HI ,
I’m a newbie,
Currently I have a question. I trained a yolov8n model.
It works well on PC.
The tflite model has been converted to vela,
However, when the program runs, it shows that the model init failed.
The following is the log converted by vela:
Accelerator configuration Ethos_U55_256
System configuration RTSS_HP_SRAM_MRAM
Memory mode Shared_Sram
Accelerator clock 400 MHz
Design peak SRAM bandwidth 2.44 GB/s
Design peak Off-chip Flash bandwidth 0.39 GB/s
Total SRAM used 2758.31 KiB
Total Off-chip Flash used 2204.52 KiB
CPU operators = 5 (1.7%)
NPU operators = 287 (98.3%)
Average SRAM bandwidth 2.00 GB/s
Input SRAM bandwidth 48.18 MB/batch
Weight SRAM bandwidth 11.96 MB/batch
Output SRAM bandwidth 28.89 MB/batch
Total SRAM bandwidth 89.09 MB/batch
Total SRAM bandwidth per input 89.09 MB/inference (batch size 1)
Average Off-chip Flash bandwidth 0.05 GB/s
Input Off-chip Flash bandwidth 0.02 MB/batch
Weight Off-chip Flash bandwidth 2.14 MB/batch
Output Off-chip Flash bandwidth 0.00 MB/batch
Total Off-chip Flash bandwidth 2.17 MB/batch
Total Off-chip Flash bandwidth per input 2.17 MB/inference (batch size 1)
Neural network macs 1014000800 MACs/batch
Network Tops/s 0.05 Tops/s
NPU cycles 13486835 cycles/batch
SRAM Access cycles 12643265 cycles/batch
DRAM Access cycles 0 cycles/batch
On-chip Flash Access cycles 0 cycles/batch
Off-chip Flash Access cycles 18347 cycles/batch
Total cycles 17795775 cycles/batch
Batch Inference time 44.49 ms, 22.48 inferences/s (batch size 1)
The following is the log of the program running:
Loading model from address: 0x0x8018bcb0
Model size: 2082048 bytes
Loading op resolver
INFO - Added ethos-u support to op resolver
Creating allocator using tensor arena at address: 0x0x2000000
INFO - Creating allocator using tensor arena at 0x0x2000000
Created new allocator at address: 0x0x226ff8c
INFO - Allocating tensors
ERROR - tensor allocation failed!
