PROCESSOR-SDK-AM62A: PROCESSOR-SDK-AM62A

chao chen

1. AM62A seems to lack PROCESSOR-SDK-RTOS-J721E (environment SDK that supports onnx/tensorflow model transformation and deployment to TI hardware) similar to J721E.
2. Learned from the forum that FIRMWARE-BUILDER-AM62A exists, not sure whether this is the environment SDK for model transformation and deployment.
3. Is the operation of FIRMWARE-BUILDER-AM62A similar to that of PROCESSOR-SDK-RTOS-J721E?
Can I directly use PROCESSOR-SDK-RTOS-J721E and set the SOC compilation to AM62A for model transformation and deployment?
Someone who can answer questions about whether AM62A can be transformed and deployed using a development process similar to J721E.

6 个月前

0 Vivian Gao 6 个月前

TI__Mastermind 36650 points

可以参考下方三个链接的回复。

1）(+) PROCESSOR-SDK-AM62A: Tensorflow Lite operations and runtime support in TIDL - Processors forum - Processors - TI E2E support forums

2）(+) PROCESSOR-SDK-AM62A: Running AI model on the chip. - Processors forum - Processors - TI E2E support forums

3）(+) [FAQ] EDGE-AI-STUDIO: Is SDK version important for Edge AI and TI Deep learning (TIDL) with C7x on AM6xA SoCs [AM62A, AM67A, AM68A, AM68PA, AM69A]? - Processors forum - Processors - TI E2E support forums

0 chao chen 6 个月前回复 Vivian Gao

Prodigy 10 points

Thanks! for reading the following link, I seem to understand it, but I still lack the tidl-tools library and firmware builder, which seems to be under application. Could you please pass it? Are the tidl_model_import.out and tidl_model*.so libraries also compiled based on edgeai-tidl-tools? I'm still conflicted about it.

0 Vivian Gao 6 个月前回复 chao chen

TI__Mastermind 36650 points

这个FAQ里面有您需要的资源。

(+) [FAQ] [参考译文] [常见问题解答] EDGE-AI-STUDIO：AM6xA SoC 的边缘 AI 资源 - 处理器（参考译文帖）(Read Only) - 处理器（参考译文帖） - E2E 设计支持

0 chao chen 6 个月前回复 Vivian Gao

Prodigy 10 points

thanks! I seem to know how to convert this model to the AM62A platform, but I have a problem that the operator is not supported. A721A and tdA4 can be transferred normally before, how do I deal with it? What documents do I need to provide to help you troubleshoot this problem?

0 chao chen 6 个月前回复 chao chen

Prodigy 10 points

(am62a) docker@8084c2d7f6f9:/ti-am62a/tools/AM62
A/tidl_tools$ bash run.sh 
 00eyeclose Start to transform 
 01eyeclose Start to transform 
========================= [Model Compilation Started] =========================

Model compilation will perform the following stages:
1. Parsing
2. Graph Optimization
3. Quantization & Calibration
4. Memory Planning

============================== [Version Summary] ==============================

-------------------------------------------------------------------------------
|          TIDL Tools Version          |              10_01_04_00             |
-------------------------------------------------------------------------------
|         C7x Firmware Version         |              10_01_00_01             |
-------------------------------------------------------------------------------

ONNX model (Proto) file      : ./model/eye.onnx  
TIDL network file            : ./model_out/eye/tidl_net_eye_am62a_Input96X96_int8_mix_Batch1.bin  
TIDL IO info file            : ./model_out/eye/tidl_io_eye_am62a_Input96X96_int8_mix_Batch1_  
Current ONNX OpSet version   : 9  
============================ [Optimization started] ============================

----------------------------- Optimization Summary -----------------------------
--------------------------------------------------------------------------------
|         Layer         | Nodes before optimization | Nodes after optimization |
--------------------------------------------------------------------------------
| TIDL_BatchNormLayer   |                         0 |                        1 |
| TIDL_EltWiseLayer     |                        10 |                       10 |
| TIDL_CLipLayer        |                        59 |                        0 |
| TIDL_SigmoidLayer     |                         1 |                        0 |
| TIDL_ConcatLayer      |                         3 |                        3 |
| TIDL_ResizeLayer      |                         3 |                        3 |
| TIDL_ConvolutionLayer |                        61 |                       61 |
| TIDL_FlattenLayer     |                         1 |                        0 |
| TIDL_PoolingLayer     |                         7 |                        7 |
--------------------------------------------------------------------------------

=========================== [Optimization completed] ===========================


-------- Running Calibration in Float Mode to Collect Tensor Statistics --------
[=============================================================================] 100 %

------------------ Fixed-point Calibration Iteration [1 / 1]: ------------------
[=============================================================================] 100 %

==================== [Quantization & Calibration Completed] ====================

========================== [Memory Planning Started] ==========================


------------------------- Network Compiler Traces ------------------------------
Successful Memory Allocation
Successful Workload Creation

========================= [Memory Planning Completed] =========================

Rerunning network compiler...
========================== [Memory Planning Started] ==========================


------------------------- Network Compiler Traces ------------------------------
Successful Memory Allocation
Successful Workload Creation

========================= [Memory Planning Completed] =========================

[TIDL Import]  UNSUPPORTED: Allowlisting : Layer name - Upsample_11 : Resize is only supported along width and height axis -- [tidlAllowlistingConstraints/tidl_constraint.cpp, 85]
[TIDL Import]  UNSUPPORTED: Allowlisting : Layer name - Upsample_14 : Resize is only supported along width and height axis -- [tidlAllowlistingConstraints/tidl_constraint.cpp, 85]
[TIDL Import]  UNSUPPORTED: Allowlisting : Layer name - Upsample_100 : Resize is only supported along width and height axis -- [tidlAllowlistingConstraints/tidl_constraint.cpp, 85]

============== Subgraph Compiled with   0 Warnings and   3 Errors ==============


[TIDL Import]  WARNING: 
*****************************************************************************
*     TIDL Import returned with error, model artifacts are not reliable     *
*                   Please resolve the compilation issues                   *
*****************************************************************************

[TIDL Import]  FATAL ERROR: Quantization/NC Pass/Model checker failed -- [tidl_import_main.cpp, 510]
[TIDL Import] Aborting

This is the overall log

0 Vivian Gao 6 个月前回复 chao chen

TI__Mastermind 36650 points

新问题请重新发帖咨询。

处理器

处理器论坛

PROCESSOR-SDK-AM62A: PROCESSOR-SDK-AM62A