-
when running the code: diffopt.step(buff_losses)
the following problem occurs: RuntimeError: trying to differentiate twice a function that was markedwith @once_differentiable
but when setting track_…
-
- Some key ideas and history behind the EVT, especially focused on Gumbel
- Application of Gumbel softmax to backpropagate through discrete categorical sampling
- Extension of ideas behind Gumbel so…
-
I have not found---in CLP3---a definition of what it means for a fcn of two vars to be differentiable.
"Example 2.2.9. Optional — A Little Weirdness" flirts with this: partials existing when fcn i…
-
## 🚀 Feature
Let's design below what an API for differential optimizers would look like. The goal is to facilitate the implementation of differential optimizer for libraries such as [higher](https:…
-
Hi, thanks for the nice work! I am quite new to this, but wondering if we can do differentiable rendering with t3 as done in https://github.com/BachiLi/redner ?
oeway updated
3 years ago
-
Vectorize over `knots` and `intervals` instead of enforcing `knots` and `intervals` , but we need to make `knots` and `intervals` part of `params` possibly (in the compute function definition)
See …
-
It's ocassionally not possible to test `@non_differentiable` using `test_rrule` because finite differencing gives the wrong / unreliable results.
For example, this is the case in the `Vector{Floa6…
-
Images of differentiable maps from a domain to a higher-dimensional codomain cannot have a nonempty subset that is open in the codomain.
We define a subclass `DiffMapImage` of `ContinuousMapImage` …
-
## 🚀 Feature
It would be nice to add Differentiable Unpooling layer to this project. It is an [ICCV paper](https://arxiv.org/abs/1908.04964) and we have just released the code [OANet](https://githu…
-
In the following example, `aesop` does not terminate (for at least 20 minutes). Besides my hope that aesop might actually solve the goal, is "running forever" an intended behavior? This is the first t…