Task 1: local tensor zero values and base model array empty

[dskhanirfan]

Hi
If I get 0 value in the local_tensor like local_tensors [LocalTensor(col_name='01', tensor=array(0., dtype=float32), weight=0.5), LocalTensor(col_name='02', tensor=array(0., dtype=float32), weight=0.5)] and base_model_nparray.shape = () Can you tell me what it means?

[sarthakpati]

Hi,

I can take an educated guess, but I will wait for @brandon-edwards and/or @psfoley to respond for added clarity.

Cheers,
Sarthak

[psfoley]

@dskhanirfan Thanks for the question. The custom aggregation functions are applied layer-wise to the model, and each of local_tensors correspond to each of the contributions coming from each of the collaborators who participated in that training round. In your example, it appears there are two collaborators ('01' and '02'), that both have weight values of 0 for the layer being processed (each with weight of 50%). Let me know if any other clarification is needed and I would be glad to expand!