Create model-support.md for NeMoRL by snowmanwwg · Pull Request #1705 · NVIDIA-NeMo/RL

snowmanwwg · 2026-01-02T06:17:48Z

Added documentation for model support and acceleration recipes.

What does this PR do ?

Add a one line overview of what this PR aims to accomplish.

Issues

List issues that this PR closes (syntax):

Usage

You can potentially add a usage example below

# Add a code snippet demonstrating how to use this

Before your PR is "Ready for review"

Pre checks:

Make sure you read and followed Contributor guidelines
Did you write any new necessary tests?
Did you run the unit tests and functional tests locally? Visit our Testing Guide for how to run tests
Did you add or update any necessary documentation? Visit our Document Development Guide for how to write, build and test the docs.

Additional Information

...

Summary by CodeRabbit

Documentation
- Added documentation detailing broad model support for Hugging Face models (LLMs and VLMs), including supported model sizes, lists of compatible models, and acceleration optimization guidance with performance benchmarks.

_{✏️ Tip: You can customize this high-level summary in your review settings.}

Added documentation for model support and acceleration recipes. Signed-off-by: Wenwen Gao <94138584+snowmanwwg@users.noreply.github.com>

coderabbitai · 2026-01-02T06:20:36Z

📝 Walkthrough

Walkthrough

Added documentation file describing Hugging Face model support in NeMo, covering LLMs and VLMs, supported model sizes, acceleration optimization via NeMo Megatron-bridge, and lists of compatible models.

Changes

Cohort / File(s)	Summary
Documentation `docs/about/model-support.md`	New documentation file detailing NeMo's support for Hugging Face models, including model types (LLMs, VLMs), supported sizes, acceleration guidance, and reference lists of supported models.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Suggested labels

CI:docs

Suggested reviewers

terrykong

Pre-merge checks

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'Create model-support.md for NeMoRL' directly and accurately describes the main change—adding a new documentation file. It is concise, clear, and specific enough for a reviewer to understand the primary purpose of the PR.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Test Results For Major Changes	✅ Passed	PR contains only documentation changes (new markdown file), which are minor changes exempt from test result requirements.

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

coderabbitai

Actionable comments posted: 1

📜 Review details

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between f017fd8 and 1f59f37.

📒 Files selected for processing (1)

docs/about/model-support.md

🧰 Additional context used

📓 Path-based instructions (2)

docs/**/*.md

📄 CodeRabbit inference engine (CODING_GUIDELINES.md)

Update docs/index.md when a new markdown doc is added under docs/**/*.md or a markdown file is renamed, ensuring the document appears in the most appropriate section

Files:

docs/about/model-support.md

!(**/tests/**|**/test_*.py|**/test_*.sh)

📄 CodeRabbit inference engine (CODING_GUIDELINES.md)

Add the NVIDIA copyright header to all Python files and shell scripts (excluding tests). The header should include the current year

Files:

docs/about/model-support.md

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (4)

GitHub Check: build-container / main
GitHub Check: Lint check
GitHub Check: Post submodule check comment / Comment on PR
GitHub Check: Post automodel integration comment / Comment on PR

🔇 Additional comments (1)

docs/about/model-support.md (1)

1-28: Verify that docs/index.md has been updated to reference this new documentation file.

Per the coding guidelines, when a new markdown doc is added under docs/**/*.md, the docs/index.md file must be updated to include a reference in the most appropriate section. Please confirm that docs/index.md has been updated with an entry for this model-support.md documentation.

coderabbitai · 2026-01-02T06:20:39Z

+
+## Broad coverage for 🤗Hugging Face models via [NeMo AutoModel](https://github.com/NVIDIA-NeMo/Automodel)
+
+NeMo-RL support 🤗Hugging Face models from the following classes


⚠️ Potential issue | 🟡 Minor

Fix grammatical error: "support" should be "supports".

Line 5 is missing the verb conjugation.

-NeMo-RL support 🤗Hugging Face models from the following classes +NeMo-RL supports 🤗Hugging Face models from the following classes

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

NeMo-RL support 🤗Hugging Face models from the following classes

NeMo-RL supports 🤗Hugging Face models from the following classes

🤖 Prompt for AI Agents

In docs/about/model-support.md around line 5, the sentence "NeMo-RL support 🤗Hugging Face models from the following classes" uses incorrect verb conjugation; change "support" to "supports" so the sentence reads "NeMo-RL supports 🤗Hugging Face models from the following classes."

terrykong · 2026-01-20T23:57:08Z

closing in favor of #1799 which puts the changes all together

- train.py: remove the obsolete use_cache/activation-checkpointing incompatibility note. Automodel NVIDIA-NeMo#1705 (pinned 6de0c361) keeps use_cache=True for KV-sharing models under activation checkpointing, so the E4B VLM recipe's activation_checkpointing: true is safe. - dtensor_policy_worker.py (v1): remove the Gemma4 mm_token_type_ids injection. The v1 DTensor worker is being deprecated; all shipped Gemma4 recipes use _v2: true, which threads use_cache/mm_token_type_ids correctly. - setup.py: drop the Nemotron-H projection-dtype patch. A module forward-hook cannot reach the fused Mamba kernel's internal out_proj F.linear, so it cannot make nemotron-h LoRA train; the proper fix is the Automodel r0.5.0 restore-dtype change (tracked as a separate migration). - recipes: migrate enable_deepep: true -> experts: gmm + dispatcher: deepep for the gemma4/qwen3.5 automodel recipes (enable_deepep is deprecated in Automodel BackendConfig; behavior-preserving). Verified: 26B-A4B trains 20 steps, gen_kl 0.0009, gates pass. - tests: harden the E4B VLM gate with median(token_mult_prob_error) < 1.05 (observed 1.011 in CI); add a reward-ordering invariant to the reward-model env test; add hermetic unit tests for _needs_kv_cache_for_shared_layers and the Gemma4 mm_token_type_ids injection. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Signed-off-by: Shuang Yu <shuangy@nvidia.com>

Create model-support.md for NeMoRL

1f59f37

Added documentation for model support and acceleration recipes. Signed-off-by: Wenwen Gao <94138584+snowmanwwg@users.noreply.github.com>

snowmanwwg requested a review from a team as a code owner January 2, 2026 06:17

github-actions Bot added the Documentation Improvements or additions to documentation label Jan 2, 2026

snowmanwwg temporarily deployed to nemo-ci January 2, 2026 06:18 — with GitHub Actions Inactive

coderabbitai Bot reviewed Jan 2, 2026

View reviewed changes

snowmanwwg temporarily deployed to nemo-ci January 2, 2026 06:21 — with GitHub Actions Inactive

terrykong closed this Jan 20, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Create model-support.md for NeMoRL#1705

Create model-support.md for NeMoRL#1705
snowmanwwg wants to merge 1 commit into
mainfrom
snowmanwwg-patch-4

snowmanwwg commented Jan 2, 2026 •

edited by coderabbitai Bot

Loading

Uh oh!

coderabbitai Bot commented Jan 2, 2026

Walkthrough

Changes

Estimated code review effort

Suggested labels

Suggested reviewers

Uh oh!

coderabbitai Bot left a comment

Uh oh!

coderabbitai Bot Jan 2, 2026

Uh oh!

terrykong commented Jan 20, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants


		## Broad coverage for 🤗Hugging Face models via [NeMo AutoModel](https://github.com/NVIDIA-NeMo/Automodel)

		NeMo-RL support 🤗Hugging Face models from the following classes

	NeMo-RL support 🤗Hugging Face models from the following classes
	NeMo-RL supports 🤗Hugging Face models from the following classes

Conversation

snowmanwwg commented Jan 2, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What does this PR do ?

Issues

Usage

Before your PR is "Ready for review"

Additional Information

Summary by CodeRabbit

Uh oh!

coderabbitai Bot commented Jan 2, 2026

Walkthrough

Changes

Estimated code review effort

Suggested labels

Suggested reviewers

Pre-merge checks

Uh oh!

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

coderabbitai Bot Jan 2, 2026

Choose a reason for hiding this comment

Uh oh!

terrykong commented Jan 20, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

snowmanwwg commented Jan 2, 2026 •

edited by coderabbitai Bot

Loading