Yuanhao Ji

[Question] How to push the docker image with different tags to multiple registries when deploying

1

Hi, there! Thank you for your contributions first! I just have an question. How to push the docker image with different tags to multiple registries when deploying? I just want...

Fix device type issue in `_get_device_handle`

9

Fix #124327 `device_type`, the first arg of [init_device_mesh()](https://github.com/pytorch/pytorch/blob/a0466061e17358fb621cfde3f85e0bd6d13cfc55/torch/distributed/device_mesh.py#L503), does not support types with indexes, such as `cuda:0`. If `cuda:0` is used as a parameter, `_get_device_handle()` will not correctly return `torch.cuda`....

Fix the duplicated line in `advanced_source/privateuseone.rst`

3

## Description Fix the duplicated line in `advanced_source/privateuseone.rst` ```diff 1. Register new backend module to Pytorch. -2. Generate methods and properties related to the new backend. +2. Rename PrivateUse1 to...

Doc: Improve usage of the word language model example

3

- add usage of `word_language_model/generate.py`

## Standards checklist: - [x] The PR title is descriptive. - [x] The PR doesn't replicate another PR which is already open. - [x] I have read the contribution guide...

Add autoloading tutorial

8

Fixes #3039 This PR adds a tutorial about the out-of-tree extension autoloading mechanism introduced in https://github.com/pytorch/pytorch/issues/122468 and implemented in https://github.com/pytorch/pytorch/pull/127074 cc: @jgong5 @Yikun @hipudding @FFFrog @albanD

[WIP] Update privateuse1 doc

1

Fixes https://github.com/pytorch/pytorch/issues/134293 cc @NmomoN @mengpenghui @fwenguang @cdzhan @1274085042 @PHLens @ailzhang @albanD

💡 [REQUEST] - Tutorial on out-of-tree extension autoloading mechanism

1

### 🚀 Describe the improvement or the new tutorial This tutorial will help the third party extension writers better to improve seamless integration within PyTorch. ### Existing tutorials on this...

Add `--device` argument to run examples on a specific device

5

All examples can only be run on CUDA or MPS using `--cuda` or `--mps`. This change adds an argument `--device` to make our examples run on a specific device, e.g.,...