From 32a3900cad17e351c34602d99b8eea66b0557540 Mon Sep 17 00:00:00 2001 From: Ananth Subramaniam Date: Thu, 3 Sep 2026 17:17:53 -0700 Subject: [PATCH] docs: update NeMo RL tutorial container Signed-off-by: Ananth Subramaniam --- .../nemo-rl-grpo/multi-node-training.mdx | 6 +++--- .../pages/training-tutorials/nemo-rl-grpo/setup.mdx | 8 ++++---- 2 files changed, 7 insertions(+), 7 deletions(-) diff --git a/fern/versions/latest/pages/training-tutorials/nemo-rl-grpo/multi-node-training.mdx b/fern/versions/latest/pages/training-tutorials/nemo-rl-grpo/multi-node-training.mdx index e1b057da65..0a24bb3ce5 100644 --- a/fern/versions/latest/pages/training-tutorials/nemo-rl-grpo/multi-node-training.mdx +++ b/fern/versions/latest/pages/training-tutorials/nemo-rl-grpo/multi-node-training.mdx @@ -57,14 +57,14 @@ cd /path/to/nemo/rl # WANDB_API_KEY: Your Weights & Biases API key for logging # EXP_NAME: Experiment name # NUM_ACTOR_NODES: Number of GPU nodes to use (2, 4, 8, etc.) -# CONTAINER_IMAGE_PATH: Nano recipe container (see RL Framework Compatibility table) +# CONTAINER_IMAGE_PATH: NeMo RL release container # SLURM_ACCOUNT: Slurm account # SLURM_PARTITION: Slurm partition WANDB_API_KEY={your W&B API key} \ EXP_NAME=nemo_gym_grpo/nemotron_nano_v2_9b/2nodes/workplace_assistant_001 \ NUM_ACTOR_NODES=2 \ REPO_LOCATION=$PWD \ -CONTAINER_IMAGE_PATH=nvcr.io/nvidia/nemo-rl:v0.4.0.nemotron_3_nano \ +CONTAINER_IMAGE_PATH=nvcr.io/nvidia/nemo-rl:v0.7.0 \ SLURM_ACCOUNT={your Slurm account} \ SLURM_PARTITION={your Slurm partition} \ examples/nemo_gym/launch_nemo_gym_multinode_training.sh \ @@ -77,7 +77,7 @@ SLURM_PARTITION={your Slurm partition} \ If you are using enroot following the steps in the [Setup](/tutorials/training-tutorials/nemo-rl-grpo/setup) doc and downloaded the container locally, use the local container filepath instead: ```bash -CONTAINER_IMAGE_PATH=$PWD/../nvcr.io/nvidia/nemo-rl:v0.4.0.nemotron_3_nano \ +CONTAINER_IMAGE_PATH=$PWD/../nvcr.io/nvidia/nemo-rl:v0.7.0 \ ``` diff --git a/fern/versions/latest/pages/training-tutorials/nemo-rl-grpo/setup.mdx b/fern/versions/latest/pages/training-tutorials/nemo-rl-grpo/setup.mdx index ba9a561bb5..2a11cb6e39 100644 --- a/fern/versions/latest/pages/training-tutorials/nemo-rl-grpo/setup.mdx +++ b/fern/versions/latest/pages/training-tutorials/nemo-rl-grpo/setup.mdx @@ -112,13 +112,13 @@ CONTAINER_IMAGE_PATH=./$CONTAINER_IMAGE_PATH -This tutorial trains [Nemotron Nano 9B v2](https://huggingface.co/nvidia/NVIDIA-Nemotron-Nano-9B-v2), not Nemotron 3 Ultra. Use the Nano NGC container from the [RL Framework Compatibility](/reference/rl-framework-compatibility) table (v0.1.1 / Nemotron 3 Nano row). For Nemotron 3 Ultra training with NeMo Gym v0.3.0, build the `ultra-v3` Dockerfile instead. +This tutorial trains [Nemotron Nano 9B v2](https://huggingface.co/nvidia/NVIDIA-Nemotron-Nano-9B-v2), not Nemotron 3 Ultra. The container and NeMo RL checkout below use the same release so their Python and uv versions remain compatible. For Nemotron 3 Ultra training with NeMo Gym v0.3.0, build the `ultra-v3` Dockerfile instead. ```bash -# Nano recipe container — see RL Framework Compatibility table +# NeMo RL release container # https://catalog.ngc.nvidia.com/orgs/nvidia/containers/nemo-rl -CONTAINER_IMAGE_PATH=nvcr.io/nvidia/nemo-rl:v0.4.0.nemotron_3_nano +CONTAINER_IMAGE_PATH=nvcr.io/nvidia/nemo-rl:v0.7.0 NUM_ACTOR_NODES=1 ACCOUNT= @@ -153,7 +153,7 @@ For the first setup on your local filesystem: ```bash # Clone NeMo RL repository -git clone https://github.com/NVIDIA-NeMo/RL +git clone --branch v0.7.0 https://github.com/NVIDIA-NeMo/RL cd RL # Initialize all submodules (Gym, Megatron, AutoModel, etc.)