Container Compute Service:training-nv-pytorch 26.03

Main features

• Upgraded torch to version 2.10.

• Upgraded vllm to version 0.17.0.

• Upgraded megatron-core to version 0.16.0.

• Upgraded deepspeed to version 0.18.8.

• Upgraded transformer_engine to version 2.12.

Bug fixes

No bug fixes in this release.

Assets

Driver requirements

• The 26.03 release supports CUDA 12.8.0 and CUDA 13.0.2, depending on the driver version. CUDA 13.0.2 requires NVIDIA Driver version 580 or later, and CUDA 12.8.0 requires NVIDIA Driver version 575 or later. For a complete list of supported drivers, see the CUDA Application Compatibility topic. For more information, see CUDA Compatibility and Upgrades.

PyTorch compiling optimization

torch.compile(), introduced in PyTorch 2.0, is effective for single-GPU training but provides limited or negative benefit for large language model (LLM) training, which depends on GPU memory optimization and distributed frameworks such as Fully Sharded Data Parallel (FSDP) or DeepSpeed.

This release improves torch.compile() for distributed LLM training through two optimizations:

• Communication granularity control in DeepSpeed: Controlling communication granularity gives the compiler a complete compute graph, enabling wider compiling optimization.

• Frontend improvements: The PyTorch compiler frontend now compiles even when a graph break occurs, with enhanced mode matching and dynamic shape capabilities.

Result: ~20% higher end-to-end throughput in 8B-parameter LLM training.

GPU memory optimization for recomputation

Based on performance tests across different clusters and parameter configurations, this release integrates the optimal number of activation recomputation layers directly into PyTorch. Enable it with a single environment variable—no manual tuning required.

This feature is available in the DeepSpeed framework.

End-to-end performance evaluation

Using the cloud-native AI performance analysis tool CNP, we conducted a comprehensive end-to-end performance comparison against a standard base image with mainstream open-source models and framework configurations. We also performed ablation studies to evaluate the contribution of each optimized component to the overall model training performance.

Image comparison: Base image vs. iterative evaluation

End-to-end performance contribution analysis of core GPU components

The following tests evaluate and compare end-to-end training performance on a multi-node GPU cluster using this image release. The comparison items include:

1. Base: NGC PyTorch image

2. ACS AI Image (Base + ACCL): NGC PyTorch image using the Alibaba Cloud Communication Library (ACCL).

3. ACS AI Image (AC2 + ACCL): Golden image using AC2 Base OS with no optimizations enabled.

4. ACS AI Image (AC2 + ACCL + CompilerOpt): Golden image using AC2 Base OS with only the torch compile optimization enabled.

5. ACS AI Image (AC2 + ACCL + CompilerOpt + CkptOpt): Golden image using AC2 Base OS with both torch compile and selective gradient checkpointing optimizations enabled.

Step 1: Select an image

docker pull registry.cn-wulanchabu.cr.aliyuncs.com/egslingjun/training-nv-pytorch:[tag]

Step 2: Enable compiler and memory optimizations

• Enable compilation optimization

  Use the transformers Trainer API:

  

• Enable selective gradient checkpointing

  export CHECKPOINT_OPTIMIZATION=true

Step 3: Start the container

The image includes the built-in model training tool ljperf. The following steps describe how to start the container and run a training task.

# Start and enter the container
docker run --rm -it --ipc=host --net=host --privileged registry.cn-wulanchabu.cr.aliyuncs.com/egslingjun/training-nv-pytorch:[tag]

# Run the training demo
ljperf benchmark --model deepspeed/llama3-8b 

Step 4: Usage recommendations

• The image contains modifications to libraries such as PyTorch and DeepSpeed. Do not reinstall them.

• In the DeepSpeed configuration, leave the zero_optimization.stage3_prefetch_bucket_size parameter empty or set it to auto.