-
### 🐛 Describe the bug
cannot using:
```
python -m examples.models.llama.export_llama --checkpoint "checkpoint.pth" --params "original_params.json" -kv --use_sdpa_with_kv_cache -X -d …
-
I'm experiencing an issue when using DeepSpeed's universal checkpointing. After converting my DeepSpeed checkpoint to a universal checkpoint using `ds_to_universal.py`, resuming training from the co…
-
Platforms: linux
This test was disabled because it is failing in CI. See [recent examples](https://hud.pytorch.org/flakytest?name=test_comprehensive_full_cpu_float16&suite=TestInductorOpInfoCPU&limit…
-
Platforms: linux
This test was disabled because it is failing in CI. See [recent examples](https://hud.pytorch.org/flakytest?name=test_comprehensive_nn_functional_cosine_embedding_loss_cpu_float16&su…
-
Platforms: mac, macos
This test was disabled because it is failing in CI. See [recent examples](https://hud.pytorch.org/flakytest?name=test_where_broadcast_cpu&suite=CpuTests&limit=100) and the most …
-
### 🐛 Describe the bug
DTensor random numbers provide an offset based RNG state tracker OffsetBasedRNGTracker for CUDA. However, for CPU, this offset based RNG state tracker is not available, and it …
-
### Your current environment
The output of `python collect_env.py`
```text
WARNING 10-16 11:39:10 cuda.py:22] You are using a deprecated `pynvml` package. Please install `nvidia-ml-py` instea…
-
### 🐛 Describe the bug
```python
scale_tril = sigma_t.view(-1, 1, 1) * torch.eye(3072, device=mean.device).unsqueeze(0)
scale_tril = repeat(scale_tril, "b p1 p2 -> (b n) p1 p2", n=all_images[…
-
ET could not be build with python 3.8 atm. Creating this issue to hide the failure on CI until 3.8 build could be included/excluded properly.
Related issues:
* https://github.com/pytorch/executorc…
-
Platforms: asan, linux, mac, macos, win, windows
This test was disabled because it is failing in CI. See [recent examples](https://hud.pytorch.org/flakytest?name=test_device_mode_ops_sparse_mm_reduce…