-
### 🐛 Describe the bug
For usual backward engine, intermediate tensors can be removed within checkpoints.
For forward autograd, intermediate tensors can be always removed if not directly available t…
-
I have hardened my sandbox (desktop, documents, deletion of floppy disks, CPU cores, RAM, HDD size...) and I still notice Pafish detects the sandbox (running on VMWare vSphere).
I have succeeded i…
-
I am facing an issue where the Windows Server VM gets stuck on the "TIANO CORE" screen after a reboot. A simple reboot trigger, such as a "Windows Update", causes the problem. It doesn't happen consis…
-
### Your current environment
The output of `python collect_env.py`
```text
Collecting environment information...
PyTorch version: 2.4.0+cu121
Is debug build: False
CUDA used to build PyTorch…
-
### 🐛 Describe the bug
When generating multiple samples from a compiled int4 model on CUDA, a runtime error occurs relating to Linear.weight swapping:
```
Traceback (most recent call last):
Fi…
-
### 🐛 Describe the bug
In the model available at https://github.com/alita-moore/img-to-text, increasing the length of the input IDs beyond 1 causes the process to slow down by approximately 2x compar…
-
### 🐛 Describe the bug
It seems that tensor.to(device) isn't copying the data correctly when it is moved from one GPU to another GPU.
Copying to/fro a GPU and a CPU seems to work as intended.
M…
-
from @secondsun
> Details https://github.com/minishift/minishift/pull/2241 and https://github.com/openshift/origin/pull/19209
>
> There's some work to do to update our docs because it also look…
-
### 🐛 Describe the bug
When processing `NaN` input on CPU and CUDA, the results seem inconsistent.
```python
import torch
a = torch.tensor(float('nan'))
b = a.cpu().type(torch.int32)
c = a…
-
### 🐛 Describe the bug
__torch_dispatch__ leads to:
```
RuntimeError: false INTERNAL ASSERT FAILED at "aten/src/ATen/RegisterFunctionalization_0.cpp":2151, please report a bug to PyTorch. mutatin…