You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
For reader.hpp we need methods for static_matrix(i, j) and static_vector(i) to read in a static matrix. One small issue atm is that when we call log_prob_grad or log_prob_propto we pass in a vector of var types to the model's log prob that is then sent into the reader to be deserialized into the model's containers. This means that in the static_matrix etc functions we will have to do a to_var_matrix() and do a full copy. Though idt there's any non-trivial way to get rid of that copy.
Current Version:
v2.24.0
The text was updated successfully, but these errors were encountered:
The only thing I can think of is actually passing in a var<mat> to the model's log_prob and doing the unpacking for the dynamic case. On one hand that feels nice but unless we did something clever with the views in #2024 I think this would end up with more copies (since right now a "copy" of a var to a var is really just a copy of a pointer)
Summary:
For reader.hpp we need methods for
static_matrix(i, j)
andstatic_vector(i)
to read in a static matrix. One small issue atm is that when we calllog_prob_grad
orlog_prob_propto
we pass in a vector ofvar
types to the model's log prob that is then sent into the reader to be deserialized into the model's containers. This means that in thestatic_matrix
etc functions we will have to do ato_var_matrix()
and do a full copy. Though idt there's any non-trivial way to get rid of that copy.Current Version:
v2.24.0
The text was updated successfully, but these errors were encountered: