-
PyPy project is looking for CI that could host Windows buildbots for building and testing PyPy interpreter and JIT compiler on Windows with x86 architecture. Is Appveyor capable of sustaining 2-hour b…
-
### 🐛 Describe the bug
custom autograd function defined in the model, jit.trace reports error
```
class GradientMultiplyLayer(torch.autograd.Function):
@staticmethod
def forward(ctx, …
-
### 🐛 Describe the bug
Running
```python
from torch import nn
import torch
from torch._decomp import get_decompositions
from torch.fx.experimental.proxy_tensor import make_fx
from torch._functo…
-
## 🐛 Bug
JIT can not recognize the imported function when I want to script a class outside the file where it is defined.
## To Reproduce
Suppose we have the following file structure：
```
main…
-
When I try to run the following
import numba
@jit(nopython = True)
def numba_func(x)
x_norm = np.linalg.norm(dx, axis = 1)
return(x_norm)
x = np.ones([3,3])…
-
A potential fix for this issues is to add a guard to ensure inputs to nvFuser kernels are non-empty tensor. Fallback to eager for empty tensor.
Invoked with: , , RuntimeError('sizes[i] > 0 INTERNAL…
-
I am trying to use tensorflow serving to serve a keras bert model, but I have problem to predict with rest api, below are informations. Can you please help me to resolve this problem.
# predict out…
-
## 🐛 Bug
For using a custom PyTorch mobile build a list of operators is required. If the model uses custom classes however `torch.jit.export_opnames` fails, even though the model can be successfull…
-
## 🐛 Bug
The following warnings is seen while executing jit tests with newer versions of PyTorch (v1.5.0 onwards)
`OpenBLAS Warning : Detect OpenMP Loop and this application may hang. Please rebui…
-
## 🐛 Bug
When calling torch.jit.script() on code using torch.finfo, a Runtime Error is generated with:
`Unknown builtin op: aten::finfo.`
## To Reproduce
Steps to reproduce the behavior:
…