-
## 🚀 Feature
Intermediate tensors are currently re-computed instead of re-used in fragmented computation. This could cause regression comparing to eager, which reuses tensor and avoids wasted re-co…
-
### Describe the bug
I just updated my er-x build to latest main and eth1 has stopped working. This would seem to be a regression in the past few weeks and I do notice some patches around ethernet su…
-
Context: I'm using IJava to teach Java via Google Colab Jupyter notebooks, and I just realized that no import statements are necessary for, say, access to the contents of the `java.io` package. I'd l…
-
Hi,
I am currently implementing a custom kernel and facing problems with the gradients' computation: dL_dK is an NxN matrix (for N data points) whereas dK_dtheta should be N^2 x M, for M parameters…
yhalk updated
6 years ago
-
When running Clear-Linux with a custom kernel which happens to not match the current default entry, adding a new kernel bundle through swupd bundle-add results in a failure to install the new bundle's…
-
I understand from the doc and conversation that Jupyverse is modular by design :+1:
So I understand it should be feasible to add endpoints, aka server extensions. aka asphalt plugins (?).
Do such…
-
Please review my comments here:
https://github.com/X16Community/x16-rom/commit/f629cff8cdacc4b5dba8a86068833d57b99a5d83#r147340847
Copied here for reference:
```
ld81 lda verck
dec
beq l…
-
### Describe the feature request
Custom ops may have inputs/outputs on different memory locations. Internally we can specify inputs/outputs memory type by calling KernelDefBuilder__InputMemoryType() …
-
### Issue type
Bug
### Have you reproduced the bug with TensorFlow Nightly?
Yes
### Source
source
### TensorFlow version
tf2.12.0
### Custom code
Yes
### OS platform and …
-
I successfully booted Linux with FPU disabled. When I enable FPU, I get the following errors:
```
[ 0.156632] smp: Bringing up secondary CPUs ... …