We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
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
Hey, so I've been using this repo as a starting point for running my own models on a GAPuino v2 (Revision B).
Overall its been working well, but the speeds I see are much slower than the speed reported in the README of this repo.
I've seen the same issue as discussed in #1, but that's not a problem as I don't need large models for my project.
However, when I run MobileNetV1 0.25 on 224x224 images I can average about 2 FPS, which is an order of magnitude slower than reported.
If I run the model using GVSOC I see a similar FPS to what is reported.
I'm using the newest version of the GAP SDK and running the code from this repo unaltered.
I see that your results were generated with a RevC, but I'm not sure if that would explain such a big difference in performance.
The text was updated successfully, but these errors were encountered:
What frequencies do you have the Cluster and FC set to?
Sorry, something went wrong.
I've been running it with FREQ_FC=250MHz and FREQ_CL=175MHz, following this repo.
hi @colinski the Makefile sets 175MhZ for the cluster only if you have a gap8 RevC (GAP V3):
ifeq '$(TARGET_CHIP)' 'GAP8_V3' FREQ_CL?=175 else FREQ_CL?=50 endif
You can use this API in your code to verify that:
int32_t cur_fc_freq = pi_freq_get(PI_FREQ_DOMAIN_FC); int32_t cur_cl_freq = pi_freq_get(PI_FREQ_DOMAIN_CL);
int32_t cur_fc_freq = pi_freq_get(PI_FREQ_DOMAIN_FC);
int32_t cur_cl_freq = pi_freq_get(PI_FREQ_DOMAIN_CL);
No branches or pull requests
Hey, so I've been using this repo as a starting point for running my own models on a GAPuino v2 (Revision B).
Overall its been working well, but the speeds I see are much slower than the speed reported in the README of this repo.
I've seen the same issue as discussed in #1, but that's not a problem as I don't need large models for my project.
However, when I run MobileNetV1 0.25 on 224x224 images I can average about 2 FPS, which is an order of magnitude slower than reported.
If I run the model using GVSOC I see a similar FPS to what is reported.
I'm using the newest version of the GAP SDK and running the code from this repo unaltered.
I see that your results were generated with a RevC, but I'm not sure if that would explain such a big difference in performance.
The text was updated successfully, but these errors were encountered: