Closed kyleaitgroup closed 5 months ago
Hi, Which deployment code are you using, it looks like the registration of the TRANSPOSE
operator is missing.
https://github.com/Seeed-Studio/SSCMA-Micro/blob/dd57e1cdd0317b7ae97838ae53fbd18f92acb020/porting/espressif/el_config_porting.h#L69
also thanks for trying, we will fix the training issue you mentioned as soon as possible, thanks!
Hi, Thanks for the quick response!
I'm currently using https://github.com/Seeed-Studio/sscma-example-esp32
I've checked that the line#define CONFIG_EL_TFLITE_OP_TRANSPOSE
does exist in the code, but the error persists.
LynnL4 commented 21 minutes ago Do you use this example? I will check it https://github.com/Seeed-Studio/sscma-example-esp32/blob/main/examples/fomo_detection_demo/main/app_main.cpp
Yes ,but I didn't notice that I'm not using the latest version of sscma-example-esp32. I'll try it first .
Hi, If you're using an older version of the repository, perhaps the problem is here https://github.com/Seeed-Studio/sscma-example-esp32/blob/d926450d530216340c46a0daf0f9435294abce27/components/modules/algorithm/algo_fomo.cpp#L223
Thanks. I've tested the latest version, but unfortunately, I encountered a new error.
Failed to resize buffer. Requested: 2092448, available 1044760, missing: 1047688 [ASSERT] Failed assertion 'interpreter != nullptr' E (5922) task_wdt: Task watchdog got triggered. The following tasks/users did not reset the watchdog in time: E (5922) task_wdt: - IDLE (CPU 0) E (5922) task_wdt: Tasks currently running: E (5922) task_wdt: CPU 0: main E (5922) task_wdt: CPU 1: IDLE E (5922) task_wdt: Print CPU 0 (current core) backtrace
Could it be caused by setting the epoch too large? I set the epoch to 300.
I see that your configuration is height=416 width=416, this will cause the model to require a lot of running memory, suggest changing it to 192x192
Hi @kyleaitgroup, the problem in training is caused by different types of ti
and po
. We fixed it in the latest version. Thank you!
Describe the bug A clear and concise description of what the bug is.
Environment Environment you use when bug appears:
python tools/train.py configs/fomo/fomo_person.py --cfg-options \work_dir=work_dirs/fomo_300 num_classes=1 epochs=300 height=416 width=416 data_root=datasets/coco_person/
Additional context I'm using latest version of ModelAssistant .
after I change
site = np.sum([ti, po], axis=0)
inModelAssistant/sscma/models/heads/fomo_head.py
withsite = np.concatenate([ti, po], axis=0)
it stat training,but I get another error when training the modelAdditional context if i set tp = fp = fn = 0 ,it works. but when I convert the model to binary and deploy it on the ESP32-S3 Eye, I encounter the following errors:
When I convert the model to binary and deploy it on the ESP32-S3 Eye, I encounter the following errors:
and