# Training a model with Primus and JAX MaxText Training performance validation with the ROCm JAX MaxText training Docker image on AMD Instinct accelerators. ## Overview MaxText for ROCm is a specialized fork of upstream MaxText, designed to enable training of large language models (LLMs) on AMD GPUs. By leveraging AMD Instinct™ MI300X and MI355X GPUs, MaxText delivers scalability, performance, and resource utilization for AI workloads. See the GitHub repository at [ROCm/maxtext](https://github.com/ROCm/maxtext/). AMD provides a ready-to-use Docker image for AMD Instinct MI300X and MI355X GPUs containing essential components, including JAX, XLA, ROCm libraries, and MaxText utilities. For the full software stack of this image (ROCm, JAX, Transformer Engine, hipBLASLt, RCCL, TensorFlow, and the rest), see [Release notes → `rocm/jax-training:maxtext-v26.7`](../01-getting-started/release-notes.md#rocmjax-trainingmaxtext-v267). The release notes are the single source of truth for image contents. > **Primus source:** use the `release/v26.7` branch rather than the Primus copy baked into the image — see [Release notes → Primus source for v26.7](../01-getting-started/release-notes.md#primus-source-for-v267) for why. --- ## Important notes for v26.7 Read this section before starting a training run. It collects the settings this release requires, the architecture-specific workarounds, and the known issues. The contents change from release to release, so re-read it when you move to a new image tag. ### Required settings **Enable Shardy.** Shardy is the partitioning system in JAX. The v26.7 image ships JAX 0.11.0, which requires it, so set `shardy=True` during the training run. You may see partitioning-related errors if it is not configured correctly. See the [Shardy migration guide](https://docs.jax.dev/en/latest/shardy_jax_migration.html) for details. ### Architecture-specific settings **MI355X (gfx950) — disable RCCL WarpSpeed.** RCCL's WarpSpeed feature (`RCCL_WARP_SPEED_AUTO`) is a gfx950-only optimization that is enabled by default in gfx950 builds, and it can cause **NaN losses** during training. Primus automatically sets `RCCL_WARP_SPEED_AUTO=0` when a gfx950 (MI355X) device is detected, so `primus-cli` handles it for you. If you launch training manually on MI355X outside of Primus, export it yourself: ```bash export RCCL_WARP_SPEED_AUTO=0 ``` This variable is a no-op on MI300X (gfx942). ### Known issues **Loss curve discrepancy with `packing=false`.** With `packing=false` the loss converges at a slightly higher value than in previous images. To reproduce the earlier convergence, set `NVTE_CK_USES_FWD_V3=0`, which uses Flash Attention v2 for the forward pass instead of v3. This is being tracked and will be addressed in a future release. **The ROCm 10.0.0 move is untested against these recipes.** v26.7 rebuilds the whole JAX stack on ROCm 10.0.0 and renames the plugin wheels; the per-model settings below were validated on the v26.6 (ROCm 7.14.0) stack. Re-check throughput and convergence on your own hardware before trusting a number here. --- ## Supported features and models MaxText supports the following key features to train large language models efficiently: - Transformer Engine (TE) - Flash Attention (FA) 3, with or without input sequence packing - GEMM tuning - Multi-node support - NANOO FP8 (for MI300X) or FP8 (for MI355X) The following models are pre-optimized for performance on the AMD Instinct MI300X, MI325X, and MI355X accelerators. Each cell lists the precisions that ship as experiment configs for that device; the config path is `examples/maxtext/configs//--pretrain.yaml`. | Model | Config prefix | MI300X / MI325X | MI355X | | ----- | ------------- | --------------- | ------ | | Llama 2 7B | `llama2_7B` | `bf16`, `nanoo_fp8` | `bf16`, `fp8` | | Llama 2 70B | `llama2_70B` | `bf16`, `nanoo_fp8` | `bf16`, `fp8` | | Llama 3/3.1 8B | `llama3_8B` | `bf16`, `nanoo_fp8` | `bf16`, `fp8` | | Llama 3/3.1 70B | `llama3_70B` | `bf16` | `bf16`, `fp8` | | Llama 3.1 405B | `llama3.1_405B` | — | `bf16`, `fp8` | | Llama 3.3 70B | `llama3.3_70B` | `bf16` | `bf16`, `fp8` | | DeepSeek-V2-lite (16B) | `deepseek_v2_16B` | `bf16`, `nanoo_fp8` | `bf16`, `fp8` | | Gemma4 26B | `gemma4_26B` | `bf16`, `nanoo_fp8` | `bf16`, `fp8` | | Gemma4 31B | `gemma4_31B` | `bf16`, `nanoo_fp8` | `bf16`, `fp8` | | Grok-1 | `grok1` | `nanoo_fp8` | — | | Mixtral 8x7B | `mixtral_8x7B` | `bf16`, `nanoo_fp8` | `bf16`, `fp8` | | Mixtral 8x22B | `mixtral_8x22B` | `bf16` | `bf16` | | Qwen3 14B | `qwen3_14B` | `bf16`, `nanoo_fp8` | `bf16`, `fp8` | | Qwen3 30B-A3B | `qwen3_30B_A3B` | `bf16`, `nanoo_fp8` | `bf16`, `fp8` | Browse `examples/maxtext/configs/` for the authoritative list — it tracks whatever configs ship in your checkout. > **Note:** Some models, such as Llama 3, require an external license agreement through a third party (for example, Meta). The only models supported in this workflow are those listed above. --- ## System validation If you have already validated your system, skip this step. Otherwise, complete the [system validation and optimization steps](https://rocm.docs.amd.com/en/latest/how-to/rocm-for-ai/training/prerequisite-system-validation.html#train-a-model-system-validation) to set up your system before starting training. --- ## Environment setup This Docker image is optimized for specific model configurations outlined below. Performance can vary for other training workloads, as AMD doesn't validate configurations and run conditions outside those described. This container should not be expected to provide generalized performance across all training workloads. Expect the container to perform in the model configurations described below; other configurations and run conditions are not validated by AMD. ### Multi-node prerequisites (RDMA only) For multi-node runs, make sure all packages are installed based on the network device you use. You only need the setup below if you are using multi-node with RDMA—otherwise skip this part. Install the packages below for building and installing the RDMA driver: ```bash apt install iproute2 -y apt install -y linux-headers-"$(uname -r)" libelf-dev apt install -y gcc make libtool autoconf librdmacm-dev rdmacm-utils infiniband-diags ibverbs-utils perftest ethtool libibverbs-dev rdma-core strace libibmad5 libibnetdisc5 ibverbs-providers libibumad-dev libibumad3 libibverbs1 libnl-3-dev libnl-route-3-dev ``` Refer to your NIC manufacturer's webpage for further steps about compiling and installing the RoCE driver. For Broadcom, see the section **Compiling Broadcom NIC Software from Source** in the [Ethernet Networking Guide for AMD Instinct MI300X GPU Clusters](https://docs.broadcom.com/doc/957608-AN2XX). ### Multi-node environment variables Set the following environment variables. **Master address**—change `localhost` to the master node's hostname: ```bash export MASTER_ADDR="${MASTER_ADDR:-localhost}" ``` **Number of nodes**—set the number of nodes you want to train on (for example, 2, 4, 8): ```bash export NNODES="${NNODES:-1}" ``` **Node rank**—set the rank of each node (0 for master, 1 for the first worker node, and so on): ```bash export NODE_RANK="${NODE_RANK:-0}" ``` **Network interface**—update the network interface in the script to match your system's network interface. To find your network interface, run this outside the container: ```bash ip a ``` Then update the following variable in the script: ```bash export NCCL_SOCKET_IFNAME=ens50f0np0 ``` **RDMA interface**—first make sure the packages above are installed on all the nodes. Then set the RDMA interfaces to use for communication: ```bash # If using Broadcom NIC export NCCL_IB_HCA=rdma0,rdma1,rdma2,rdma3,rdma4,rdma5,rdma6,rdma7 # If using Mellanox NIC export NCCL_IB_HCA=mlx5_0,mlx5_1,mlx5_2,mlx5_3,mlx5_4,mlx5_5,mlx5_8,mlx5_9 ``` For Primus-specific networking guidance, see [Multi-node networking](../04-technical-guides/multi-node-networking.md). --- ## Training and benchmarking options Use the following instructions to set up the environment, configure the script to train models, and reproduce the benchmark results on the MI300X, MI325X, MI350X, and MI355X accelerators with the Docker image. All training runs through `primus-cli`. There are two ways to reach it: | Method | Use it when | | ------ | ----------- | | [**`primus-cli`**](#running-training-with-primus-cli-recommended) **(recommended)** | Any new work. One CLI covers direct, container, and Slurm launches, and the same YAML configurations work across every Primus backend. | | [Standalone benchmarking](#standalone-benchmarking) | You want to launch the container yourself and reproduce a single benchmark configuration by hand. | JAX MaxText is integrated into [Primus](https://github.com/AMD-AGI/Primus), which supports multiple backends including Megatron-LM, TorchTitan, and JAX MaxText alongside ROCm-optimized components. The unified `primus-cli` runs training jobs with the JAX MaxText backend and is the path we recommend. --- ## Running training with primus-cli (recommended) **Clone the Primus branch matching the image.** Do this on the host — every command below runs from this directory: ```bash git clone --recurse-submodules https://github.com/AMD-AGI/Primus.git cd Primus git checkout release/v26.7 git submodule update --init third_party/maxtext/ ``` That is all the setup required. `primus-cli container` starts the image for you, mounts this checkout into it at the same path, and runs the training inside — so this is the code that executes, and the `/workspace/Primus` copy baked into the image is not used. It also forwards environment variables you export on the host; the forwarded list is `container.options.env` in `runner/.primus.yaml`. > **Pass `--image` for MaxText.** The default image in `runner/.primus.yaml` is `rocm/primus`, which is the PyTorch image. MaxText runs need `--image rocm/jax-training:maxtext-v26.7` in container mode, or the image set in your Slurm config file. For detailed usage of `primus-cli`, refer to the [CLI reference](./cli-reference.md). The examples below target MI355X. Primus automatically sets `RCCL_WARP_SPEED_AUTO=0` when it detects a gfx950 (MI355X) device — see [Architecture-specific settings](#architecture-specific-settings). The variable is a no-op on MI300X, so it is safe to leave in place on either architecture if you are exporting it manually. **Container mode (recommended)**—run from the host; `primus-cli` starts the container: ```bash ./runner/primus-cli container --image rocm/jax-training:maxtext-v26.7 \ -- train pretrain --config examples/maxtext/configs/MI355X/llama2_7B-bf16-pretrain.yaml ``` **Direct mode**—only when you already have a shell **inside** the MaxText container (or a bare-metal JAX install); no `--image` needed: ```bash ./runner/primus-cli direct \ -- train pretrain --config examples/maxtext/configs/MI355X/llama2_7B-bf16-pretrain.yaml ``` **Slurm mode**—distributed training on a Slurm cluster: ```bash # Use a custom config file, where you can specify the docker image and set environment variables. ./runner/primus-cli --config my_maxtext_config.yaml slurm srun -N 8 \ -- train pretrain --config examples/maxtext/configs/MI355X/llama2_7B-bf16-pretrain.yaml ``` To run a different model or GPU architecture, swap the `--config` path. Configurations live under `examples/maxtext/configs/MI300X/` and `examples/maxtext/configs/MI355X/`. Config filenames follow the pattern `--pretrain.yaml`, where `` is `bf16`, `fp8` (MI355X), or `nanoo_fp8` (MI300X). See [End-to-end training recipes](./end-to-end-training-recipes.md) for the full inventory and [MaxText parameters](../03-configuration-reference/maxtext-parameters.md) for the YAML fields. --- ## Standalone benchmarking Use the following command to pull the Docker image from Docker Hub: ```bash docker pull rocm/jax-training:maxtext-v26.7 ``` ### Single-node training #### Setup > **Note:** Adjust the following variables based on your environment. Export variables: - `HF_TOKEN` is your Hugging Face token to access models, tokenizers, and data. See [User access tokens](https://huggingface.co/docs/hub/en/security-tokens) for more information. - `HF_HOME` is where `huggingface_hub` will store local data. Refer to the [Hugging Face CLI documentation](https://huggingface.co/docs/huggingface_hub/main/en/guides/cli#hf-download) on how to download the data. If you already have downloaded or cached Hugging Face artifacts, set this variable to that path. Downloaded files typically get cached to a place like `~/.cache/huggingface`. ```bash export HF_TOKEN= export HF_HOME= ``` Launch the Docker container: ```bash docker run -it \ --device /dev/dri --device /dev/kfd \ --network host --ipc host --group-add video \ --cap-add SYS_PTRACE --security-opt seccomp=unconfined --privileged \ -v $HOME:$HOME -v $HOME/.ssh:/root/.ssh \ -v $HF_HOME:/hf_cache -e HF_HOME=/hf_cache \ -e HF_TOKEN=$HF_TOKEN \ --shm-size 64G --name training_env \ rocm/jax-training:maxtext-v26.7 ``` Execute the `training_env` container (optional if you are already in the container): ```bash docker start training_env docker exec -it training_env bash ``` Inside the container, the Primus repository (with the MaxText backend) is available at `/workspace/Primus`. Run training with `primus-cli` in direct mode; configs live under `examples/maxtext/configs//` where `` is `MI300X` or `MI355X`. ```bash cd /workspace/Primus # Unquantized (bf16), e.g. Llama 2 7B on MI300X # Note: RCCL_WARP_SPEED_AUTO=0 is auto-set by Primus on MI355X (gfx950). ./primus-cli direct -- train pretrain \ --config examples/maxtext/configs/MI300X/llama2_7B-bf16-pretrain.yaml ``` For quantized training, replace `-bf16-` in the config name with `-fp8-` (MI355X) or `-nanoo_fp8-` (MI300X): ```bash # nanoo_fp8 on MI300X ./primus-cli direct -- train pretrain \ --config examples/maxtext/configs/MI300X/llama2_7B-nanoo_fp8-pretrain.yaml # fp8 on MI355X ./primus-cli direct -- train pretrain \ --config examples/maxtext/configs/MI355X/llama2_7B-fp8-pretrain.yaml ``` ### Benchmarking examples Every listed model has a bf16 variant. Quantization is device-specific: MI300X uses NANOO FP8 (`-nanoo_fp8-`) and MI355X uses FP8 (`-fp8-`). The columns below show which quantized variant is also available: | Model | MI300X (bf16 + …) | MI355X (bf16 + …) | | ---------------- | ----------------- | ----------------- | | Llama 2 7B | `-nanoo_fp8-` | `-fp8-` | | Llama 2 70B | `-nanoo_fp8-` | `-fp8-` | | Llama 3/3.1 8B | `-nanoo_fp8-` | `-fp8-` | | Llama 3/3.1 70B | bf16 only | `-fp8-` | | Llama 3.3 70B | bf16 only | `-fp8-` | | DeepSeek-V2-lite | `-nanoo_fp8-` | `-fp8-` | | Gemma4 26B | `-nanoo_fp8-` | `-fp8-` | | Gemma4 31B | `-nanoo_fp8-` | `-fp8-` | | Mixtral-8x7B | `-nanoo_fp8-` | `-fp8-` | | Qwen3 14B | `-nanoo_fp8-` | `-fp8-` | | Qwen3 30B-A3B | `-nanoo_fp8-` | `-fp8-` | ### Multi-node training > **Note:** These scripts will launch the Docker container and execute the benchmark, so **run them outside of any Docker container**. Multi-node training is launched through the unified `primus-cli` in Slurm mode. The general form for a multi-node run is: ```bash # From /workspace/Primus (or a cloned Primus checkout) # RCCL_WARP_SPEED_AUTO=0 is auto-set by Primus on MI355X (gfx950). ./primus-cli --config my_maxtext_config.yaml slurm srun -N \ -- train pretrain --config examples/maxtext/configs//--pretrain.yaml ``` where `` is `MI300X` or `MI355X`, `` is one of the MaxText configs (for example, `llama2_7B`, `llama2_70B`, `llama3_8B`, `llama3_70B`, `gemma4_26B`, `gemma4_31B`, `mixtral_8x7B`, `qwen3_14B`, `qwen3_30B_A3B`), and `` is `bf16`, `fp8` (MI355X), or `nanoo_fp8` (MI300X). #### Example commands 1. **Multi-node training with the Llama 2 7B model on 2 nodes:** ```bash ./primus-cli --config my_maxtext_config.yaml slurm srun -N 2 \ -- train pretrain --config examples/maxtext/configs/MI300X/llama2_7B-bf16-pretrain.yaml ``` 2. **Multi-node training with the Llama 2 70B model on 4 nodes:** ```bash ./primus-cli --config my_maxtext_config.yaml slurm srun -N 4 \ -- train pretrain --config examples/maxtext/configs/MI300X/llama2_70B-bf16-pretrain.yaml ``` 3. **Multi-node training with the Llama 3 8B model on 2 nodes:** ```bash ./primus-cli --config my_maxtext_config.yaml slurm srun -N 2 \ -- train pretrain --config examples/maxtext/configs/MI300X/llama3_8B-bf16-pretrain.yaml ``` 4. **Multi-node training with the Llama 3 70B model on 8 nodes:** ```bash ./primus-cli --config my_maxtext_config.yaml slurm srun -N 8 \ -- train pretrain --config examples/maxtext/configs/MI300X/llama3_70B-bf16-pretrain.yaml ``` 5. **Multi-node training with the Llama 3.1 405B model on MI355X (gfx950) with 8 nodes:** ```bash ./primus-cli --config my_maxtext_config.yaml slurm srun -N 8 \ -- train pretrain --config examples/maxtext/configs/MI355X/llama3.1_405B-bf16-pretrain.yaml ``` --- ## Profiling with JAX XPlane Profiler MaxText has built-in XPlane profiling support via JAX's profiler. Traces capture GPU kernel timelines, RCCL collectives, HLO graphs, and more. The output can be viewed in TensorBoard's Trace Viewer or analyzed with TraceLens. ### Key MaxText profiler flags The following MaxText config keys control profiling: ```text profiler=xplane # Use xplane format (produces .xplane.pb files) skip_first_n_steps_for_profiler=2 # Skip compilation/warmup steps profiler_steps=5 # Number of steps to profile upload_all_profiler_results=True # Save all GPU profiles (not just GPU0) ``` **Choosing step counts:** - `steps` should be greater than `skip_first_n_steps_for_profiler` + `profiler_steps` (for example, `steps=12` with `skip=2` and `profile=5` gives 5 warmup + 5 profiled + 2 cooldown) - `skip_first_n_steps_for_profiler=2` skips step 0 (compilation) and step 1 (warmup) - `profiler_steps=5` is typically enough; more steps mean larger `.xplane.pb` files ### Enabling the profiler in an experiment config The Primus MaxText experiment configs (`examples/maxtext/configs//--pretrain.yaml` in `/workspace/Primus`) already include a `profiler` key under `overrides` (set to `""` by default). To profile a config without passing flags on every launch, edit its `overrides` block and set the profiler fields: ```yaml profiler: "xplane" skip_first_n_steps_for_profiler: 2 profiler_steps: 5 upload_all_profiler_results: True steps: 12 ``` Then run the config as usual: ```bash ./primus-cli direct -- train pretrain \ --config examples/maxtext/configs/MI300X/llama3_8B-bf16-pretrain.yaml ``` Profile output will be written under the `base_output_directory` specified in the YAML (see [Output structure](#output-structure) below). ### Example: profile a model standalone in Docker ```bash #!/bin/bash set -e IMAGE="$1" # Docker image, e.g. rocm/jax-training:maxtext-v26.7 TAG="$2" # Short tag for output folder, e.g. v26.6_llama2_7b PROFILE_DIR="/path/to/profiles/${TAG}" mkdir -p "${PROFILE_DIR}" docker run --rm --privileged --network=host \ --device=/dev/dri --device=/dev/kfd --ipc=host \ -v "${PROFILE_DIR}:/mnt/profile" \ "${IMAGE}" bash -c ' export XLA_PYTHON_CLIENT_MEM_FRACTION=.97 export LD_LIBRARY_PATH=/usr/local/lib/:/opt/rocm/lib:$LD_LIBRARY_PATH export XLA_FLAGS="--xla_gpu_enable_latency_hiding_scheduler=True --xla_gpu_enable_command_buffer= " export GPU_MAX_HW_QUEUES=2 # On MI355X (gfx950), disable RCCL WarpSpeed to avoid NaN losses (no-op on MI300X) export RCCL_WARP_SPEED_AUTO=0 cd /workspace/maxtext python3 -m MaxText.train src/MaxText/configs/base.yml \ run_name=profile \ base_output_directory=/mnt/profile \ hardware=gpu \ steps=12 \ model_name= \ dataset_type=synthetic \ enable_checkpointing=False \ enable_goodput_recording=False \ monitor_goodput=False \ \ profiler=xplane \ skip_first_n_steps_for_profiler=2 \ profiler_steps=5 \ upload_all_profiler_results=True ' 2>&1 | tee "${PROFILE_DIR}/run.log" echo "Profile files:" find "${PROFILE_DIR}" -name "*.xplane.pb" -o -name "*.trace.json.gz" 2>/dev/null ``` ### Output structure MaxText writes profiles in TensorBoard format: ```text / └── profile/ └── tensorboard/ └── plugins/ └── profile/ └── / ├── .xplane.pb # Raw XPlane proto (GPU timelines) ├── .trace.json.gz # Trace viewer data └── *.hlo_proto.pb # HLO graphs for each compiled module ``` ### Viewing traces in TensorBoard ```bash pip install tensorboard tensorboard-plugin-profile # Point --logdir at the directory containing the tensorboard/ folder tensorboard --logdir /path/to/profiles//profile --port 6006 ``` Navigate to **Profile > Trace Viewer** in the TensorBoard UI. **Tips:** - Zoom into a single training step (skip the first profiled step as it may have residual warmup) - Look at individual GPU streams to see compute/RCCL overlap ### Keeping profile files small - Use `profiler_steps=5` (not more) to keep `.xplane.pb` under approximately 100 MB - Too many steps can produce files over 500 MB that TensorBoard struggles to load - `enable_checkpointing=False` avoids checkpoint I/O noise in the trace - `dataset_type=synthetic` eliminates data loading variability --- ## Profiling with rocprofv3 If you need to collect a trace and the JAX profiler isn't working, you can use `rocprofv3` as a temporary workaround: ```bash rocprofv3 --hip-trace --kernel-trace --memory-copy-trace --rccl-trace --output-format pftrace -d ./v3_traces -- python3 app.py ``` - Replace `python3 app.py` with any command line command that you want to run, such as `./primus-cli direct -- train pretrain --config examples/maxtext/configs/MI300X/llama2_7B-bf16-pretrain.yaml` (run from `/workspace/Primus`). - You can set the directory where you want the `.json` traces to be saved using `-d `. - The resulting traces can be opened in [Perfetto](https://ui.perfetto.dev/). --- ## Related documentation - [End-to-end training recipes](./end-to-end-training-recipes.md) - [Pretraining workflows](./pretraining.md) - [MaxText parameters](../03-configuration-reference/maxtext-parameters.md) - [CLI reference](./cli-reference.md) - [Multi-node networking](../04-technical-guides/multi-node-networking.md)