-
nvFuser generated code for a fusion block present in DiT has worse than expected performance. The subgraph is performing a `LayerNorm + + Mul + Add + Add` computation as shown in the code below. nvFus…
-
### Before You Begin
- [X] I confirm that I have downloaded the latest version of the addon.
- [X] I am not playing on a private server.
- [X] I checked for an [existing, open ticket](https://git…
-
`grad` transform has to be applied after `thunder.jit` while `autocast` has to be applied before `thunder.jit`
```python
import torch
import torch.nn as nn
import thunder
from thunder.core.tran…
-
I'm trying to investigate a huge performance drop affecting some low-end cards.
I currently suspect:
- a bug in the driver itself.
- something stupid done on our side similar to the way our sky…
-
In this opportunity[ the hacker](https://etherscan.io/txs?a=0x2a2c200af4e659348c4182dd9806a340851df42e&p=1) drained native assets from the protocol without deploying any contract.
Considering the h…
-
Mostly just what the title says. TC is going the same path as Insomnia sadly, and I'd like to use Bruno as an alternative.
Feel free to point me in the right direction, I could probably implement i…
-
**What happened**:
War Thunder reports connection issues to game servers and heavy packet loss.
After a failed attempt to join a match with a "connection error", a second attempt will succeed but wi…
-
### 🐛 Bug
NeMo's Stable Diffusion uses [CLIPTokenizer](https://github.com/NVIDIA/NeMo/blob/v1.23.0/nemo/collections/multimodal/modules/stable_diffusion/encoders/modules.py#L264) from HuggingFace. A…
-
Group norm is calculated as:
```
x0 = [N, C, H, W]
x1 = x0.cast(fp32).reshape(N, C, H, W) --> (N, G, C/G, H, W)
x2 = x1 / x1.sum(C/G, H, W)
x3 = x2.reshape(N, G, C/G, H, W) --> (N, C, H, W)
x4…
-
### Is Your Feature Request Related to a Problem?
If Thunder is unable to connect to the server, it shows a blank screen with unresponsive buttons. After quite a long period of time it will respond w…