-
## š Bug
Tracing code that modifies a boolean tensor fails with TracingCheckError. Tracing works only with `torch.jit.trace(..., check_trace=False)`.
## To Reproduce
Small repro case:
```
ā¦
-
I wanted to share the method to export detic model to ONNX format with custom vocabulary.
**Step 1)** First of all, comment out this line `box_features = _ScaleGradient.apply(box_features, 1.0 / seā¦
-
Starting with Xcode 15 and iOS 17, `ios-deploy` no longer works on iOS 17+ physical devices and [sounds like ios-deploy will not be supporting it](https://github.com/ios-control/ios-deploy/issues/588#ā¦
-
Run: [runtime-coreclr r2r-extra 20220710.1 ](https://dev.azure.com/dnceng/public/_build/results?buildId=1871413&view=ms.vss-test-web.build-test-results-tab&runId=49032984&resultId=108641&paneView=debuā¦
-
### What happened?
I encountered a runtime error when running FrankenPHP in worker mode alongside an endpoint that uses AMPHP. The error does not occur in non-worker mode.
Environment
FrankenPHP ā¦
-
## Build Information
Build: https://dev.azure.com/dnceng-public/public/_build/results?buildId=876794&view=logs&j=07effcfd-c63d-5008-e811-03b29530f5cb&t=506b47fe-4551-54d6-41af-334dd2c09566
Build errorā¦
-
## š Bug
A minimal repro for the fixed issue (https://github.com/Lightning-AI/lightning-thunder/issues/461#issuecomment-2178023346) doesn't work anymore with a failure in Thunder's interpreter:
``ā¦
-
`grad` transform has to be applied after `thunder.jit` while `autocast` has to be applied before `thunder.jit`
```python
import torch
import torch.nn as nn
import thunder
from thunder.core.tranā¦
-
#### The next comment explains what this issue aims at
When `thunder.jit` trace `optimizer._init_group` before this initialization, calls of the jitted `optimizer.step` will all reset the inner stā¦
-
## š Feature
[jit] Support `torch.Tensor.numel()` and `torch.Size.numel()` properly in tracing.
## Motivation
Currently, tracing `x.numel()` causes the number of elements of `x` to be staticallyā¦
lqf96 updated
3 years ago