-
Last year I followed the implementation logic in this file, https://github.com/openxla/xla/blob/7954169ccfb6290d94af3ea3634229b097682ba8/xla/service/gpu/runtime/gemm.cc
and the input parameters were…
-
![image](https://user-images.githubusercontent.com/19290577/213022789-054fb230-34f8-452b-ac92-ea590f3a98fd.png)
-
### 🐛 Describe the bug
The `torch.compile` mode and the eager mode differ in generating tensor API behavior, which usually results in the eager mode running without problems, while the `torch.compile…
-
I developed a Triton kernel that is replacing the `Linear` layer in a framework that I am currently developing.
This kernel is integrated in the standard way:
```
class TritonLinear(Function):
…
-
The incompatibility is that during backwards, fused_rmsnorm does dynamic control flow over strides, which isn't safe for export tracing used by PP.
```
dy = dy.view(-1, dy.shape[-1])
…
-
Thank you for your great work. We tried fine-tuning with a learning rate of 1e-5, frame stride set to random in 1-6, resolution: [576, 1024], video_length: 16. After training for a period of time, the…
-
Not sure how to recreate this, but it seems like during `npm install` process in `prepare` phase, if it takes a while, it crashes strider. No errors that I can see.
Have strict caching enabled and no…
-
I am re-implementing the enhancement of DP-SGD through the [random sparsification](https://github.com/JunyiZhu-AI/RandomSparsification) of gradients on my UNet Model.
Here is a Debug info on extend…
-
I'm curious how hard it might be to change to CImg addressing calculations to allow arbitrary row to row and slice to slice strides?
ie, instead of index= x + width * ( y + height * z ) ), index = …
-
![image](https://github.com/threestudio-project/threestudio/assets/44741080/fa84d180-d948-485a-aa90-129bba0bc66c)
Great job! I encountered a warning during runtime. Could this have an impact on the f…