-
Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace opera…
-
Hi, I'm modifying torch-ngp code to create a model that updates the pose. I created a new custom autograd function, with forward and backward. However, I wonder why the backward I defined is not execu…
-
I've noticed the superloss implementation is similar with AlanChou's unoffical implementation (https://github.com/AlanChou/Super-Loss). Both of which used scipy to calculate lambertw. However, as stat…
-
The bwd and fwd_bwd tests for layer_norm failed.
Error string is `RuntimeError: This backward function was compiled with non-empty donated buffers which requires create_graph=False and retain_grap…
-
**Describe the bug**
When turning the flag with_mixed_precision = False the code crashes when computing the attentions [here](https://github.com/clessig/atmorep/blob/main/atmorep/transformer/transf…
-
### 🚀 The feature, motivation and pitch
`torch.compile` can takes order of seconds to compile a decent size model like Llama2 7B with a `aot-autogra` enabled backend. Note that I only include the `dy…
-
Hi,
I am interested of your work. I'm trying to get it installed for visualizing detection results but ended up with an error while running `python setup.py build develop`. Is there a fix for this?…
-
when running examples in **Basic Usage**, it shows errors:
```shell
2024-09-07 20:09:29,033 - modelscope - INFO - PyTorch version 2.0.1 Found.
2024-09-07 20:09:29,033 - modelscope - INFO - Loadin…
-
### 🐛 Describe the bug
NaViT (Native Resolution Vision Transformer) is a variant of the ViT. It can be draft-exported but failed when running `aoti_compile_and_package`
To repro:
1. Install vit-…
-
Hello,
I am using a tcnn Hash Encoding + Cutlass MLP to train a neural field. I would like to access second derivatives to implement a smoothing loss during training, as well as obtain gradients fo…