＋ Add to collection

Install (CLI)

npx skillmds add nvidia/nemo-mbridge-recipe-recommender

Or Copy and Paste in Your Agent chat

Please install the Agent Skill "nvidia/nemo-mbridge-recipe-recommender" for me from skillmd.com. A SkillMD skill is a SKILL.md file — plain markdown plus a few companion files — that you save into my skills directory. Use whichever fetch route fits your setup: (1) if you have a terminal, run npx -y skillmds add nvidia/nemo-mbridge-recipe-recommender; (2) otherwise fetch the file manifest at https://api.skillmd.com/api/skills/nvidia/nemo-mbridge-recipe-recommender/bundle?format=manifest and download each listed file from its "url" into nemo-mbridge-recipe-recommender/ in your skills directory; (3) if the SkillMD MCP tools are connected, call skillmd_download with slug "nvidia/nemo-mbridge-recipe-recommender" and save the returned files under nemo-mbridge-recipe-recommender/. Show me the SKILL.md before saving if you'd like me to review it. If your skills directory is read-only (e.g. the Claude desktop app), save the files to a working folder instead and tell me — I'll add them through my app's skills settings.

README badge

Show this skill's registry status in your repo:

[![SkillMD](https://skillmd.com/badge/nvidia/nemo-mbridge-recipe-recommender.svg)](https://skillmd.com/skills/nvidia/nemo-mbridge-recipe-recommender)

nemo-mbridge-recipe-recommender

Name: nemo-mbridge-recipe-recommender
Rating: 4.2 (5 reviews)
Author: NVIDIA

nvidia/nemo-mbridge-recipe-recommender · Agent Skill (multi-file)

by NVIDIA · bundle

Published Jun 29, 2026 · Last updated Jul 6, 2026

Indexes Megatron Bridge recipes and recommends the best starting config based on model, GPU count, and training goal.

SKILL.md

Auto Recipe — Recipe Index & Recommendation

This skill indexes every shipped recipe and helps users pick the right starting config, adjust parallelism, and avoid common pitfalls.

How to Use This Skill

Ask the user for: model name/size, GPU count & type, training goal (pretrain / SFT / PEFT), and sequence length (if non-default).
Look up the best-match recipe in the index below.
Recommend the recipe function name + entry-point command.
Provide adjustment advice (parallelism resizing, batch tuning, pitfalls).

First Answer Checklist

When recommending recipes, always include these distinctions before the long index details:

Library recipes under src/megatron/bridge/recipes/ are for functional training and use scripts/training/run_recipe.py.
Performance recipes under scripts/performance/ are for upper-bound throughput benchmarks. They use mock data and should not be presented as production training recipes.
For a first-time Bridge smoke test, recommend llama3_8b_sft_config with mock data via --dataset llm-pretrain-mock. Do not use llm-finetune for the setup-only tryout unless the user specifically asks for an SFT data path.
For normal SFT recommendations, use --dataset llm-finetune; for pretrain and mock validation recommendations, use --dataset llm-pretrain-mock.
After the recipe and dataset, give the required resizing rules: TP must divide num_key_value_heads, keep TP within one node unless using NVL72-class interconnect, enable SP when TP > 1, configure CP for long context, DP is implicit, and reduce micro_batch_size first on OOM.

Entry Points

Library recipes (functional training)

# Pretrain with mock data
uv run python -m torch.distributed.run --nproc_per_node=8 scripts/training/run_recipe.py \
    --recipe <recipe_function_name> \
    --dataset llm-pretrain-mock

# SFT with SQuAD
uv run python -m torch.distributed.run --nproc_per_node=8 scripts/training/run_recipe.py \
    --recipe <recipe_function_name> \
    --dataset llm-finetune

# Override any field via CLI
uv run python -m torch.distributed.run --nproc_per_node=8 scripts/training/run_recipe.py \
    --recipe llama3_8b_pretrain_config \
    --dataset llm-pretrain-mock \
    'model.tensor_model_parallel_size=2' \
    'training.global_batch_size=64'

Performance recipes (throughput benchmarks)

python scripts/performance/run_script.py \
    --recipe <model_family> \
    --gpu_type h100 \
    --num_gpus 64 \
    --data mock

See the Performance Recipe Index for important caveats before using these for anything beyond throughput benchmarking.

Recipe Unification (Coming Soon — PR #2803)

PR #2803 is unifying performance recipes into the same Python function format used by library recipes. Key changes:

Perf recipes move from scripts/performance/configs/ → src/megatron/bridge/recipes/<family>/<model>_perf.py
Each perf recipe becomes a self-contained Python function (e.g. llama3_8b_h100_bf16_pretrain_config())
The old WorkloadBaseConfig → set_workload_base_configs → get_perf_optimized_recipe pipeline is removed
Shared helpers: _benchmark_common() (50 iters, timing, TE RNG), _perf_precision() (bf16 / fp8_cs / fp8_mx / nvfp4)

Why Python, not YAML? Previous YAML-based approaches had problems: recipe logic was split across multiple indirection layers, configs were not self-contained, and the two-level pipeline made maintenance and debugging difficult. Python functions are explicit, greppable, and composable.

After #2803 lands, both library and perf recipes will be invocable through the same run_recipe.py entry point.

Library Recipe Index

All recipes live under src/megatron/bridge/recipes/. Each function returns a ConfigContainer with model, training, optimizer, and data settings.

Llama

Recipe	Mode	TP	PP	CP	SP	GPUs (min)	Seq Len
`llama2_7b_pretrain_config`	Pretrain	2	1	—	—	2	4K
`llama3_8b_pretrain_config`	Pretrain	2	1	—	✓	2	8K
`llama3_8b_16k_pretrain_config`	Pretrain	2	1	2	✓	4	16K
`llama3_8b_64k_pretrain_config`	Pretrain	2	1	4	✓	8	64K
`llama3_8b_128k_pretrain_config`	Pretrain	2	1	8	✓	16	128K
`llama3_70b_pretrain_config`	Pretrain	8	4	—	✓	32	8K
`llama3_70b_16k_pretrain_config`	Pretrain	8	4	2	✓	64	16K
`llama3_70b_64k_pretrain_config`	Pretrain	8	4	4	✓	128	64K
`llama31_405b_pretrain_config`	Pretrain	8	16	—	✓	128	8K
`llama3_8b_sft_config`	SFT	2	1	—	✓	2	8K
`llama3_70b_sft_config`	SFT	4	4	—	✓	16	8K
`llama31_405b_sft_config`	SFT	8	8	—	✓	64	8K
`llama3_8b_peft_config`	PEFT	1	1	—	—	1	8K
`llama3_70b_peft_config`	PEFT	2	4	—	✓	8	8K
`llama31_405b_peft_config`	PEFT	4	8	—	✓	32	8K

Qwen2 / Qwen2.5

Recipe	Mode	TP	PP	Sizes
`qwen2_*_{pretrain,sft,peft}_config`	All	1–8	1–4	500M, 1.5B, 7B, 14B, 32B, 72B
`qwen25_*_{pretrain,sft,peft}_config`	All	1–8	1–4	500M, 1.5B, 3B, 7B, 14B, 32B, 72B

Qwen3 (Dense)

Recipe	Mode	TP	PP	CP	Sizes
`qwen3_*_pretrain_config`	Pretrain	1–8	1–2	—	600M–32B
`qwen3_*_sft_config`	SFT	1–8	1–2	—	600M–32B
`qwen3_600m_sft_128k_config`	SFT	1	1	8	600M (128K seq)
`qwen3_*_peft_config`	PEFT	1	1	—	600M–32B

Qwen3 MoE

Recipe	Mode	TP	PP	EP	CP	GPUs
`qwen3_30b_a3b_pretrain_config`	Pretrain	1	1	8	—	8
`qwen3_30b_a3b_sft_config`	SFT	1	1	8	—	8
`qwen3_30b_a3b_peft_config`	PEFT	1	1	1	—	1
`qwen3_235b_a22b_pretrain_config`	Pretrain	4	16	8	2	512+
`qwen3_235b_a22b_sft_config`	SFT	4	8	8	—	256
`qwen3_235b_a22b_peft_config`	PEFT	1	4	4	—	16

Qwen3-Next

Recipe	Mode	TP	PP	EP
`qwen3_next_80b_a3b_pretrain_config`	Pretrain	1	4	8
`qwen3_next_80b_a3b_sft_config`	SFT	1	2	8
`qwen3_next_80b_a3b_peft_config`	PEFT	1	1	4

DeepSeek

Recipe	Mode	TP	PP	EP	GPUs
`deepseek_v2_lite_pretrain_config`	Pretrain	1	1	8	8
`deepseek_v2_pretrain_config`	Pretrain	1	4	32	128
`deepseek_v3_pretrain_config`	Pretrain	2	16	64	2048
`deepseek_v3_pretrain_config_32nodes`	Pretrain	2	8	32	256

GLM-4.5

Recipe	Mode	TP	PP	EP	GPUs
`glm45_355b_pretrain_config`	Pretrain	2	8	16	256
`glm45_air_106b_pretrain_config`	Pretrain	1	4	8	32
`glm45_355b_sft_config`	SFT	2	8	16	256
`glm45_air_106b_sft_config`	SFT	1	4	8	32
`glm45_355b_peft_config`	PEFT	2	4	4	32
`glm45_air_106b_peft_config`	PEFT	1	2	4	8

Gemma

Recipe	Mode	TP	PP	Sizes
`gemma2_*_{pretrain,sft,peft}_config`	All	2–8	1–2	2B, 9B, 27B
`gemma3_1b_{pretrain,sft,peft}_config`	All	1	1	1B (32K seq)

NemotronH / Nemotron

Recipe	Mode	TP	PP	EP	Notes
`nemotronh_{4b,8b,47b,56b}_*_config`	P/S/PEFT	1–8	1–4	—	Dense SSM-hybrid
`nemotron_3_nano_*_config`	P/S/PEFT	varies	1	8	MoE + Mamba
`nemotron_3_super_*_config`	P/S/PEFT	4	1	8	MoE + Mamba, ~40% CUDA graph gain
`nemotron_nano_{9b,12b}_v2_*_config`	P/S/PEFT	varies	1	—	Dense

Other Models

Recipe	Mode	Notes
`moonlight_16b_{pretrain,sft,peft}_config`	All	MoE EP=8
`olmoe_7b_{pretrain,sft,peft}_config`	All	MoE EP=8
`ministral3_{3b,8b,14b}_{sft,peft}_config`	SFT/PEFT	Dense
`gpt_oss_20b_*_config`	All	MoE + FP8/MXFP8 variants
`gpt_oss_120b_*_config`	All	MoE
`vanilla_gpt_pretrain_config`	Pretrain	MLM/Bridge parity baseline
`gpt3_175b_pretrain_config`	Pretrain	TP=4, PP=8, VP=6
`kimi_k2_pretrain_config`	Pretrain	1T MoE, TP=2 PP=16 EP=32

VLM Recipes

Recipe	Mode	TP	PP	EP	GPUs
`gemma3_vl_{4b,12b,27b}_{sft,peft}_config`	SFT/PEFT	1–8	1–2	—	1–16
`qwen25_vl_{3b,7b,32b,72b}_{sft,peft}_config`	SFT/PEFT	1–8	1–4	—	1–32
`qwen3_vl_{8b,30b_a3b,235b_a22b}_{sft,peft}_config`	SFT/PEFT	1–4	1–8	1–32	1–512
`qwen35_vl_*_{sft,peft}_config`	SFT/PEFT	varies	varies	varies	varies
`glm_45v_{sft,peft}_config`	SFT/PEFT	1	8	4–16	64–512
`nemotron_nano_v2_vl_12b_{sft,peft}_config`	SFT/PEFT	2–4	1	—	8

Diffusion Recipes

Recipe	Mode	TP	CP
`wan_1_3B_{pretrain,sft}_config`	P/SFT	1	8
`wan_14B_{pretrain,sft}_config`	P/SFT	2	4
`flux_12b_{pretrain,sft}_config`	P/SFT	2	1

Performance Recipe Index

All perf recipes live under scripts/performance/. They are invoked via run_script.py and use WorkloadBaseConfig presets per GPU type.

Important: Perf recipes are designed for upper-bound throughput benchmarks, not production training. They run 50 iterations on mock data by default. Throughput numbers are aspirational targets, not validated convergence configs.

Llama 3 / 3.1

Model	GPUs	GPU Types	Key Features
Llama 3 8B	8	H100, B200, B300, GB200, GB300, R100	CUDA graphs (local), FSDP on GB variants
Llama 3 70B	64	H100, B200, B300, GB200, GB300	TP comm overlap (userbuffers), FSDP, CUDA graphs
Llama 3.1 405B	128–1024	H100, B200, B300, GB200, GB300	TP+CP comm overlap (userbuffers), FSDP, heavy PP/VP

SFT/LoRA variants also exist (e.g. 8B SFT with packed sequences, 70B SFT on 32 GPUs).

DeepSeek V3

Model	GPUs	GPU Types	Key Features
DeepSeek V3 (671B MoE)	256–1024	H100, B200, B300, GB200, GB300	HybridEP dispatcher, MLA recompute, CUDA graphs (TE scoped)

Qwen3 MoE

Model	GPUs	GPU Types	Key Features
Qwen3 30B-A3B	8–16	H100, B200, B300, GB200, GB300	MoE alltoall/flex dispatcher
Qwen3 235B-A22B	64–256	H100, B200, B300, GB200, GB300	TP comm overlap, CUDA graphs, MoE a2a overlap
Qwen3-Next 80B-A3B	64–128	H100, B200, B300, GB200, GB300	EP 64–128

Qwen3-VL

Model	GPUs	GPU Types	Key Features
Qwen3-VL 30B-A3B	8–16	H100, B200, B300, GB200, GB300	VLM + MoE
Qwen3-VL 235B-A22B	64–256	H100, B200, B300, GB200, GB300	VLM + MoE, TP comm overlap

Kimi K2

Model	GPUs	GPU Types	Key Features
Kimi K2 (1T MoE)	256–1024	H100, B200, B300, GB200, GB300	Muon/Adam optimizer, HybridEP, pipeline layout helpers

NemotronH

Model	GPUs	GPU Types	Key Features
Nemotron 3 Nano (30B MoE+Mamba)	8–16	H100, B200, B300, GB200, GB300	TE CUDA graphs (attn+mamba+moe), HybridEP
Nemotron 3 Super	64	H100, B200, B300, GB200, GB300	TE CUDA graphs, EP=64
NemotronH 56B	64	H100, B200, B300	TP=2–8, TE graphs (mamba+attn)

GPT-OSS

Model	GPUs	GPU Types	Key Features
GPT-OSS 120B	64	H100, B200, GB200	EP=64, HybridEP on GB200

Recommendation Decision Tree

User wants to train a model
│
├─ Know the model name?
│   ├─ Yes → Look up in Library Recipe Index above
│   │   ├─ Has a recipe for their size + mode? → Use it directly
│   │   └─ No exact match? → Use closest size, adjust parallelism
│   └─ No → Ask for model name, size, and HF model ID
│
├─ What's the training goal?
│   ├─ Pretrain → Use *_pretrain_config
│   ├─ SFT (full fine-tune) → Use *_sft_config
│   └─ PEFT (LoRA/DoRA) → Use *_peft_config (lowest GPU requirement)
│
├─ How many GPUs?
│   ├─ 1 GPU → Only PEFT recipes work (TP=1, PP=1)
│   ├─ 8 GPUs (1 node) → Most 8B–16B models, small MoE (EP=8)
│   ├─ 16–64 GPUs → 70B dense, medium MoE
│   └─ 128+ GPUs → 405B+, large MoE (DeepSeek V3, Kimi K2)
│
├─ Want throughput benchmarks?
│   ├─ Yes → Use perf recipes (scripts/performance/)
│   │   └─ ⚠️ These run on mock data for upper-bound perf only
│   └─ No → Use library recipes (scripts/training/run_recipe.py)
│
└─ Long context?
    ├─ > 8K → Need CP (context parallelism), check *_16k / *_64k / *_128k variants
    └─ ≤ 8K → Default recipes work

Adjustment Advice (When Recommending)

Parallelism Resizing Rules

When the user's GPU count differs from the recipe default:

TP must divide num_key_value_heads (GQA constraint). E.g. if num_key_value_heads=8, valid TP = {1, 2, 4, 8}.
TP should stay within a single node (NVLink). TP > 8 requires inter-node NVLink (e.g., GB200 NVL72).
PP adds pipeline bubbles. Minimize PP; only increase when TP alone can't fit the model. Use VP (virtual pipeline) to mitigate bubble overhead.
EP doesn't reduce dense-layer memory. Only expert parameters shard with EP. Shared attention/embeddings are replicated. For "OOM with MoE", increase EP first, not TP.
SP should be True whenever TP > 1. It eliminates redundant activation copies and is essentially free.
CP requires all-to-all or ring attention. Check cp_comm_type. For GQA models, a2a+p2p hierarchical CP allows CP > num_kv_heads.
world_size = DP × TP × PP × CP × EP. DP is implicit. Make sure the product of explicit parallelisms divides your total GPU count.

Batch Size Tuning

Start with the recipe's micro_batch_size. If OOM, reduce to 1.
global_batch_size determines learning dynamics. Scale with DP: GBS = micro_batch_size × DP × gradient_accumulation_steps.
For MoE, micro_batch_size=1 is typical at scale.

Common Pitfalls to Warn About

Pitfall	Symptom	Fix
TP > num_kv_heads	Crash: "TP must divide num_query_groups"	Reduce TP to a divisor of num_kv_heads
PP without VP	Poor throughput (large bubble)	Set `virtual_pipeline_model_parallel_size`
EP too low for large MoE	OOM on expert params	Increase EP; each expert lives on EP/num_experts ranks
CUDA graphs + packed sequences	Assert: "CUDA graph accepts only Tensor inputs"	Disable packing or use `local` full-iteration graphs
CUDA graphs + full recompute	Assert: "full recompute only with full iteration CUDA graph"	Disable recompute or switch to `local` impl
`use_te_rng_tracker` not set	Assert on provider init when CUDA graphs enabled	Set `cfg.model.use_te_rng_tracker = True` and `cfg.rng.te_rng_tracker = True`
FSDP + TP > 1 on H100	Possible comm bottleneck	Prefer FSDP with TP=1 or TP=2 on H100; FSDP shines on GB/B-series
Long context without CP	OOM on activations	Add CP=2/4/8; use `_16k`, `_64k`, or `*_128k` recipe variants
MoE `overlap_grad_reduce` on H100	May hurt perf (False in many H100 presets)	Set `overlap_grad_reduce=False` for MoE on H100
VLM SFT missing image data	Runs but produces garbage	Provide actual multimodal dataset or use mock VLM data
Qwen35-VL MoE FSDP	Tested on Blackwell only	May not work on H100; validate first

Recipe Override Examples

# Scale Llama3 8B from 2 GPUs to 8 GPUs (increase DP)
uv run python -m torch.distributed.run --nproc_per_node=8 scripts/training/run_recipe.py \
    --recipe llama3_8b_pretrain_config \
    --dataset llm-pretrain-mock

# Reduce parallelism for Qwen3-MoE 30B to fit on 4 GPUs
uv run python -m torch.distributed.run --nproc_per_node=4 scripts/training/run_recipe.py \
    --recipe qwen3_30b_a3b_sft_config \
    --dataset llm-finetune \
    'model.expert_model_parallel_size=4'

# Add long context to an existing recipe
uv run python -m torch.distributed.run --nproc_per_node=8 scripts/training/run_recipe.py \
    --recipe llama3_8b_pretrain_config \
    --dataset llm-pretrain-mock \
    'model.seq_length=32768' \
    'model.context_parallel_size=4'

# Enable CUDA graphs on any recipe
uv run python -m torch.distributed.run --nproc_per_node=8 scripts/training/run_recipe.py \
    --recipe qwen3_30b_a3b_pretrain_config \
    --dataset llm-pretrain-mock \
    'model.cuda_graph_impl=transformer_engine' \
    'model.cuda_graph_scope=[attn,moe_router,moe_preprocess]' \
    'model.use_te_rng_tracker=True' \
    'rng.te_rng_tracker=True'

Quick Reference: Which Recipe for My Situation?

I want to...	Start with	GPUs needed
Try Bridge for the first time	`llama3_8b_sft_config` + mock data	2
Fine-tune a 7-8B model	`llama3_8b_sft_config` or `qwen3_8b_sft_config`	2–8
LoRA on 1 GPU	`llama3_8b_peft_config` or `qwen3_8b_peft_config`	1
Pretrain a dense 70B	`llama3_70b_pretrain_config`	32–64
Train a small MoE	`qwen3_30b_a3b_pretrain_config`	8
Train a large MoE (235B+)	`qwen3_235b_a22b_pretrain_config`	256–512
Benchmark throughput	Perf recipes via `run_script.py`	Varies
Long-context training	`llama3_8b_128k_pretrain_config` or add CP override	16+
VLM fine-tuning	`qwen3_vl_8b_sft_config` or `gemma3_vl_*_sft_config`	4–8
Diffusion training	`wan_1_3B_pretrain_config` or `flux_12b_pretrain_config`	8

Code Anchors

What	Path
Library recipes root	`src/megatron/bridge/recipes/`
Recipe `__init__.py` (all exports)	`src/megatron/bridge/recipes/__init__.py`
Common recipe helpers	`src/megatron/bridge/recipes/common.py`
Training entry point	`scripts/training/run_recipe.py`
Perf recipes root	`scripts/performance/`
Perf entry point	`scripts/performance/run_script.py`
Perf workload configs	`scripts/performance/configs/<family>/`
Perf overrides (benchmark defaults)	`scripts/performance/utils/overrides.py`

---
name: nemo-mbridge-recipe-recommender
description: Indexes Megatron Bridge recipes and recommends the best starting config based on model, GPU count, and training goal.
license: Apache-2.0
---

# Auto Recipe — Recipe Index & Recommendation

This skill indexes every shipped recipe and helps users pick the right starting
config, adjust parallelism, and avoid common pitfalls.

## How to Use This Skill

1. Ask the user for: **model name/size**, **GPU count & type**, **training goal**
   (pretrain / SFT / PEFT), and **sequence length** (if non-default).
2. Look up the best-match recipe in the index below.
3. Recommend the recipe function name + entry-point command.
4. Provide adjustment advice (parallelism resizing, batch tuning, pitfalls).

## First Answer Checklist

When recommending recipes, always include these distinctions before the long
index details:

1. **Library recipes** under `src/megatron/bridge/recipes/` are for functional
   training and use `scripts/training/run_recipe.py`.
2. **Performance recipes** under `scripts/performance/` are for upper-bound
   throughput benchmarks. They use mock data and should not be presented as
   production training recipes.
3. For a first-time Bridge smoke test, recommend `llama3_8b_sft_config` with
   mock data via `--dataset llm-pretrain-mock`. Do not use `llm-finetune` for
   the setup-only tryout unless the user specifically asks for an SFT data path.
4. For normal SFT recommendations, use `--dataset llm-finetune`; for pretrain
   and mock validation recommendations, use `--dataset llm-pretrain-mock`.
5. After the recipe and dataset, give the required resizing rules: TP must
   divide `num_key_value_heads`, keep TP within one node unless using
   NVL72-class interconnect, enable SP when TP > 1, configure CP for long
   context, DP is implicit, and reduce `micro_batch_size` first on OOM.

---

## Entry Points

### Library recipes (functional training)

```bash
# Pretrain with mock data
uv run python -m torch.distributed.run --nproc_per_node=8 scripts/training/run_recipe.py \
    --recipe <recipe_function_name> \
    --dataset llm-pretrain-mock

# SFT with SQuAD
uv run python -m torch.distributed.run --nproc_per_node=8 scripts/training/run_recipe.py \
    --recipe <recipe_function_name> \
    --dataset llm-finetune

# Override any field via CLI
uv run python -m torch.distributed.run --nproc_per_node=8 scripts/training/run_recipe.py \
    --recipe llama3_8b_pretrain_config \
    --dataset llm-pretrain-mock \
    'model.tensor_model_parallel_size=2' \
    'training.global_batch_size=64'
```

### Performance recipes (throughput benchmarks)

```bash
python scripts/performance/run_script.py \
    --recipe <model_family> \
    --gpu_type h100 \
    --num_gpus 64 \
    --data mock
```

See the Performance Recipe Index for important caveats before using these for anything beyond throughput benchmarking.

---

## Recipe Unification (Coming Soon — PR #2803)

PR [#2803](https://github.com/NVIDIA-NeMo/Megatron-Bridge/pull/2803) is
unifying performance recipes into the same **Python function** format used by
library recipes. Key changes:

- Perf recipes move from `scripts/performance/configs/` → `src/megatron/bridge/recipes/<family>/<model>_perf.py`
- Each perf recipe becomes a **self-contained Python function** (e.g. `llama3_8b_h100_bf16_pretrain_config()`)
- The old `WorkloadBaseConfig` → `set_workload_base_configs` → `get_perf_optimized_recipe` pipeline is removed
- Shared helpers: `_benchmark_common()` (50 iters, timing, TE RNG), `_perf_precision()` (bf16 / fp8_cs / fp8_mx / nvfp4)

**Why Python, not YAML?** Previous YAML-based approaches had problems:
recipe logic was split across multiple indirection layers, configs were not
self-contained, and the two-level pipeline made maintenance and debugging
difficult. Python functions are explicit, greppable, and composable.

After #2803 lands, both library and perf recipes will be invocable through the
same `run_recipe.py` entry point.

---

## Library Recipe Index

All recipes live under `src/megatron/bridge/recipes/`. Each function returns a
`ConfigContainer` with model, training, optimizer, and data settings.

### Llama

| Recipe | Mode | TP | PP | CP | SP | GPUs (min) | Seq Len |
|--------|------|----|----|----|----|------------|---------|
| `llama2_7b_pretrain_config` | Pretrain | 2 | 1 | — | — | 2 | 4K |
| `llama3_8b_pretrain_config` | Pretrain | 2 | 1 | — | ✓ | 2 | 8K |
| `llama3_8b_16k_pretrain_config` | Pretrain | 2 | 1 | 2 | ✓ | 4 | 16K |
| `llama3_8b_64k_pretrain_config` | Pretrain | 2 | 1 | 4 | ✓ | 8 | 64K |
| `llama3_8b_128k_pretrain_config` | Pretrain | 2 | 1 | 8 | ✓ | 16 | 128K |
| `llama3_70b_pretrain_config` | Pretrain | 8 | 4 | — | ✓ | 32 | 8K |
| `llama3_70b_16k_pretrain_config` | Pretrain | 8 | 4 | 2 | ✓ | 64 | 16K |
| `llama3_70b_64k_pretrain_config` | Pretrain | 8 | 4 | 4 | ✓ | 128 | 64K |
| `llama31_405b_pretrain_config` | Pretrain | 8 | 16 | — | ✓ | 128 | 8K |
| `llama3_8b_sft_config` | SFT | 2 | 1 | — | ✓ | 2 | 8K |
| `llama3_70b_sft_config` | SFT | 4 | 4 | — | ✓ | 16 | 8K |
| `llama31_405b_sft_config` | SFT | 8 | 8 | — | ✓ | 64 | 8K |
| `llama3_8b_peft_config` | PEFT | 1 | 1 | — | — | 1 | 8K |
| `llama3_70b_peft_config` | PEFT | 2 | 4 | — | ✓ | 8 | 8K |
| `llama31_405b_peft_config` | PEFT | 4 | 8 | — | ✓ | 32 | 8K |

### Qwen2 / Qwen2.5

| Recipe | Mode | TP | PP | Sizes |
|--------|------|----|----|-------|
| `qwen2_*_{pretrain,sft,peft}_config` | All | 1–8 | 1–4 | 500M, 1.5B, 7B, 14B, 32B, 72B |
| `qwen25_*_{pretrain,sft,peft}_config` | All | 1–8 | 1–4 | 500M, 1.5B, 3B, 7B, 14B, 32B, 72B |

### Qwen3 (Dense)

| Recipe | Mode | TP | PP | CP | Sizes |
|--------|------|----|----|-----|-------|
| `qwen3_*_pretrain_config` | Pretrain | 1–8 | 1–2 | — | 600M–32B |
| `qwen3_*_sft_config` | SFT | 1–8 | 1–2 | — | 600M–32B |
| `qwen3_600m_sft_128k_config` | SFT | 1 | 1 | 8 | 600M (128K seq) |
| `qwen3_*_peft_config` | PEFT | 1 | 1 | — | 600M–32B |

### Qwen3 MoE

| Recipe | Mode | TP | PP | EP | CP | GPUs |
|--------|------|----|----|----|----|------|
| `qwen3_30b_a3b_pretrain_config` | Pretrain | 1 | 1 | 8 | — | 8 |
| `qwen3_30b_a3b_sft_config` | SFT | 1 | 1 | 8 | — | 8 |
| `qwen3_30b_a3b_peft_config` | PEFT | 1 | 1 | 1 | — | 1 |
| `qwen3_235b_a22b_pretrain_config` | Pretrain | 4 | 16 | 8 | 2 | 512+ |
| `qwen3_235b_a22b_sft_config` | SFT | 4 | 8 | 8 | — | 256 |
| `qwen3_235b_a22b_peft_config` | PEFT | 1 | 4 | 4 | — | 16 |

### Qwen3-Next

| Recipe | Mode | TP | PP | EP |
|--------|------|----|----|-----|
| `qwen3_next_80b_a3b_pretrain_config` | Pretrain | 1 | 4 | 8 |
| `qwen3_next_80b_a3b_sft_config` | SFT | 1 | 2 | 8 |
| `qwen3_next_80b_a3b_peft_config` | PEFT | 1 | 1 | 4 |

### DeepSeek

| Recipe | Mode | TP | PP | EP | GPUs |
|--------|------|----|----|-----|------|
| `deepseek_v2_lite_pretrain_config` | Pretrain | 1 | 1 | 8 | 8 |
| `deepseek_v2_pretrain_config` | Pretrain | 1 | 4 | 32 | 128 |
| `deepseek_v3_pretrain_config` | Pretrain | 2 | 16 | 64 | 2048 |
| `deepseek_v3_pretrain_config_32nodes` | Pretrain | 2 | 8 | 32 | 256 |

### GLM-4.5

| Recipe | Mode | TP | PP | EP | GPUs |
|--------|------|----|----|-----|------|
| `glm45_355b_pretrain_config` | Pretrain | 2 | 8 | 16 | 256 |
| `glm45_air_106b_pretrain_config` | Pretrain | 1 | 4 | 8 | 32 |
| `glm45_355b_sft_config` | SFT | 2 | 8 | 16 | 256 |
| `glm45_air_106b_sft_config` | SFT | 1 | 4 | 8 | 32 |
| `glm45_355b_peft_config` | PEFT | 2 | 4 | 4 | 32 |
| `glm45_air_106b_peft_config` | PEFT | 1 | 2 | 4 | 8 |

### Gemma

| Recipe | Mode | TP | PP | Sizes |
|--------|------|----|----|-------|
| `gemma2_*_{pretrain,sft,peft}_config` | All | 2–8 | 1–2 | 2B, 9B, 27B |
| `gemma3_1b_{pretrain,sft,peft}_config` | All | 1 | 1 | 1B (32K seq) |

### NemotronH / Nemotron

| Recipe | Mode | TP | PP | EP | Notes |
|--------|------|----|----|-----|-------|
| `nemotronh_{4b,8b,47b,56b}_*_config` | P/S/PEFT | 1–8 | 1–4 | — | Dense SSM-hybrid |
| `nemotron_3_nano_*_config` | P/S/PEFT | varies | 1 | 8 | MoE + Mamba |
| `nemotron_3_super_*_config` | P/S/PEFT | 4 | 1 | 8 | MoE + Mamba, ~40% CUDA graph gain |
| `nemotron_nano_{9b,12b}_v2_*_config` | P/S/PEFT | varies | 1 | — | Dense |

### Other Models

| Recipe | Mode | Notes |
|--------|------|-------|
| `moonlight_16b_{pretrain,sft,peft}_config` | All | MoE EP=8 |
| `olmoe_7b_{pretrain,sft,peft}_config` | All | MoE EP=8 |
| `ministral3_{3b,8b,14b}_{sft,peft}_config` | SFT/PEFT | Dense |
| `gpt_oss_20b_*_config` | All | MoE + FP8/MXFP8 variants |
| `gpt_oss_120b_*_config` | All | MoE |
| `vanilla_gpt_pretrain_config` | Pretrain | MLM/Bridge parity baseline |
| `gpt3_175b_pretrain_config` | Pretrain | TP=4, PP=8, VP=6 |
| `kimi_k2_pretrain_config` | Pretrain | 1T MoE, TP=2 PP=16 EP=32 |

### VLM Recipes

| Recipe | Mode | TP | PP | EP | GPUs |
|--------|------|----|----|-----|------|
| `gemma3_vl_{4b,12b,27b}_{sft,peft}_config` | SFT/PEFT | 1–8 | 1–2 | — | 1–16 |
| `qwen25_vl_{3b,7b,32b,72b}_{sft,peft}_config` | SFT/PEFT | 1–8 | 1–4 | — | 1–32 |
| `qwen3_vl_{8b,30b_a3b,235b_a22b}_{sft,peft}_config` | SFT/PEFT | 1–4 | 1–8 | 1–32 | 1–512 |
| `qwen35_vl_*_{sft,peft}_config` | SFT/PEFT | varies | varies | varies | varies |
| `glm_45v_{sft,peft}_config` | SFT/PEFT | 1 | 8 | 4–16 | 64–512 |
| `nemotron_nano_v2_vl_12b_{sft,peft}_config` | SFT/PEFT | 2–4 | 1 | — | 8 |

### Diffusion Recipes

| Recipe | Mode | TP | CP |
|--------|------|----|----|
| `wan_1_3B_{pretrain,sft}_config` | P/SFT | 1 | 8 |
| `wan_14B_{pretrain,sft}_config` | P/SFT | 2 | 4 |
| `flux_12b_{pretrain,sft}_config` | P/SFT | 2 | 1 |

---

## Performance Recipe Index

All perf recipes live under `scripts/performance/`. They are invoked via
`run_script.py` and use `WorkloadBaseConfig` presets per GPU type.

> **Important:** Perf recipes are designed for **upper-bound throughput
> benchmarks**, not production training. They run **50 iterations** on **mock
> data** by default. Throughput numbers are aspirational targets, not validated
> convergence configs.

### Llama 3 / 3.1

| Model | GPUs | GPU Types | Key Features |
|-------|------|-----------|--------------|
| Llama 3 8B | 8 | H100, B200, B300, GB200, GB300, R100 | CUDA graphs (local), FSDP on GB variants |
| Llama 3 70B | 64 | H100, B200, B300, GB200, GB300 | TP comm overlap (userbuffers), FSDP, CUDA graphs |
| Llama 3.1 405B | 128–1024 | H100, B200, B300, GB200, GB300 | TP+CP comm overlap (userbuffers), FSDP, heavy PP/VP |

SFT/LoRA variants also exist (e.g. 8B SFT with packed sequences, 70B SFT on 32 GPUs).

### DeepSeek V3

| Model | GPUs | GPU Types | Key Features |
|-------|------|-----------|--------------|
| DeepSeek V3 (671B MoE) | 256–1024 | H100, B200, B300, GB200, GB300 | HybridEP dispatcher, MLA recompute, CUDA graphs (TE scoped) |

### Qwen3 MoE

| Model | GPUs | GPU Types | Key Features |
|-------|------|-----------|--------------|
| Qwen3 30B-A3B | 8–16 | H100, B200, B300, GB200, GB300 | MoE alltoall/flex dispatcher |
| Qwen3 235B-A22B | 64–256 | H100, B200, B300, GB200, GB300 | TP comm overlap, CUDA graphs, MoE a2a overlap |
| Qwen3-Next 80B-A3B | 64–128 | H100, B200, B300, GB200, GB300 | EP 64–128 |

### Qwen3-VL

| Model | GPUs | GPU Types | Key Features |
|-------|------|-----------|--------------|
| Qwen3-VL 30B-A3B | 8–16 | H100, B200, B300, GB200, GB300 | VLM + MoE |
| Qwen3-VL 235B-A22B | 64–256 | H100, B200, B300, GB200, GB300 | VLM + MoE, TP comm overlap |

### Kimi K2

| Model | GPUs | GPU Types | Key Features |
|-------|------|-----------|--------------|
| Kimi K2 (1T MoE) | 256–1024 | H100, B200, B300, GB200, GB300 | Muon/Adam optimizer, HybridEP, pipeline layout helpers |

### NemotronH

| Model | GPUs | GPU Types | Key Features |
|-------|------|-----------|--------------|
| Nemotron 3 Nano (30B MoE+Mamba) | 8–16 | H100, B200, B300, GB200, GB300 | TE CUDA graphs (attn+mamba+moe), HybridEP |
| Nemotron 3 Super | 64 | H100, B200, B300, GB200, GB300 | TE CUDA graphs, EP=64 |
| NemotronH 56B | 64 | H100, B200, B300 | TP=2–8, TE graphs (mamba+attn) |

### GPT-OSS

| Model | GPUs | GPU Types | Key Features |
|-------|------|-----------|--------------|
| GPT-OSS 120B | 64 | H100, B200, GB200 | EP=64, HybridEP on GB200 |

---

## Recommendation Decision Tree

```text
User wants to train a model
│
├─ Know the model name?
│   ├─ Yes → Look up in Library Recipe Index above
│   │   ├─ Has a recipe for their size + mode? → Use it directly
│   │   └─ No exact match? → Use closest size, adjust parallelism
│   └─ No → Ask for model name, size, and HF model ID
│
├─ What's the training goal?
│   ├─ Pretrain → Use *_pretrain_config
│   ├─ SFT (full fine-tune) → Use *_sft_config
│   └─ PEFT (LoRA/DoRA) → Use *_peft_config (lowest GPU requirement)
│
├─ How many GPUs?
│   ├─ 1 GPU → Only PEFT recipes work (TP=1, PP=1)
│   ├─ 8 GPUs (1 node) → Most 8B–16B models, small MoE (EP=8)
│   ├─ 16–64 GPUs → 70B dense, medium MoE
│   └─ 128+ GPUs → 405B+, large MoE (DeepSeek V3, Kimi K2)
│
├─ Want throughput benchmarks?
│   ├─ Yes → Use perf recipes (scripts/performance/)
│   │   └─ ⚠️ These run on mock data for upper-bound perf only
│   └─ No → Use library recipes (scripts/training/run_recipe.py)
│
└─ Long context?
    ├─ > 8K → Need CP (context parallelism), check *_16k / *_64k / *_128k variants
    └─ ≤ 8K → Default recipes work
```

---

## Adjustment Advice (When Recommending)

### Parallelism Resizing Rules

When the user's GPU count differs from the recipe default:

1. **TP must divide `num_key_value_heads`** (GQA constraint). E.g. if
   `num_key_value_heads=8`, valid TP = {1, 2, 4, 8}.
2. **TP should stay within a single node** (NVLink). TP > 8 requires
   inter-node NVLink (e.g., GB200 NVL72).
3. **PP adds pipeline bubbles.** Minimize PP; only increase when TP alone can't
   fit the model. Use VP (virtual pipeline) to mitigate bubble overhead.
4. **EP doesn't reduce dense-layer memory.** Only expert parameters shard with
   EP. Shared attention/embeddings are replicated. For "OOM with MoE", increase
   EP first, not TP.
5. **SP should be True whenever TP > 1.** It eliminates redundant activation
   copies and is essentially free.
6. **CP requires all-to-all or ring attention.** Check `cp_comm_type`. For
   GQA models, `a2a+p2p` hierarchical CP allows CP > num_kv_heads.
7. **world_size = DP × TP × PP × CP × EP.** DP is implicit. Make sure the
   product of explicit parallelisms divides your total GPU count.

### Batch Size Tuning

- Start with the recipe's `micro_batch_size`. If OOM, reduce to 1.
- `global_batch_size` determines learning dynamics. Scale with DP:
  `GBS = micro_batch_size × DP × gradient_accumulation_steps`.
- For MoE, `micro_batch_size=1` is typical at scale.

### Common Pitfalls to Warn About

| Pitfall | Symptom | Fix |
|---------|---------|-----|
| TP > num_kv_heads | Crash: "TP must divide num_query_groups" | Reduce TP to a divisor of num_kv_heads |
| PP without VP | Poor throughput (large bubble) | Set `virtual_pipeline_model_parallel_size` |
| EP too low for large MoE | OOM on expert params | Increase EP; each expert lives on EP/num_experts ranks |
| CUDA graphs + packed sequences | Assert: "CUDA graph accepts only Tensor inputs" | Disable packing or use `local` full-iteration graphs |
| CUDA graphs + full recompute | Assert: "full recompute only with full iteration CUDA graph" | Disable recompute or switch to `local` impl |
| `use_te_rng_tracker` not set | Assert on provider init when CUDA graphs enabled | Set `cfg.model.use_te_rng_tracker = True` and `cfg.rng.te_rng_tracker = True` |
| FSDP + TP > 1 on H100 | Possible comm bottleneck | Prefer FSDP with TP=1 or TP=2 on H100; FSDP shines on GB/B-series |
| Long context without CP | OOM on activations | Add CP=2/4/8; use `*_16k`, `*_64k`, or `*_128k` recipe variants |
| MoE `overlap_grad_reduce` on H100 | May hurt perf (False in many H100 presets) | Set `overlap_grad_reduce=False` for MoE on H100 |
| VLM SFT missing image data | Runs but produces garbage | Provide actual multimodal dataset or use mock VLM data |
| Qwen35-VL MoE FSDP | Tested on Blackwell only | May not work on H100; validate first |

### Recipe Override Examples

```bash
# Scale Llama3 8B from 2 GPUs to 8 GPUs (increase DP)
uv run python -m torch.distributed.run --nproc_per_node=8 scripts/training/run_recipe.py \
    --recipe llama3_8b_pretrain_config \
    --dataset llm-pretrain-mock

# Reduce parallelism for Qwen3-MoE 30B to fit on 4 GPUs
uv run python -m torch.distributed.run --nproc_per_node=4 scripts/training/run_recipe.py \
    --recipe qwen3_30b_a3b_sft_config \
    --dataset llm-finetune \
    'model.expert_model_parallel_size=4'

# Add long context to an existing recipe
uv run python -m torch.distributed.run --nproc_per_node=8 scripts/training/run_recipe.py \
    --recipe llama3_8b_pretrain_config \
    --dataset llm-pretrain-mock \
    'model.seq_length=32768' \
    'model.context_parallel_size=4'

# Enable CUDA graphs on any recipe
uv run python -m torch.distributed.run --nproc_per_node=8 scripts/training/run_recipe.py \
    --recipe qwen3_30b_a3b_pretrain_config \
    --dataset llm-pretrain-mock \
    'model.cuda_graph_impl=transformer_engine' \
    'model.cuda_graph_scope=[attn,moe_router,moe_preprocess]' \
    'model.use_te_rng_tracker=True' \
    'rng.te_rng_tracker=True'
```

---

## Quick Reference: Which Recipe for My Situation?

| I want to... | Start with | GPUs needed |
|---|---|---|
| Try Bridge for the first time | `llama3_8b_sft_config` + mock data | 2 |
| Fine-tune a 7-8B model | `llama3_8b_sft_config` or `qwen3_8b_sft_config` | 2–8 |
| LoRA on 1 GPU | `llama3_8b_peft_config` or `qwen3_8b_peft_config` | 1 |
| Pretrain a dense 70B | `llama3_70b_pretrain_config` | 32–64 |
| Train a small MoE | `qwen3_30b_a3b_pretrain_config` | 8 |
| Train a large MoE (235B+) | `qwen3_235b_a22b_pretrain_config` | 256–512 |
| Benchmark throughput | Perf recipes via `run_script.py` | Varies |
| Long-context training | `llama3_8b_128k_pretrain_config` or add CP override | 16+ |
| VLM fine-tuning | `qwen3_vl_8b_sft_config` or `gemma3_vl_*_sft_config` | 4–8 |
| Diffusion training | `wan_1_3B_pretrain_config` or `flux_12b_pretrain_config` | 8 |

---

## Code Anchors

| What | Path |
|------|------|
| Library recipes root | `src/megatron/bridge/recipes/` |
| Recipe `__init__.py` (all exports) | `src/megatron/bridge/recipes/__init__.py` |
| Common recipe helpers | `src/megatron/bridge/recipes/common.py` |
| Training entry point | `scripts/training/run_recipe.py` |
| Perf recipes root | `scripts/performance/` |
| Perf entry point | `scripts/performance/run_script.py` |
| Perf workload configs | `scripts/performance/configs/<family>/` |
| Perf overrides (benchmark defaults) | `scripts/performance/utils/overrides.py` |

Files

This skill is a package of 5 files. Install with the command above, or download the folder.

📄SKILL.md entry
📁evals
◆evals.json 2.0 KB
📄BENCHMARK.md 4.2 KB
📄skill-card.md 3.8 KB
📄skill.oms.sig GitHub 0 B

NVIDIA/skills/tree/main/skills/nemo-mbridge-recipe-recommender commit · 63c02e122e

▌
slime-rl-training · orchestra-research bundle
Post-train LLMs with reinforcement learning using the slime framework, which integrates Megatron-LM for training and SGLang for rollout generation.

10.4k
repo stars
▌
miles-rl-training · orchestra-research bundle
Train large-scale MoE models with FP8/INT4 low-precision RL, speculative decoding, and train-inference alignment using the miles framework.

10.4k
repo stars
▌
slime-rl-training · lord1egypt
Guides LLM post-training with RL using slime, a Megatron+SGLang framework for training GLM, Qwen, DeepSeek, and Llama models with GRPO, async, and multi-turn workflows.

2
repo stars
▌
prompt-refine · gabrielmoreira bundle
Silently restructures natural-language prompts into the format best suited for the model currently executing the skill, then answers the rewritten version.

17
repo stars
▌
gemma-dev · google-gemma bundle
Selects the right Gemma model for a task, recommends deployment tooling (Gradio, Transformers.js, Vertex AI, MLX), and applies optimizations like MTP and QAT.

6
installs
▌
implementing-llms-litgpt · orchestra-research bundle
Train, fine-tune, and deploy LLMs using LitGPT's clean implementations of 20+ architectures like Llama, Gemma, and Phi.

10.4k
repo stars

Frequently asked questions

How do I install the nemo-mbridge-recipe-recommender skill?

Run npx skillmds add nvidia/nemo-mbridge-recipe-recommender in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

What does the nemo-mbridge-recipe-recommender skill do?

Indexes Megatron Bridge recipes and recommends the best starting config based on model, GPU count, and training goal. It is listed under AI & ML, Model Training & Fine-tuning on SkillMD.

Is nemo-mbridge-recipe-recommender safe to use?

SkillMD's automated safety review verdict for this skill is PASS. Independent scanners report: SkillSpector: PASS, Skill Scanner: PASS. Capability flags: makes network calls. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with nemo-mbridge-recipe-recommender?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is nemo-mbridge-recipe-recommender free to use?

Yes. Installing skills from SkillMD is free. This skill is licensed under Apache-2.

Who published nemo-mbridge-recipe-recommender?

NVIDIA (@nvidia) published this skill as a verified publisher. Their other Agent Skills are listed on their SkillMD profile.

nemo-mbridge-recipe-recommender

Auto Recipe — Recipe Index & Recommendation

How to Use This Skill

First Answer Checklist

Entry Points

Library recipes (functional training)

Performance recipes (throughput benchmarks)

Recipe Unification (Coming Soon — PR #2803)

Library Recipe Index

Llama

Qwen2 / Qwen2.5

Qwen3 (Dense)

Qwen3 MoE

Qwen3-Next

DeepSeek

GLM-4.5

Gemma

NemotronH / Nemotron

Other Models

VLM Recipes

Diffusion Recipes

Performance Recipe Index

Llama 3 / 3.1

DeepSeek V3

Qwen3 MoE

Qwen3-VL

Kimi K2

NemotronH

GPT-OSS

Recommendation Decision Tree

Adjustment Advice (When Recommending)

Parallelism Resizing Rules

Batch Size Tuning

Common Pitfalls to Warn About

Recipe Override Examples

Quick Reference: Which Recipe for My Situation?

Code Anchors

Files

Related

Frequently asked questions