6th IDRIS hackathon May 12, 19, 20 and 21

You don't have to be an IDRIS user to take part in this event, so please feel free to spread this message to your colleagues. On this occasion, you'll have access to IDRIS computing resources, so you can test your code on recent GPUs up to NVIDIA H100 on Jean Zay and NVIDIA GB200 NVL72 on the DALIA prototype machine.

For HPC projects, this can be as much about starting to port code to GPU using OpenACC, OpenMP, Kokkos, GPU libraries, CUDA, etc. as it can be about optimizing an existing GPU implementation. In the case of large code, it is recommended to prepare a representative mini-application to be more efficient during the event.

For AI projects, this hackathon offers a unique opportunity to explore finetuning, developing or evaluating state-of-the-art models such as LLMs and GenAI applications. You'll be able to optimize your models using advanced techniques such as mixed precision, multi-GPU or multi-node parallelization (DDP, PP, TP, CP), using basic frameworks such as PyTorch or TensorFlow, or specialized tools such as DeepSpeed, Megatron-LM, Nanotron or NVIDIA NeMo. Although complete pre-training is out of reach within the duration of the hackathon, preparing pre-training code for large-scale models remains possible.

Registration is via the event page (in English): https://www.openhackathons.org/s/siteevent/a0CUP00003Ahxc22AB/se000449.

--------------------------------------------------------------------

IDRIS is organizing an HPC and AI GPU Hackathon together with NVIDIA.

The Hackathon will be held on May 12th 2026 (remotely) and from May 19th through May 21st (on site at IDRIS). The application deadline is March 10th.

GPU Hackathons provide opportunities for teams of three or four developers and researchers to accelerate their HPC or AI codes under the guidance of expert mentors from universities, national labs, computing centers, governmental institutions and hardware manufacturers in a collaborative environment.

The event is open to everyone even if you are not already a user of IDRIS computing resources so feel free to forward this message to your colleagues. During the Hackathon, you will be provided with access to the computing resources of IDRIS, in order to be able to benchmark your code on recent GPUs, including NVIDIA H100s on Jean Zay and NVIDIA GB200s NLV72 on the DALIA prototype machine.

For HPC projects, it could be the opportunity to start porting a code to GPU using OpenACC, OpenMP, Kokkos, GPU libraries, CUDA, etc or to optimize an existing GPU implementation. If your codebase is large in size, we recommend selecting and extracting a representative mini-app from the original application in order to be more efficient during the Hackathon.

For AI projects, it will be the occasion to explore finetuning, developing or evaluating cutting-edge models such as LLMs or GenAI applications. You will be able to improve your models thanks to advanced techniques like mixed precision, multi-GPU or multi-node parallelization (DDP, PP, TP, CP), by using standard frameworks such as PyTorch or TensorFlow, or specialized tools like DeepSpeed, Megatron-LM, Nanotron or NVIDIA NeMo. Even though a full pretraining cannot be achieved in the time frame of the Hackathon, you can get a pretraining code ready for large scale models.

Please register on https://www.openhackathons.org/s/siteevent/a0CUP00003Ahxc22AB/se000449.