training-nv-pytorch 25.05

Features

• The base image CUDA is upgraded to 12.9.0.

Bug fixes

• PyTorch is upgraded to 2.6.0.7.post1, which fixes the profile crash issue in the open-source community.

25.05

egslingjun-registry.cn-wulanchabu.cr.aliyuncs.com/egslingjun/training-nv-pytorch:25.05-serverless

VPC image

acs-registry-vpc.{region-id}.cr.aliyuncs.com/egslingjun/{image:tag}

Replace the placeholders with actual values:

PyTorch compilation optimization

torch.compile(), introduced in PyTorch 2.0, often delivers strong gains for small-scale, single-GPU workloads. But LLM training depends on GPU memory optimization and distributed frameworks such as FSDP or DeepSpeed, so torch.compile() may offer limited benefits or even degrade performance.

• Control communication granularity in the DeepSpeed framework. This helps the compiler capture a more complete compute graph and apply broader compilation optimizations.

• Use an optimized PyTorch build:

  • The PyTorch compiler frontend is improved to ensure that compilation succeeds even if a graph break occurs in the compute graph.

  • Pattern matching and dynamic shape support are strengthened to improve post-compilation performance.

With these optimizations, 8B-parameter LLM training typically achieves an end-to-end throughput gain of about 20%.

GPU memory optimization for recomputation

A predictive model for GPU memory overhead, built on large-scale performance data—including different models, clusters, and training parameter settings, as well as system metrics such as GPU memory utilization collected during benchmarking—recommends the optimal number of activation recomputation layers. This approach is integrated into PyTorch, allowing you to achieve the performance gains of GPU memory optimization with minimal effort. This feature is now supported in the DeepSpeed framework.

ACCL

ACCL is Alibaba Cloud’s high-performance communication library built for Lingjun. ACCL-N is the GPU-focused version. ACCL-N is a high-performance communication library customized from NVIDIA NCCL. It is fully compatible with NCCL, fixes issues in the upstream NCCL release, and includes performance and stability improvements.

1. Pull the image

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

2. Enable optimizations

Enable compiling optimization

Use the transformers Trainer API:

Enable GPU memory optimization for activation recomputation

export CHECKPOINT_OPTIMIZATION=true

3. Launch a container

The image includes ljperf, a built-in model training tool for launching containers and running training tasks.

LLM example:

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

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

Usage notes

• This release modifies the PyTorch and DeepSpeed libraries. Do not reinstall them.

• Leave zero_optimization.stage3_prefetch_bucket_size blank or set it to auto in your DeepSpeed configuration.