-
## 🐛 Bug
Whenever I use `generate` function on a TPU (I use v5e litepod8), I have a crash with a C++ stack trace but no info on the python side and no way to catch/recover.
## To Reproduce
I…
-
## Description
We have identified an issue with our tensor operation tests, particularly in how they handle data types. This inconsistency can lead to false positives in our test suite, potentially…
-
In Python it's possible to write just this:
`3 * x**2 - 4*x + 5`
With Candle it currently looks like this, which is difficult to read and write:
`Ok((&(&(&x.powf(2.0)? * 3.0)? - &x * 4.0)? + 5.0)…
-
Current implementation of symmetries has limited handling of symmetries that exclude for instance the symmetries of the Riemann tensor.
One would naturally expect:
* To be able to (anti)symmetrie…
-
### Is there an existing issue for this?
- [X] I have searched the existing issues
### Bug summary
A memory leak issue is observed in TorchIO version 20.0.1 during prolonged training sessions…
-
# Convolutional neural networks for solving PDEs | Arthur Bauville
Writing an iterative multigrid PDE solver as a convolutional neural network
[https://abauville.github.io/blog/neural%20networks/202…
-
Context:
- mixed precision training
- model: bf16
- optimizer: fp32
When we dump a V2-format optimizer state dict, it contains the params fields which we expect to be high-precision versions of the m…
-
Today, we have separate registration for edit & display ui on the "component ui registry". Whenever we want to display a value and there's no display ui, we fall back to a disabled version of the edit…
Wumpf updated
3 months ago
-
### System Info
- TensorRT-LLM version: 0.10.0.dev2024050700
(I doubt any other information is relevant)
### Who can help?
@kaiyux
### Information
- [ ] The official example scripts
- […
-
# Name of the type
For now we will assume it's `Tensor`. However it's opionated, because it might be useful to avoid calling it the same name as other libs.
# Operations
Namespace: `GenericTens…