-
Notifications
You must be signed in to change notification settings - Fork 349
staging: Add tensorflow micro module. #9938
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
|
@lyakh see the Readme for the extra steps needed to link archives to the llext module. We are linking a C and C++17 library to the llext module. |
|
@lgirdwood please try this version of your second patch. |
7d6b224 to
9283180
Compare
Thanks, this seems to have done the trick, I've had to add |
486217e to
8e7d05f
Compare
8e7d05f to
091c6ca
Compare
10042e9 to
c3cfacb
Compare
|
Ready for staging. |
jsarha
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Tried to get this to compile, but it failed. The make file tried to look for sources under ${BASE_DIR}/tflm-micro , but the code is cloned under ${BASE_DIR}/tflite-micro.
marcinszkudlinski
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why special scripts/tensorflow-clone.sh script? We do use west, just add the repos there
c3cfacb to
afa8164
Compare
Thanks fixed, the built-in config was using the wrong directory but llext was building fine. |
That was my initial intention too, but I could only get this working with a west manifest in each child module repo (and this would be difficult/slow to upstream). I think the script is fine to unblock us today as either someone with better west know how or updates to west could fix this in the future. |
This is staging quality and adds a TFLM module based on micro_speech example to do audio classification. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
kv2019i
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Some minor comments, but nothing worth blocking the PR.
| ``` | ||
| CONFIG_CPP=y | ||
| CONFIG_STD_CPP17=y | ||
| CONFIG_SOF_STAGING=y |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This could be an overlay alike sof/app/shell_overlay.conf or sof/app/perf_overlay.conf.
| g_arena, kArenaSize); | ||
|
|
||
| // and allocate the tensors | ||
| if (interpreter->AllocateTensors() != kTfLiteOk) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is more of a tflite API question, but wondering can "interpreter" be NULL/nullptr here? Maybe this is ok and the new operator is just overloaded.
Integrates Tensorflow micro edition using xtensa optimized kernels into SOF as a module. WIP.