feat(am62dx-edgeai): Add AM62DX EdgeAI documentation by PrathamTI · Pull Request #818 · TexasInstruments/processor-sdk-doc

PrathamTI · 2026-10-26T12:16:02Z

This PR adds documentation for the AM62DX EdgeAI software stack — audio deep-learning inference (speech enhancement and audio classification) running on the C7x DSP via TVM .

cshilwant · 2026-10-27T06:14:05Z

Are 17 granular commits needed? If not, please plan to squash them

cshilwant · 2026-10-27T06:19:58Z

+
+|__SDK_DOWNLOAD_URL__|
+
+Look for the file: ``tisdk-edgeai-image-am62dxx-evm.rootfs.wic.xz``


#816 (comment)

Could you please go through all the comments on the draft PR and make sure they are addressed? Ideally, the same draft PR should have been opened for review, which would make it easier to verify that all comments have been handled.

We had some changes which modified build.yml and deploy.yml file to run github actions on the fork (And the generated github.io link was shared amongst other members). Hence we created a separate PR with clean commits.

cshilwant · 2026-10-27T06:22:16Z

+
+.. _edgeai-introduction:
+
+########################


#816 (comment)

cshilwant · 2026-10-27T06:24:02Z

+At run time on the AM62D, the TVM runtime loads the artifact and dispatches the
+offloaded subgraphs to TIDL-RT on the C7\ |trade| NPU.
+
+.. figure:: ../../../images/audioai/tvm_tidl_flow.svg


why relative path is required here?
/images/audioai/tvm_tidl_flow.svg should work I believe

cshilwant

Please review all comments from the draft PR and ensure they are addressed in this PR. Several comments from the draft PR are still unresolved here.

PrathamTI · 2026-10-27T13:45:47Z

Are 17 granular commits needed? If not, please plan to squash them

Squashed into a single commit

cshilwant

Run your commit by Claude against our CONTRIBUTING guide [0] to spot potential issues. Go through its findings, separate real non-compliance from false alarms, and fix what's actually broken.

[0] - https://github.com/TexasInstruments/processor-sdk-doc/blob/master/CONTRIBUTING.md

cshilwant · 2026-10-27T21:50:11Z

+
+.. code-block:: bash
+
+   sudo minicom -D /dev/ttyUSB1 -c on


Please highlight baudrate in command itself

Added baudrate: https://github.com/TexasInstruments/processor-sdk-doc/pull/818/changes#diff-899e4a5652d665d36b2644fedbd9e3b2b873e846f8239ab7ea1f04ad40c76f66R149

cshilwant · 2026-10-27T21:53:13Z

+
+   host# ssh root@<EVM_IP_ADDRESS>
+
+If using Ubuntu 22.04, add the following to ``~/.ssh/config``:


use role file - https://www.sphinx-doc.org/en/master/usage/restructuredtext/roles.html#role-file

Note - this comment is not just for this section. Please update it for all such file & directory references in all such references in the PR

I have addressed this as well. The role-file was missing in multiple files, thanks for identifying this issue.

cshilwant · 2026-10-27T22:22:00Z

+   root@am62dxx-evm:~# git clone https://github.com/TexasInstruments/rpmsg-dma.git -b main
+   root@am62dxx-evm:~# cd rpmsg-dma
+
+The example is under ``example/edge-ai/`` and contains:


https://github.com/TexasInstruments/processor-sdk-doc/pull/818/changes#r4394614169

cshilwant · 2026-10-27T22:26:50Z

+   * - GTCRN
+     - Speech enhancement (VoiceBank-DEMAND-16k)
+     - ``ase-0100``
+     - Arm core (ONNX runtime, FP32)


is this correct that GTCRN runs entirely on Arm that too with ONNX-RT ?
@PrathamTI / @v-singh1

let's drop all references of GTCRN from SDK docs as the AM62D SDK intent is to support network execution on MMA &/OR C7x with TVM-RT only and no fallback to Arm Cortex A53.

There are 7 references which needs to be dropped.

Removed all the GTCRN references

Akshay-Abdar · 2026-10-28T09:10:42Z

+- BOOTMODE [8:15] (SW3) = 0100 0000
+- BOOTMODE [0:7] (SW2) = 1100 0010
+
+.. figure:: /images/am62dx_boot_pin.png


Just for confirmation. I can see SW1 in the image (the board image displayed). Can you check if it is SW1? The other bootmode switch naming is also not clear

Akshay-Abdar · 2026-10-28T09:13:10Z

+a network, the web portal can be accessed from any device on the same network without
+any additional setup.
+
+The portal showcases the three core EdgeAI pipelines — home screen, speech


home screen as EdgeAI pipeline?

Removed this statement.

Akshay-Abdar · 2026-10-28T09:22:14Z

Any release notes section planned to be added somewhere at the bottom?

Akshay-Abdar · 2026-10-28T09:45:33Z

+
+Connect the components to the AM62D EVM as shown below:
+
+.. figure:: /images/am62dx_board_config.png


In this image, there are 3 unmarked arrows (the arrows are pointing towards a component) without any labelling. Kindly mark some label and also the image is not clear

@Akshay-Abdar

Thanks for identifying this issue. @paresh-bhagat is working on adding the image with proper labels

@Akshay-Abdar

We have updated the image: https://prathamti.github.io/processor-sdk-doc/processor-sdk-edgeai-AM62DX/esd/docs/master/devices/AM62DX/edgeai/evm_hardware_setup.html

Akshay-Abdar · 2026-10-28T09:47:47Z

+For **Windows** use Tera Term:
+
+https://learn.sparkfun.com/tutorials/terminal-basics/tera-term-windows


Do we need this part since we haven't tested this for windows

Akshay-Abdar · 2026-10-28T09:53:19Z

+On-target measured latency and RTF for all four models are **TBD** - to be
+filled in from the Model Zoo inference scripts on the AM62D EVM:
+
+.. list-table::
+   :header-rows: 1
+   :widths: 20 20 30 30
+
+   * - Model
+     - On target
+     - Measured latency
+     - RTF
+   * - VGGish11
+     - C7\ |trade| NPU
+     - TBD
+     - - (clip classifier)
+   * - YAMNet
+     - C7\ |trade| NPU
+     - TBD
+     - - (clip classifier)
+   * - GCRN
+     - C7\ |trade| NPU
+     - TBD
+     - TBD
+   * - GTCRN
+     - Arm core


Small question. If its TBD do we need to mention this part in the docs? Also add the arm core GTCRN model if the section will be included

PrathamTI · 2026-10-28T13:15:10Z

Any release notes section planned to be added somewhere at the bottom?

Yes we will create a separate section for release notes, but its not planned as of yet.

Adds the full EdgeAI documentation set for AM62DX covering audio deep-learning inference (speech enhancement and audio classification) on the C7x DSP via TVM + TIDL. The guide spans platform overview, hardware setup, the out-of-box web GUI demo, the C++ inference application, GStreamer audio pipeline development, the model workflow (zoo, compile/evaluate, inspect), integrating a new model, and performance benchmarking. Co-authored-by: JuneChul Roh <jroh@ti.com> Co-authored-by: Paresh Bhagat <p-bhagat@ti.com> Signed-off-by: Pratham Deshmukh <p-deshmukh@ti.com>

PrathamTI requested review from StaticRocket, VeeruPrudhvi, cshilwant, gehariprasath, jeevantelukula, praneethbajjuri and uditkumarti as code owners October 26, 2026 12:16

PrathamTI changed the title ~~Am62d docs~~ feat(am62dx-edgeai): Add AM62DX EdgeAI documentation Oct 26, 2026

PrathamTI force-pushed the am62d-docs branch from 6aa1a39 to 7d44da3 Compare October 27, 2026 05:29

cshilwant reviewed Oct 27, 2026

View reviewed changes

cshilwant requested changes Oct 27, 2026

View reviewed changes

PrathamTI force-pushed the am62d-docs branch 3 times, most recently from 823790b to 569b338 Compare October 27, 2026 13:40

cshilwant requested changes Oct 27, 2026

View reviewed changes

Akshay-Abdar reviewed Oct 28, 2026

View reviewed changes

PrathamTI force-pushed the am62d-docs branch 2 times, most recently from a0a0d2e to 599dcd1 Compare October 28, 2026 12:41

paresh-bhagat force-pushed the am62d-docs branch from 599dcd1 to 1ac9d42 Compare October 29, 2026 04:51

PrathamTI force-pushed the am62d-docs branch from 1ac9d42 to 6c560ae Compare October 29, 2026 05:22

PrathamTI force-pushed the am62d-docs branch from 6c560ae to 36ff99f Compare October 29, 2026 08:31


		\|__SDK_DOWNLOAD_URL__\|

		Look for the file: ``tisdk-edgeai-image-am62dxx-evm.rootfs.wic.xz``


		host# ssh root@<EVM_IP_ADDRESS>

		If using Ubuntu 22.04, add the following to ``~/.ssh/config``:


		Connect the components to the AM62D EVM as shown below:

		.. figure:: /images/am62dx_board_config.png

		For Windows use Tera Term:

		https://learn.sparkfun.com/tutorials/terminal-basics/tera-term-windows


		.. _edgeai-introduction:

		########################

Conversation

PrathamTI commented Oct 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

cshilwant commented Oct 27, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cshilwant left a comment

Choose a reason for hiding this comment

Uh oh!

PrathamTI commented Oct 27, 2026

Uh oh!

cshilwant left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Akshay-Abdar commented Oct 28, 2026

Uh oh!

Akshay-Abdar Oct 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

PrathamTI commented Oct 28, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

PrathamTI commented Oct 26, 2026 •

edited

Loading

Akshay-Abdar Oct 28, 2026 •

edited

Loading