Skip to content

Conversation

@mdanilow
Copy link

  • Feature input shapes of external weights layers are saved to io_shape_dict driver dictionary under "external_weights_input_shapes" key.
  • These shapes are then used to recognize convolutional layers and send the parameters proper number of times in driver_base.py
  • tests/end2end/test_ext_weights.py tests also the cnv-w2a2 topology along tfc-w2a2 and properly verifies the number of runtime weights files ("idmaX.npy"). This test is changed to use only hls backends for MVAU due to a different existing bug - to be addressed.
  • Fixed a bug in validate.py script in the deployment package caused by the case of all wrong outputs in the test batch (np.bincount would return just number of 0's instead of number of 0's and 1's)

@mdanilow mdanilow force-pushed the feature/conv_ext_weights branch from 2c30803 to c124579 Compare September 5, 2024 12:25
@mdanilow mdanilow force-pushed the feature/conv_ext_weights branch from c124579 to b806a85 Compare October 10, 2024 16:19
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant