-
Hi. I've found, that you unfreeze the whole GAN, and making steps only via specific optimizer (for generator and discriminator). But when you do loss.backward, gradients are computed for the WHOLE GAN…
-
http://arxiv.org/abs/1606.03401
-
Some operators create extra arrays, which can cause memory problems with larger models. In some cases this is avoidable. For example, the [ReLu step math](https://github.com/nengo/nengo/blob/master/ne…
-
Memory Efficient Convolution (MEC) is an alternative to `im2col` to lower the tensor and convolution kernel to feed them to BLAS for Matrix-Multiplication.
Compared to FFT and Winograd it uses much…
-
I think it will be useful to add operations akin to [`scipy.linalg.lu_facto` and `scipy.linalg.lu_solve`](https://docs.scipy.org/doc/scipy-0.14.0/reference/generated/scipy.linalg.lu_factor.html).
T…
-
```
A lot of temporary arrays are used in most NumPy programs and they are handled
quit efficient by Malloc because all Mallocs are almost immediately followed by
a Free.
However, this is not the c…
-
We need an efficient memory map to [be able to manage malloc and mmap effectively](https://github.com/Lind-Project/lind-wasm/issues/4). It seems like a good start would be to port [NaCl's memory map](…
-
### The problem you're addressing (if any)
There is no support for DDR5 SPD support in Linux, and those modules are used in Odroid-H4+; this results in a lack of information in the HCL report.Linux d…
-
We found that measurement-error mitigation that should be memory-efficient, like the TensoredFilter, is implemented with creating a vector whose length is exponential in the number of qubits. This…
-
I have been running into an issue where the FileChooser widget uses quite a bit of RAM when opening directories with a large number of files. For example, using the widget in kivy/uix/filechooser.py, …