Differences

This shows you the differences between two versions of the page.

--- lwc:hardware:jetson [2024/09/25 22:24] – John Harrison
+++ lwc:hardware:jetson [2025/03/05 17:12] (current) – [misc] John Harrison
@@ Line 2: / Line 2: @@
   * [[https://wiki.seeedstudio.com/reComputer_Industrial_Getting_Started/|getting started]]
   * [[https://github.com/Seeed-Projects/jetson-examples?tab=readme-ov-file|Jetson examples]]
-  * [[https://developer.nvidia.com/embedded/learn/jetson-ai-certification-programs#course_outline|nvidia certification]]
+  * [[https://developer.nvidia.com/embedded/learn/jetson-ai-certification-programs#course_outline|nvidia certification]] (no longer available)
+  * [[https://www.jetson-ai-lab.com/|Jetson AI lab]]
+  * [[https://learn.nvidia.com/courses/course-detail?course_id=course-v1:DLI+S-RX-02+V2|Getting Started with AI on Jetson Nano]]
+==== seeedstudio reComputer 3011 and 4012 (and probably others) ====
+  * do not try to update with seeedstudio's firmware. It's not updateable after that
+  * instead use sdkmanager from nvidia:
+    * do manual install. Automatic will try to update the seeedstudio stuff which will fail
+    * manual install requires forced recovery mode which requires a USB connection
+      * to do forced recovery with unit unplugged:
+        * use a paper clip to press the recovery button
+        * while pressing the recovery button apply power
+    * attach USB cable to "device port." For USB the cable used matters. I had luck using usbc <--> usbc. Using usba <--> usbc failed
+    * 1/2 way through the install process, sdkmanager will want to reconfirm/switch how host is communicating to the target. At this point the target is booted into the new OS
+    * at that point I switched to ethernet as both units were connected to my router. I'm not sure if sticking with usb might have worked or would have been faster
+==== Installing pyTorch ====
+//corrected from: [[https://docs.nvidia.com/deeplearning/frameworks/install-pytorch-jetson-platform/index.html]])//
+  * install ''jetpack''. (This is not the same as installing the OS for Jetson): ''sudo apt install nvidia-jetpack''
+  * get version of CUDA: ''nvcc --version''
+  * install ''cusparselt'':
+<code>
+wget raw.githubusercontent.com/pytorch/pytorch/5c6af2b583709f6176898c017424dc9981023c28/.ci/docker/common/install_cusparselt.sh
+sudo -s # set CUDA version and install as root
+export CUDA_VERSION=12.1 # as an example. Use the version returned from ''nvcc --version''
+bash ./install_cusparselt.sh
+</code>
+  * install correct numpy version: ''sudo pip install numpy==1.26.1''
+  * download correct version of pytorch for jetpack version. Search for it at [[https://developer.download.nvidia.cn/compute/redist/jp/]]
+<code>
+sudo pip3 install --no-cache http://<URL TO PYTORCH WHL>
+</code>
+==== misc ====
+  * [[https://elinux.org/Jetson_Zoo#ONNX_Runtime|onnxruntime for Jetson]]
+  * if ''onnxruntime'' fails with message ''Invalid argument. Specify the number of threads explicitly'' fix with one-time cli command: ''sudo nvpmodel -m 0'' which forces the computer to with no throttling
+  * ''tegrastats'' might show load on GPU?
+    * ''tegrastats | cut -f 14 -d " "'' shows % load on GPU
+  * you can get rid of the warning ''System Throttled due to Over-current'' by editing ''/usr/share/nvpmodel_indicator/nvpmodel_indicator.py''