docs(readme): update convergence table, latest news, and outdated links by sbhavani · Pull Request #2638 · NVIDIA/TransformerEngine

sbhavani · 2026-02-01T02:50:52Z

Description

Updates README add missing format support documentation, update news section, and fix broken/outdated links.

Type of change

Documentation change (change only to the documentation, either a fix or a new content)
Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to not work as expected)
Infra/Build change
Code refactoring

Changes

Add MXFP8 and NVFP4 format support to highlights and description
Update FP8 convergence table with MXFP8 results from arxiv paper
Remove outdated JAX Toolbox links and "available on request" entries
Update Docker container versions to 26.01
Fix DeepSpeed and Lightning integration links
Add Nemotron 3 paper to Latest News
Add quickstart notebook link after PyTorch example

Checklist:

I have read and followed the contributing guidelines
The functionality is complete
I have commented my code, particularly in hard-to-understand areas
I have made corresponding changes to the documentation
My changes generate no new warnings
I have added tests that prove my fix is effective or that my feature works
New and existing unit tests pass locally with my changes

greptile-apps · 2026-02-01T02:53:32Z

Greptile Summary

This PR updates README.rst with documentation improvements: adding a Nemotron 3 news entry, documenting MXFP8 and NVFP4 format support in the description and highlights, refreshing the FP8 convergence table with Megatron Core MXFP8 results (removing stale JAX Toolbox and "available on request" entries), updating the NGC container version to 26.01, and fixing integration links (DeepSpeed, Lightning) while removing "Coming soon" stubs for Colossal-AI and PeriFlow.

Key observations:

The quickstart.ipynb linked at line 143 (docs/examples/quickstart.ipynb) does not exist in the repository — only quickstart_utils.py and quickstart_jax_utils.py are present. This link was already broken in the base branch and is not introduced by this PR, but merging without addressing it keeps a broken link visible.
The "FP8 Convergence" section title and its introductory paragraph claim convergence only for FP8 vs BF16, but the PR description explicitly states the new LLM-8B and MoE-16B Megatron Core entries represent MXFP8 results from arxiv paper 2506.08027. The section heading and intro text are not updated to reflect this, which may mislead readers about what precision those entries demonstrate.
Removal of "Coming soon" entries for Colossal-AI and PeriFlow is unannounced — consider whether these integrations are now abandoned or simply no longer tracked.

Confidence Score: 4/5

This is a documentation-only PR with no code changes; safe to merge with minor doc accuracy concerns addressed.
All changes are documentation-only (README.rst). The link fixes, news entry, MXFP8/NVFP4 additions, and container version update are accurate and improvements. The main concern is the FP8 Convergence section's title and intro text not being updated to reflect the newly added MXFP8 results — this is a clarity issue rather than a factual error. The broken quickstart.ipynb link pre-exists in the base branch and was not introduced by this PR.
README.rst lines 379–382 (FP8 Convergence section title and intro paragraph should mention MXFP8 to match the newly added table entries)

Important Files Changed

Filename	Overview
README.rst	Documentation update with mostly good changes (news entry, link fixes, container version updates, removed "Coming soon" stubs), but includes a broken `quickstart.ipynb` link that pre-exists in the base and was not introduced by this PR, and the FP8 Convergence section's title and introductory text are not updated to reflect the addition of MXFP8 results from the cited arxiv paper.

Flowchart

flowchart TD
    A[README.rst Changes] --> B[Latest News Section]
    A --> C[What is TE? Description]
    A --> D[Highlights Section]
    A --> E[NGC Container Version]
    A --> F[FP8 Convergence Table]
    A --> G[Integrations List]

    B --> B1["+ Nemotron 3 paper (12/2025)\n NVFP4 on Transformer Engine"]

    C --> C1["+ MXFP8 / NVFP4 support\n mentioned for Blackwell GPUs"]

    D --> D1["+ MXFP8 and NVFP4 bullet\n for Blackwell GPUs"]

    E --> E1["Updated to 26.01\n (January 2026 release)"]

    F --> F2["- Removed: T5-770M (JAX/T5x)\n- Removed: GPT-5B (JAX/Paxml)\n- Removed: GPT-5B (NeMo, on-request)\n- Removed: T5-11B (JAX/T5x, on-request)\n- Removed: GPT-22B (NeMo, on-request)\n- Removed: GPT-175B (JAX/Paxml)"]
    F --> F3["+ Added: LLM-8B (Megatron Core)\n+ Added: MoE-16B (Megatron Core)"]

    G --> G1["DeepSpeed: specific test file → root repo\nLightning: GitHub issue → proper docs\n- Removed: Colossal-AI (Coming soon)\n- Removed: PeriFlow (Coming soon)\nNanotron: removed Coming soon tag"]

_{Last reviewed commit: f0bd464}

greptile-apps

_{1 file reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}

greptile-apps · 2026-02-01T02:53:35Z

README.rst

  loss = out.sum()
  loss.backward()

+For a tutorial with more details, see the `Quickstart Notebook <https://github.com/NVIDIA/TransformerEngine/blob/main/docs/examples/quickstart.ipynb>`_.


The referenced quickstart.ipynb file does not exist in docs/examples/. The actual notebooks in that directory are fp8_primer.ipynb, advanced_optimizations.ipynb, and te_jax_integration.ipynb. Consider using one of these existing notebooks or creating the quickstart notebook before merging.

greptile-apps

_{1 file reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}

greptile-apps · 2026-02-02T20:39:14Z

README.rst

+| LLM-8B     |  Megatron Core           | https://arxiv.org/abs/2506.08027                                                                        |
 +------------+------------------+---------------------------------------------------------------------------------------------------------+
 | MPT-13B    |  Mosaic Composer | https://www.databricks.com/blog/turbocharged-training-optimizing-databricks-mosaic-ai-stack-fp8         |
 +------------+------------------+---------------------------------------------------------------------------------------------------------+
-| GPT-22B    |  NeMo Framework  | Available on request                                                                                    |
+| MoE-16B    |  Megatron Core           | https://arxiv.org/abs/2506.08027                                                                        |


extra whitespace in "Megatron Core" framework names - inconsistent with other rows

Suggested change

| LLM-8B | Megatron Core | https://arxiv.org/abs/2506.08027 |

+------------+------------------+---------------------------------------------------------------------------------------------------------+

| MPT-13B | Mosaic Composer | https://www.databricks.com/blog/turbocharged-training-optimizing-databricks-mosaic-ai-stack-fp8 |

+------------+------------------+---------------------------------------------------------------------------------------------------------+

| GPT-22B | NeMo Framework | Available on request |

| MoE-16B | Megatron Core | https://arxiv.org/abs/2506.08027 |

| LLM-8B | Megatron Core | https://arxiv.org/abs/2506.08027 |

+------------+------------------+---------------------------------------------------------------------------------------------------------+

| MPT-13B | Mosaic Composer | https://www.databricks.com/blog/turbocharged-training-optimizing-databricks-mosaic-ai-stack-fp8 |

+------------+------------------+---------------------------------------------------------------------------------------------------------+

| MoE-16B | Megatron Core | https://arxiv.org/abs/2506.08027 |

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

greptile-apps

_{1 file reviewed, 2 comments}

_{Edit Code Review Agent Settings | Greptile}

greptile-apps · 2026-02-02T23:19:45Z

README.rst

 | LLama2-7B  |  Alibaba Pai     | https://mp.weixin.qq.com/s/NQT0uKXLbXyh5031zBdeBQ                                                       |
 +------------+------------------+---------------------------------------------------------------------------------------------------------+
-| T5-11B     |  JAX/T5x         | Available on request                                                                                    |
+| LLM-8B     |  Megatron Core   | https://arxiv.org/abs/2506.08027                                                                        |


greptile-apps · 2026-02-02T23:19:46Z

README.rst

 | MPT-13B    |  Mosaic Composer | https://www.databricks.com/blog/turbocharged-training-optimizing-databricks-mosaic-ai-stack-fp8         |
 +------------+------------------+---------------------------------------------------------------------------------------------------------+
-| GPT-22B    |  NeMo Framework  | Available on request                                                                                    |
+| MoE-16B    |  Megatron Core   | https://arxiv.org/abs/2506.08027                                                                        |


…t info - Add MXFP8 and NVFP4 format support to highlights and description - Update FP8 convergence table with MXFP8 results from arxiv paper - Remove outdated JAX-Toolbox links and "available on request" entries - Update Docker container versions to 26.01 - Fix DeepSpeed and Lightning integration links - Add Nemotron 3 paper to Latest News - Add quickstart notebook link after PyTorch example Signed-off-by: Santosh Bhavani <santosh.bhavani@live.com>

- Replace quickstart.ipynb link with fp8_primer.ipynb (file exists) - Fix extra whitespace in Megatron Core table rows Signed-off-by: Santosh Bhavani <santosh.bhavani@live.com>

greptile-apps

_{1 file reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}

README.rst

Signed-off-by: Santosh Bhavani <santosh.bhavani@live.com>

greptile-apps

_{1 file reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}

greptile-apps · 2026-02-10T23:26:51Z

Additional Comments (1)

README.rst
Broken reST link markup

This link uses ...__(double-underscore), while the rest of the README uses..._. In reStructuredText, __ denotes an anonymous hyperlink reference and can render incorrectly here.

* `Stable and Scalable FP8 Deep Learning Training on Blackwell | GTC 2025 <https://www.nvidia.com/en-us/on-demand/session/gtc24-s62457/>`_

greptile-apps

_{2 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}

This reverts commit 643b3d9.

greptile-apps

_{1 file reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}

greptile-apps · 2026-02-17T19:50:53Z

Additional Comments (1)

README.rst
MXFP8 results not reflected in section title or intro

The PR description states the convergence table is being "updated with MXFP8 results from arxiv paper" (arxiv.org/abs/2506.08027), and entries for LLM-8B and MoE-16B via Megatron Core are cited from that paper. However, the section heading remains FP8 Convergence and the introductory paragraph only asserts convergence for FP8 vs BF16:

"FP8 has been tested extensively... no significant difference between FP8 and BF16 training loss curves."

If the new entries represent MXFP8 convergence (not standard FP8), readers may be misled about what precision those Megatron Core results actually demonstrate. Consider either:

Adding a sentence to the intro noting MXFP8 convergence has also been validated, or
Adding a "Format" column to the table to distinguish FP8 from MXFP8 entries.

greptile-apps bot reviewed Feb 1, 2026

View reviewed changes

sbhavani force-pushed the fix/readme-updates branch from 3f01d10 to 98726c5 Compare February 2, 2026 20:36

greptile-apps bot reviewed Feb 2, 2026

View reviewed changes

sbhavani added 2 commits February 9, 2026 12:00

fix(readme): address review feedback

d5922a3

- Replace quickstart.ipynb link with fp8_primer.ipynb (file exists) - Fix extra whitespace in Megatron Core table rows Signed-off-by: Santosh Bhavani <santosh.bhavani@live.com>

sbhavani force-pushed the fix/readme-updates branch from 90af7f4 to d5922a3 Compare February 9, 2026 20:33

greptile-apps bot reviewed Feb 9, 2026

View reviewed changes

ptrendx reviewed Feb 10, 2026

View reviewed changes

README.rst Outdated Show resolved Hide resolved

Revert FP8 Primer link changes, defer to PR NVIDIA#2641

52deeda

Signed-off-by: Santosh Bhavani <santosh.bhavani@live.com>

sbhavani force-pushed the fix/readme-updates branch from cfb045e to 52deeda Compare February 10, 2026 23:24

greptile-apps bot reviewed Feb 10, 2026

View reviewed changes

ci: remove maximize-build-space from pytorch and all jobs

643b3d9

greptile-apps bot reviewed Feb 17, 2026

View reviewed changes

Revert "ci: remove maximize-build-space from pytorch and all jobs"

f0bd464

This reverts commit 643b3d9.

greptile-apps bot reviewed Feb 17, 2026

View reviewed changes

	\| LLM-8B \| Megatron Core \| https://arxiv.org/abs/2506.08027 \|
	\| LLM-8B \| Megatron Core \| https://arxiv.org/abs/2506.08027 \|

	\| MoE-16B \| Megatron Core \| https://arxiv.org/abs/2506.08027 \|
	\| MoE-16B \| Megatron Core \| https://arxiv.org/abs/2506.08027 \|

Conversation

sbhavani commented Feb 1, 2026

Description

Type of change

Changes

Checklist:

Uh oh!

greptile-apps bot commented Feb 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 4/5

Important Files Changed

Flowchart

Uh oh!

greptile-apps bot left a comment

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Feb 1, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot left a comment

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Feb 2, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot left a comment

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Feb 2, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Feb 2, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

greptile-apps bot left a comment

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot commented Feb 10, 2026

Uh oh!

greptile-apps bot left a comment

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot left a comment

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot commented Feb 17, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

greptile-apps bot commented Feb 1, 2026 •

edited

Loading