-
**Issue identified:** cuDNN SDPA JIT recompiles when the context length changes. This results in training that does not use packing to keep recompiling, resulting in the observed 500ms overhead.
--…
-
Hi, my suggestion is based on the desire to get heap information updates not only at the beginning and end of GC, but also at any other time on request outside of GC. If the application does not perfo…
ww898 updated
2 weeks ago
-
I am trying to run pytorch profiler with tensorboard tutorial from [pytorch/tutorial](https://github.com/pytorch/tutorials/blob/main/intermediate_source/tensorboard_profiler_tutorial.py) in Windows 11…
-
### 🐛 Describe the bug
Under certain circumstances, the `torch.profiler.profile` will crash with the following error message:
```
RuntimeError: stack.size() INTERNAL ASSERT FAILED at
"/opt/conda…
-
### 🚀 The feature, motivation and pitch
Add support for 'MPS' in Pytorch profiler
```
In [1]: import torch
In [2]: from torch.profiler import profile, record_function, ProfilerActivity
...…
-
Profiler GUI application does not respond to close button, stucking forever until Ctrl-C.
System: Ubuntu 20.04, 22.04
Version: v0.9, v0.10
Tried LEGACY = 0 and 1
Capstone v0.4.2 installed fro…
-
### 🐛 Describe the bug
I am using PyTorch's execution trace observer to collect traces about a basic MNIST program. While the program **does** produce valid data (everything is there), it **does not*…
-
I was trying to build this since the .whl requires AVX2 and my testing machine does not have AVX2. I was already able to get all of the oneAPI and intel extensions for tensorflow to work. I did, howev…
-
### 🐛 Describe the bug
when enabling `kineto__tensor_core_insts` or `dram__bytes_read.sum`, the pytorch profiler outputs this warning and the trace becomes unusable. I have even tried adding the foll…
-
Platforms: rocm, dynamo
This test was disabled because it is failing in CI. See [recent examples](https://hud.pytorch.org/failure/test_kineto_profiler_with_environment_variable) and the most recent…