Part Number: TDA4VM
Hello. I am having difficulty setting up a TI-TDA4VM.
We have models which have been compiled using EdgeAI TIDL tools 08_06_00_03, which according to this compatibility chart https://github.com/TexasInstruments/edgeai-tidl-tools/blob/master/docs/version_compatibility_table.md needs the Processor SDK Linux 08_06_00_11.
I have installed the image from here https://www.ti.com/tool/download/PROCESSOR-SDK-LINUX-SK-TDA4VM/08.06.00.11 using the instructions in the SDK documentation.
Trying to run a model (MobileNetV2, 1x3x224x224) for inference via a pythons script using onnxruntime gives this error:
[C7x_1 ] 2914.102595 s: VX_ZONE_ERROR:[tivxAlgiVisionAllocMem:194] Failed to Allocate memory record 5 @ space = 17 and size = 12365752 !!!
[C7x_1 ] 2914.102623 s: VX_ZONE_ERROR:[tivxAlgiVisionCreate:358] tivxAlgiVisionAllocMem Failed
[C7x_1 ] 2914.102647 s: VX_ZONE_ERROR:[tivxKernelTIDLCreate:720] tivxAlgiVisionCreate returned NULL
[C7x_1 ] 2914.102595 s: VX_ZONE_ERROR:[tivxAlgiVisionAllocMem:194] Failed to Allocate memory record 5 @ space = 17 and size = 12365752 !!!
[C7x_1 ] 2914.102647 s: VX_ZONE_ERROR:[tivxKernelTIDLCreate:720] tivxAlgiVisionCreate returned NULL
2914.123863 s: VX_ZONE_ERROR:[ownContextSendCmd:799] Command ack message returned failure cmd_status: -1
2914.123887 s: VX_ZONE_ERROR:[ownContextSendCmd:835] tivxEventWait() failed.
2914.123898 s: VX_ZONE_ERROR:[ownNodeKernelInit:527] Target kernel, TIVX_CMD_NODE_CREATE failed for node TIDLNode
2914.123906 s: VX_ZONE_ERROR:[ownNodeKernelInit:528] Please be sure the target callbacks have been registered for this core
2914.123911 s: VX_ZONE_ERROR:[ownNodeKernelInit:529] If the target callbacks have been registered, please ensure no errors are occurring within the create callback of this kernel
2914.123919 s: VX_ZONE_ERROR:[ownGraphNodeKernelInit:583] kernel init for node 0, kernel com.ti.tidl:1:1 ... failed !!!
2914.123934 s: VX_ZONE_ERROR:[vxVerifyGraph:2055] Node kernel init failed
2914.123942 s: VX_ZONE_ERROR:[vxVerifyGraph:2109] Graph verify failed
2914.124049 s: VX_ZONE_ERROR:[ownGraphScheduleGraphWrapper:799] graph is not in a state required to be scheduled
2914.124059 s: VX_ZONE_ERROR:[vxProcessGraph:734] schedule graph failed
2914.124064 s: VX_ZONE_ERROR:[vxProcessGraph:739] wait graph failed
ERROR: Running TIDL graph ... Failed !!!
I am trying to understand the cause of this error. I have seen other forum posts (https://e2e.ti.com/support/processors-group/processors/f/processors-forum/1084744/sk-tda4vm-tidl_rt_ovx-error-verifying-tidl-graph-failed and https://e2e.ti.com/support/processors-group/processors/f/processors-forum/860394/tda4-deploy-big-neural-net-on-c7) where a similar error was mentioned. There the suggested root cause was trying to use a model that was too large and the proposed suggestions included rebuilding the operating system with new memory settings. I have a hard time believing that the model I am using here is too large for the default installation using the Linux image from TI, so I assume that there is some other error being made here either in my installation or in the model compilation.
Could you give me some more information about this error? Is this a problem with the installation on the TI-TDA4VM, or could this be a problem with how the model is compiled for TIDL? What is "@ space = 17" referring to here? A location with a 17 byte size, or is 17 some reference to the location?
I did not see this error when downgrading to 08_05_00_03 for the processor Linux SDK and compiling models with EdgeAI TIDL tools 8.5. We will need to use this hardware to support models compiled with the latest versions, so if more recent images for the processor Linux SDK require additional configuration, this would be important to know.
Thank you.
