **************** 
Empty prototxt path, running calibration

~~~~~Running TIDL in PC emulation mode to collect Activations range for each layer~~~~~

Processing config file #0 : /home/developer/Documents/developpement/HelloWorld_PC_Cloud_TDA4VM_SK_EVM_v2/custom-artifacts_mobilenet_v1_1.0_224_quant/tempDir/87_tidl_io_.qunat_stats_config.txt 
 Freeing memory for user provided Net
 ----------------------- TIDL Process with REF_ONLY FLOW ------------------------

#    0 . .. T     747.07  .... ..... ... .... .....
#    1 . .. T     739.17  .... ..... ... .... .....
#    2 . .. T     748.66  .... ..... ... .... .....
#    3 . .. T     730.08  .... ..... ... .... .....
 
 
 *****************   Calibration iteration number 0 started ************************ 
 
 
 
Empty prototxt path, running calibration

~~~~~Running TIDL in PC emulation mode to collect Activations range for each layer~~~~~
...
 
 
 
Empty prototxt path, running calibration
Output is truncated. View as a scrollable element or open in a text editor. Adjust cell output settings...
Segmentation fault (core dumped)
100%|██████████| 4/4 [00:18<00:00,  4.52s/it]

------------------ Network Compiler Traces -----------------------------
Could not open /home/developer/Documents/developpement/HelloWorld_PC_Cloud_TDA4VM_SK_EVM_v2/custom-artifacts_mobilenet_v1_1.0_224_quant/tempDir/87_tidl_net/perfSimInfo.bin
WARNING: [TIDL_E_DATAFLOW_INFO_NULL] Network compiler returned with error or didn't executed, this model can only be used on PC/Host emulation mode, it is not expected to work on target/EVM.
****************************************************
**          1 WARNINGS          0 ERRORS          **
****************************************************

 Final number of subgraphs:1 , 31 nodes delegated to accelerator

Here is the folder with every files needed to reproduce the error:

project_compilation.zip

Is there any workaround available to address this error and proceed with the compilation ?

Is it possible to have an example of compilation logs to compare with mine ?

Thank you for your attention to this issue.

Best regards

9 months ago

0 Azer Tiop 9 months ago

Intellectual 350 points

Hi,

I'm still encountering the error, "Segmentation fault (core dumped)", followed by "Network compiler returned with error or didn't executed, this model can only be used on PC/Host emulation mode, it is not expected to work on target/EVM."

I tried to follow the out-of-the-box tutorial for edgeai_tidl_tools : GitHub - TexasInstruments/edgeai-tidl-tools: Edgeai TIDL Tools and Examples - This repository contains Tools and example developed for Deep learning runtime (DLRT) offering provided by TI’s edge AI solutions.

and I’ve attached the logs here:

************************************************

I also attempted to upgrade the C7X/MMA firmware version, but I couldn't find information on how to specify the firmware version we want to install according to the tutorial:

[C7X/MMA TIDL User Guide] https://software-dl.ti.com/jacinto7/esd/processor-sdk-rtos-jacinto7/10_00_00_05/exports/docs/c7x-mma-tidl/ti_dl/docs/user_guide_html/md_tidl_build_instruction.html

************************************************

Has anyone else faced this issue? If so, could you please share your setups?

I’m also including the generated file following the core dump.

ERROR: apport (pid 160049) Fri Dec 13 09:06:57 2024: called for pid 160048, signal 11, core limit 0, dump mode 1

ERROR: apport (pid 160049) Fri Dec 13 09:06:57 2024: executable: /home/developer/edgeai/edgeai-tidl-tools/tidl_tools/ti_cnnperfsim.out (command line "./ti_cnnperfsim.out /home/developer/edgeai/edgeai-tidl-tools/examples/jupyter_notebooks/custom-artifacts/tflite/mobilenetv2/tempDir/264_tidl_io_.perf_sim_config.txt 1601 0 2")

ERROR: apport (pid 160049) Fri Dec 13 09:06:57 2024: executable does not belong to a package, ignoring

Thank you for your assistance!

Best regards,

Azer

0 Wen Li 9 months ago in reply to Azer Tiop

TI__Expert 7061 points

Hi Azer; we will take a look.

Were you able to run anything on the target board ? if you did, how did you flash the board/card?

Thanks and regards

Wen Li

0 Azer Tiop 9 months ago in reply to Wen Li

Intellectual 350 points

Hi Wen Li,

I am able to run inference using ARM on the target board. How did I flash the board card?

After installing PSDK RTOS and Linux according to the Vision Apps tutorial, I followed steps to set up SD card :

umount /dev/sdc1 && umount /dev/sdc2
sudo sdk_builder/scripts/mk-linux-card.sh /dev/sdc
sdk_builder/scripts/install_to_sd_card.sh
cd ${PSDKR_PATH}/sdk_builder
make linux_fs_install_sd

After setting up SD card and using jupyter notebook that I installed on the card, I put my scripts, models and artifacts on board.

Thanks and regards

Azer

0 Azer Tiop 9 months ago in reply to Azer Tiop

Intellectual 350 points

Hi,

The compilation is now working. With the latest version of EDGEAI_TIDL_TOOLS 10_00_08_00, I followed these steps:

1. I rebuilt the firmware package by following the instruction: ./sdk_builder/scripts/setup_psdk_rtos.sh --install_tidl_deps
2. I reinstalled EDGEAI_TIDL_TOOLS 10_00_08_00.
3. I manually installed GCC version 11.3.0 on my PC.

Do you have any explanation for what was generating this error and why these steps resolved it?

Thanks and regards

Azer

0 Wen Li 8 months ago in reply to Azer Tiop

TI__Expert 7061 points

Hi Azer;

I have escalated your ticket to our expert team. They will provide answer to you soon.

Thanks and regards

Wen Li

0 Azer Tiop 7 months ago in reply to Wen Li

Intellectual 350 points

Hi,

Do you have any news?

Regards,

Azer

0 Chris Tsongas 7 months ago in reply to Azer Tiop

TI__Genius 14420 points

Hi Azer,

I am able to run this to completion on 10_00_08_00. Something is not setup right in your system. Attached is a log file. Do an ls -al in your tidl_tools directory and please send it to me.

Regards6560.log.txt,

Chris

Processors

Processors forum

TDA4VM: WARNING: [TIDL_E_DATAFLOW_INFO_NULL] Network compiler returned with error or didn't executed, this model can only be used on PC/Host emulation mode, it is not expected to work on target/EVM.