Skip to content

[docs] zero + sequence parallelism - #44605

Merged
stevhliu merged 7 commits into
huggingface:mainfrom
stevhliu:distrib-training-2
Apr 10, 2026
Merged

[docs] zero + sequence parallelism#44605
stevhliu merged 7 commits into
huggingface:mainfrom
stevhliu:distrib-training-2

Conversation

@stevhliu

Copy link
Copy Markdown
Member
  • refactors DeepSpeed ZeRO doc:
    • moves the troubleshooting section to the more general Debugging guide to keep everything in one place
    • moves the sequence parallelism section into a new doc to give it more visibility
    • update to be more lightweight and focused on the content that matters to Transformers (for example, remove features that aren't specific to DeepSpeed, non-Trainer integration, launching from Jupyter notebook, etc.)
  • lightly refactors Sequence parallelism + Debugging docs

@HuggingFaceDocBuilderDev

Copy link
Copy Markdown

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

@stevhliu
stevhliu requested a review from SunMarc March 11, 2026 17:46
@stevhliu stevhliu mentioned this pull request Mar 16, 2026
3 tasks
@SunMarc
SunMarc requested a review from kashif April 9, 2026 14:47

@SunMarc SunMarc left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks ! just some minor comments

Comment thread docs/source/en/deepspeed_alst.md
Comment thread docs/source/en/deepspeed.md Outdated
Comment thread docs/source/en/deepspeed_alst.md Outdated
Comment thread docs/source/en/deepspeed.md Outdated
@stevhliu
stevhliu requested a review from SunMarc April 10, 2026 06:47

@SunMarc SunMarc left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks a lot !

Comment on lines +95 to +107
</hfoption>
<hfoption id="Accelerate config file">

Run the [accelerate config](https://huggingface.co/docs/accelerate/en/package_reference/cli#accelerate-config) command and answer questions about your hardware and training setup to create a `default_config.yaml` file in your cache.

```yaml
distributed_type: DEEPSPEED
deepspeed_config:
deepspeed_config_file: path/to/ds_config.json
machine_rank: 0
num_machines: 1
num_processes: 4
```

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Note that with this version you don't need to pass the deepspeed argument in TrainingArguments and even if you pass it, it will be ignored.

Comment thread docs/source/en/deepspeed_alst.md
@stevhliu
stevhliu added this pull request to the merge queue Apr 10, 2026
Merged via the queue into huggingface:main with commit 09d91db Apr 10, 2026
16 checks passed
@stevhliu
stevhliu deleted the distrib-training-2 branch April 10, 2026 16:46
sirzechs66 pushed a commit to sirzechs66/transformers that referenced this pull request Apr 18, 2026
* zero + debug

* sp

* revert

* link to blog

* feedback

* feedback
eustlb pushed a commit that referenced this pull request May 12, 2026
* zero + debug

* sp

* revert

* link to blog

* feedback

* feedback
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants