Skip to content

Nemotron Ultra config for ModelOpt examples #5159

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
jenchen13 merged 3 commits into from
Jun 4, 2026
Merged

Nemotron Ultra config for ModelOpt examples #5159

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
453761c
add nemotron ultra config for modelopt examples
jenchen13 Jun 4, 2026
1067c73
update HF model
jenchen13 Jun 4, 2026
1734ebc
update readme
jenchen13 Jun 4, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions examples/post_training/modelopt/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,6 +35,7 @@ knowledge distillation, pruning, speculative decoding, and more.
| `nvidia/NVIDIA-Nemotron-Nano-9B-v2` | ✅ | - | ✅ | ✅ |
| `nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16` | ✅ | - | ✅ | ✅ |
| `nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16` | ✅ | - | ✅ | ✅ |
| `nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16` | ✅ | - | ✅ | ✅ |
| `openai/gpt-oss-{20b, 120b}` | ✅ | **Online** | ✅ | ✅ |
| `Qwen/Qwen3-{0.6B, 8B}` | ✅ | ✅ | ✅ | ✅ |
| `Qwen/Qwen3-{30B-A3B, 235B-A22B}` | **WAR** | ✅ | ✅ | ✅ |
Expand Down
59 changes: 59 additions & 0 deletions examples/post_training/modelopt/conf/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,59 @@
#!/bin/bash

if [ -z "${HF_MODEL_CKPT:-}" ]; then
HF_MODEL_CKPT=nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16
fi

if [ -z "${TOKENIZER_MODEL:-}" ]; then
TOKENIZER_MODEL=${HF_MODEL_CKPT}
fi

MODEL_ARGS=" \
--trust-remote-code \
--save-interval 100 \
--micro-batch-size 1 \
--enable-experimental \
--use-fused-weighted-squared-relu \
--cross-entropy-loss-fusion \
--cross-entropy-fusion-impl native \
--moe-permute-fusion \
--moe-latent-size 2048 \
--moe-router-score-function sigmoid \
--moe-grouped-gemm \
--num-experts 512 \
--moe-router-topk 22 \
--moe-shared-expert-intermediate-size 10240 \
--moe-aux-loss-coeff 1e-4 \
--moe-router-topk-scaling-factor 5.0 \
--moe-router-enable-expert-bias \
--moe-router-dtype fp32 \
--moe-router-load-balancing-type seq_aux_loss \
--moe-token-dispatcher-type alltoall \
\
--attention-backend flash \
--disable-gloo-process-groups \
--is-hybrid-model \
--mamba-num-heads 256 \
--mamba-head-dim 64 \
--hybrid-override-pattern MEMEMEM*EMEMEM*EMEMEMEM*EMEMEMEM*EMEMEM*EMEMEMEM*EMEMEMEM*EMEMEM*EMEMEMEM*EMEMEMEM*EMEMEM*EMEMEMEM*EMEMEMEME \
--use-mcore-models \
--untie-embeddings-and-output-weights \
--disable-bias-linear \
--init-method-std 0.0099 \
--position-embedding-type none \
--squared-relu \
--num-layers 108 \
--hidden-size 8192 \
--num-attention-heads 64 \
--group-query-attention \
--num-query-groups 2 \
--ffn-hidden-size 5120 \
--kv-channels 128 \
--normalization RMSNorm \
\
--tokenizer-type HuggingFaceTokenizer \
--bf16 \
--seq-length 8192 \
--max-position-embeddings 8192 \
--export-model-type HybridModel \
"
Loading