Hi, currently using QCS410 SoC devkit (Linux) with an Adreno GPU 612. (and SNPE version 1.41.0.2173)
Issue: I cannot run .dlc models on the GPU.
Details: When I try running .dlc models on snpe-net-run and snpe-throughput-net-run with the GPU flag enabled --use_gpu I always get this error:
error_code=1006; error_message=Invalid constraint. error_code=1006; error_message=Invalid constraint. No GPU device available for target UNKNOWN_SDM; error_component=Model Validation; line_no=79; thread_id=4013401056; error_component=Model Validation; line_no=271; thread_id=4159881232
(Sample command: snpe-throughput-net-run --container ssd_mobilenet_v1_coco_2018_01_28.dlc --duration 5 --use_gpu --perf_profile "default")
I have verified with snpe-platform-validator that the GPU runtime is in fact available:
/snpe-1.41.0.2173/bin/arm-oe-linux-gcc8.2hf # ./snpe-platform-validator --runtime gpu
Hi, I runned inception_v3.dlc model on snpe-throughput-net-run with gpu flag enabled on QCS610 it is working fine , below is the command and result.
Command : snpe-throughput-net-run --container inception_v3.dlc --duration 5 --use_gpu --perf_profile "default"
*/ Result : [Thread 0 - gpu_float32_16_hybrid] 2.41937 infs/sec - Number of images processed: 13 - Build time: 1633724 microseconds - Elapsed time: 5373303 microseconds - Teardown time: 64042 microseconds
Total throughput: 2.41937 infs/sec
*/
Excuse me,
has this problem been solved yet?
I have the same problem