docs: recipe pages for Kimi-K3, Nemotron-3-Ultra and Gemma-4 - #2381
Merged
Conversation
The page drifted from what scripts/ actually ships, in both directions. Missing entirely, though each has had a launch script for a while, and each is already advertised on the docs home page: Gemma-4 26B-A4B, Gemma-4 31B JoyAI-LLM-Flash Nemotron-3-Ultra-550B-A55B They have no recipe page yet, so they link to the script the way the home page already links them. The intro now says that a bare script link means the recipe page is not written yet, since the page otherwise promises one per model. Inkling-Small has a recipe page and was on the home page, but was never listed here. Removed, because #2356 and the branch under it retire both the launch script and the recipe page: GLM-Z1-9B-0414, Moonlight-16B-A3B and MiMo-7B-RL. MiMo had no other entry, so the family row goes with it. That stack already drops these from the docs.json nav and from the GLM and Kimi family pages; the top-level table is the one place it missed. Also: every launch script is Python after #2356, so the recipe-structure section no longer offers .sh as an alternative. The home page is untouched.
Three model families ship launch scripts and are advertised on the docs home page, but had no recipe page and no entry in the supported-models table. Nemotron-3-Ultra-550B-A55B and Gemma-4 now have pages, written from their launch scripts and model configs. Both follow the six-section shape the rest of the model docs use. The Ultra page spends most of its length on one constraint, because it explains the whole layout: Mamba n_groups=8 forces attention and Mamba tensor-parallel to divide 8, so scale comes from PP and EP rather than TP, and the rollout side runs 32-GPU engines with DP-attention sized so attn_tp lands back on 8. It also records that routing replay is off for the 108-layer model and what that costs in train/rollout logprob drift. The Gemma page covers both sizes on one page, since 26B-A4B and 31B differ only in expert parallelism, token budget and a few sglang knobs. It notes that miles trains the language stack on the base VLM checkpoint without stripping the vision tower, and that the dense 31B needs the gemma4-dense branch of Megatron-Bridge. JoyAI-LLM-Flash keeps a script link rather than a page. Its script is a single recipe with no variants to compare, and nothing in the repo documents the model itself, so a page would be a paraphrase of the script. Inkling-Small had a page and a home-page entry but was never in the table. Removals are left to #2356, which already drops GLM-Z1-9B-0414, Moonlight-16B, MiMo and DeepSeek-R1 from this table along with their scripts. Keeping this PR purely additive means the two do not conflict.
Zhichenzzz
changed the title
… tables The home page linked Nemotron-3-Ultra and both Gemma-4 sizes straight at their launch scripts, because that was the only thing to link at the time. Those three now have recipe pages, so the home page links to them instead. The supported-models table now carries Kimi-K3 and matches the home page's family order and entry order, so the two read as the same list at two levels of detail rather than two lists that disagree. Gemma is split into its two sizes the way the home page has it, both pointing at the shared recipe page. Two entries still link outside the docs, and the intro now says why. JoyAI-LLM- Flash links to its launch script: one recipe, no variants, and nothing in the repo describing the model, so a page would restate the script. Kimi-K3 links to the pull request adding its support, which is backend plumbing with no launch script or model config yet, so there is no recipe to write down. The one ordering difference left is the MiMo row, which #2356 removes. Once that lands the two tables agree entry for entry.
…nkling Kimi-K3 now has a page. It documents the day-0 LoRA RL recipe from #1825: the MXFP4 to BF16 upcast, the 32-rank torch_dist conversion, and the validated 16-node launch, with the LoRA target modules and what a healthy run looks like on the GB300 validation cluster. A line at the top says the branch it comes from, since none of it is on main yet. The page leans on two things worth calling out. LoRA is not a nicety here, it is what makes the recipe fit: the base stays frozen and only the adapters carry gradients, with one shared A factor and per-expert B factors across the 896 experts. And the memory figure in "what a healthy run looks like" is the one to watch, because an adapter export that leaks shows up as allocated memory not returning to baseline after a weight sync, long before the run dies. Inkling still said its implementation "is open at" #1683. That merged on 2026-08-03 and the code is on main, so the line is gone; the page is the documentation now. Both tables link Kimi-K3 at the page instead of the pull request. MiMo and GPT-OSS move below JoyAI so the families the home page lists appear in the same order in both, with the two it does not carry after them.
Zhichenzzz
changed the title
One conflict, in models/index.md, where main and this branch each edited the top two rows of the family table. - DeepSeek: took main's row. #2391 replaced models/deepseek/deepseek.md with a DeepSeek-V3.2 recipe, so this branch's DeepSeek-R1 and DeepSeek-V3 entries both pointed at a page that no longer exists. - Thinking Machines: kept this branch's row. inkling-small.md is present on main but was never linked from the family table; adding it is part of what this branch is for. docs.json merged cleanly, keeping this branch's Kimi-K3, Nemotron-3-Ultra and Gemma entries next to main's deepseek-v3-2 and the developer nav from #2376. Checked after resolving: docs.json parses, all 86 nav entries resolve, no model page is missing from the nav, and no internal link in any docs page is dead. The welcome and models tables still agree: identical family order, with MiMo and GPT-OSS trailing in the full list only.
…ranch name The launcher has no worker subcommand and no --head-ip option; multi-node runs join an external ray cluster and set MILES_SCRIPT_EXTERNAL_RAY, like the other recipe launchers. The Megatron-Bridge branch for the dense 31B recipe is zhichen/gemma4-dense, not gemma4-dense.
Shi-Dong
approved these changes
Aug 11, 2026
Zhichenzzz
added a commit
to zianglih/miles
that referenced
this pull request
Aug 11, 2026
One conflict, in docs.json: radixark#2395 added three redirects for the deleted Platforms pages at the same spot in the redirects array where this branch adds the fp8-low-precision one. Kept all four. Checked what the merge brought in, since this branch renames a page that others link to: the three recipe pages added by radixark#2381 (Kimi-K3, Nemotron-3-Ultra, Gemma-4) carry no link to the low-precision page, so the rename sweep is still complete. All 24 links into it agree on the Low Precision RL title. docs.json parses, all 83 nav entries resolve, every redirect destination exists, and no internal link or image reference in any docs page is dead.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Brings the model docs in line with what
scripts/actually ships, and gives the three families that only had script links their recipe pages.Added
docs/models/kimi/kimi-k3.md— the Kimi-K3 LoRA RL recipe. The implementation is the open PR #1825; the page says so up front and documents that branch's scripts, the pinnedradixark/miles:kimi-k3image, and thesglang-miles-k3SGLang branch, all of which are live.docs/models/nemotron/nemotron-3-ultra.md— Nemotron-3-Ultra-550B-A55B via Megatron AutoBridge (scripts/run_nemotron_3_ultra_550b_a55b.py), including the 4-layer single-node slice.docs/models/gemma/gemma-4.mdplus a new Gemma family index — Gemma-4 26B-A4B (MoE) and 31B (dense) via the HF bridge (scripts/run_gemma_4_26b_a4b.py,scripts/run_gemma_4_31b.py).Index and nav updates
docs/index.md: the Kimi-K3, Nemotron-3-Ultra and Gemma-4 rows now point at the new recipe pages instead of a PR link and raw script links.docs/models/index.md: adds the Kimi-K3 and Nemotron-3-Ultra rows, Inkling-Small (which was missing), and the JoyAI-LLM-Flash script row.docs/docs.json: new Gemma nav group; Kimi-K3 and Nemotron-3-Ultra entries.docs/models/thinkingmachines/inkling.md: drops the merged-PR pointer.No rows are removed: GLM-Z1, Moonlight and MiMo stay listed until the #2356 stack that retires their scripts actually lands.
Checked
mainscripts/models/*, and the launcher snapshots (for Kimi-K3, from the Kimi-K3 lora RL day-0 support #1825 branch the page cites)