Hi
I got this error message while I was trying convert a S3FD Model which is trained on Caffe (
Link)
After tracing down the api code, i find the error message is come from /usr/local/bin/ncsdk/Controllers/
FileIO.py:153
...
def get_zero_buffer(for_data, datatype):
...
if zero_data_offset - buffer_size + pad + buffer_index > 100 * 1024 * 1024:
throw_error(ErrorTable.NoResourses)
...
then I slightly modified the code to know how many does the model really need.
The conversion is then successful and it showed that it needs 300 MB roughly.
I've tried to run the graph then i got a timeout error from device.
I know that It's possible to split model into several parts and run them on multiple sticks(detail was mentioned here)
Is there any other way that i can do avoid this error ?
or do you have any future plan to raise on limit of Intermediate layer buffer ?
