Eric Langlois

This can also be fixed by removing the `www.`. I would make a PR but I can't find an open source license for this code.

I've discovered the strategy you use in the [DDPG example](https://github.com/LaurentMazare/tch-rs/blob/master/examples/reinforcement-learning/ddpg.rs#L130) of constructing a new module instance then using the var store to copy the variables. That should work for me....

Having COptimizer available is sufficient for my purposes in terms of a non-templated optimizer type. It might still be nice in general to have the methods of Optimizer be available...

Replying to https://github.com/LaurentMazare/tch-rs/issues/281#issuecomment-855287935 from @jerry73204 > Another concern is that whether `.clone()` is shallow or deep affects the semantics of ownership. Currently, the self-modifying methods like `.absolute_(&mut self)` expects unique...

Is this safe? What happens if you create a view on a tensor `t` then call `t.resize_`?

I wanted this functionality too so you inspired me to try creating a hopefully-safe version: https://gist.github.com/edlanglois/5444813b409a9b44f4067591bf0f40b0 All views are provided through `ExclusiveTensor`, which creates the tensor and keeps it private...

I think that part is OK, there is an elided lifetime in the returned `ArrayView` so the signature is actually `array_view ArrayView

I don't think that `Copy` can be implemented for `Tensor` because `Copy` is for ["types whose values can be duplicated simply by copying bits"](https://doc.rust-lang.org/std/marker/trait.Copy.html). If you did that with `Tensor`...

Traceback from stopping the test: ``` File "/home/eric/dev/pycid/pycid/core/cpd.py", line 219, in initialize_tabular_cpd [[complete_dictionary(self.stochastic_function(**i))[t] for i in self.parent_values(cid)] for t in domain] File "/home/eric/dev/pycid/pycid/core/cpd.py", line 219, in [[complete_dictionary(self.stochastic_function(**i))[t] for i in...

Thanks for explaining the underlying cause. Yes, my issue is that I'm unable to merge a .yaml config into the nested structured config.