-
### 🐛 Describe the bug
Microsoft directml custom backend for pytorch gpu acceleration in WSL receives error in huggingface transormer .generate method.
directml_torch reference:
https://learn.m…
-
We hope to have as many CI workflows across the PyTorch Ecosystem using the Generic CI Jobs. Here is the link to the documentation/migration guide to the generic CI jobs: https://github.com/pytorch/te…
-
Can anyone generate me a working `pip freeze --all` including all dependencies and versions needed by audiocraft?
Even since a couple of weeks ago `pip install -e .` stopped working for me because …
-
Currently the user has to pip install the nightly explicitly. Instead once torchdata is released with StatefulDataLoader, add it to requirements.txt and pyproject.toml and remove the pip install from …
-
### 🚀 The feature
Allows for sourcing datasets from SQL queries. Should allow to substitute in different backends. eg Athena, Presto, Postgres. Should do smart batching to minimize number of query re…
-
### 🐛 Describe the bug
I am training a model using pytorch lightning using mixed precision. I have a region of the network that requires the dynamic range of float32, so I am using the `with torch.a…
-
Trying to fix nightly ci uncovered some actual bugs in the code that should have been found with normal CI. Are we not testing everything?
-
### 🐛 Describe the bug
It's known that layer norm needs to compute the variance and mean of its input. So we can expect that `torch.var_mean` runs faster than `LayerNorm`. But, when I time them, I …
-
### 🚀 The feature
```python
class MultiProcessingReadingService(ReadingServiceInterface):
num_workers: int = get_number_of_cpu_cores()
pin_memory: bool = True
timeout: float
wo…
-
## 🚀 Feature
**Motivation**
* To avoid pitfall with shuffling and sharding of datapipes in distributed training environments
* To ensure consistent experience of TorchData based datasets ac…