-
### 🐛 Describe the bug
In some scenarios, user may add custom functions/methods for torch.Tensor/torch.XXX (such as this official API torch.utils.generate_methods_for_privateuse1_backend #121723 ), t…
-
### Your current environment
```text
Collecting environment information...
PyTorch version: 2.3.0+cu121
Is debug build: False
CUDA used to build PyTorch: 12.1
ROCM used to build PyTorch: N/A
…
-
### 🐛 Describe the bug
On a toy regression model with pytorch 2.1.2 on cpu, `torch.autocast` is really slow.
Without `with torch.autocast(device_type="cpu", dtype=torch.bfloat16, cache_enabled=True)…
ldv1 updated
5 months ago
-
### 🐛 Describe the bug
Hello there,
This is the same issue as https://github.com/pytorch/pytorch/issues/80979, which had been closed even though I can still reproduce the error.
I am trying t…
-
Hi everyone,
in [this branch](https://github.com/Eclectic-Sheep/sheeprl/tree/feature/compile) one can use `torch.compile` to compile the Dreamer-V3 agent. In particular:
* in the `sheeprl/configs…
-
### 🐛 Describe the bug
This fails in inductor but works in eager and aot_eager. I believe inductor is just calling the "normal" clone but needs to handle the sparse tensor clone specially.
```
…
-
## Issue description
I've been observing an issue with the `torch.cuda.memory_reserved("cuda:0")` function in PyTorch. Despite having a model training on the GPU, the function `torch.cuda.memory_…
-
### 🐛 Describe the bug
As of `torch==2.4`, we're unable to register an out-of-tree backend using the `PrivateUse1` mechanism described in [this documentation](https://pytorch.org/tutorials/advanced/e…
-
### 🐛 Describe the bug
torch.export.export fails to trace through a binary operator.
Here is the failing code
```python
import torch
class MyTensor:
def __init__(self, tensor):
…
-
### 🐛 Describe the bug
When using FSDP (Fully Sharded Data Parallel) to save a model, some parameters are not fully gathered on rank 0 and therefore not properly saved. This issue occurs specifical…