The ACSS Institute provides computing resources for researchers and advanced students working on social science projects.

The 4H100 and 2H100 servers, equipped with NVIDIA H100 GPUs, support work such as running open language models, processing text corpora and adapting models to a research question.

Fillable PDF · English · 3 pages

Open resource

Good practices on the GPU servers

A collection of practical sheets written by the ACSS team from its experience of the cluster: sharing GPUs, serving a language model with vLLM, estimating memory needs, adapting a model, spreading a computation over several GPUs and organising projects. Freely available.

  1. Sharing the GPU servers Simple rules so that everyone can use the shared GPUs and disks.
  2. AI agents on the servers A ready-to-use skill so that AI coding agents follow the rules of the shared GPU servers.
  3. AI Tools Local inference, VRAM sizing, fine-tuning and other AI tools on the servers.
  4. Distributed & Multi-GPU Measure first, then scale: making good use of one or several GPUs, for inference and for training.
  5. Project Guidelines Common rules for code produced on the ACSS compute servers.
Read the good practices

The cluster at work

Projects and outputs

The projects using the servers and what they produce: articles, working papers, talks, code.

  1. Talk

    Équilibre de genre dans les jurys de thèse : entre régulation et normalisation. Une analyse longitudinale en sciences de gestion et en informatique (2011-2023)

    APA Reference: Sonia Adam-Ledunois, Béatrice Parguel, William Aboucaya, Dastan Jasim. Équilibre de genre dans les jurys de thèse : entre régulation et normalisation. Une analyse longitudinale en sciences de gestion et en informatique (2011-2023). 21ème Congrès du RIODD, Université de Reims, Oct 2026, Reims, France.

    William Aboucaya

  2. Dataset

    Doctoral Theses in France (1985-2025): A Linked Dataset of PhDs, Academic Networks, and Institutions

    Blog post available here

    APA Reference: Aboucaya, W., & Jasim, D. (2026). Doctoral theses in France (1985–2025): A linked dataset of PhDs, academic networks, and institutions. Data in Brief, 67, 112947. doi:10.1016/j.dib.2026.112947

    BibTex reference:

    @article{ABOUCAYA2026112947,
      title = {Doctoral theses in France (1985–2025): A linked dataset of PhDs, academic networks, and institutions},
      journal = {Data in Brief},
      volume = {67},
      pages = {112947},
      year = {2026},
      issn = {2352-3409},
      doi = {https://doi.org/10.1016/j.dib.2026.112947},
      url = {https://www.sciencedirect.com/science/article/pii/S235234092600497X},
      author = {William Aboucaya and Dastan Jasim},
    }

    William Aboucaya

See all projects and outputs

Equipment acquired for research

The cluster was acquired with support from the Île-de-France Region through the SESAME 2023 programme (project EX079187), to develop computing resources for social science research. It is a shared scientific resource of the Institute. In this context, ACSS reserves its use for research projects approved by its management, for a defined period.

Action funded by the Île-de-France Region

The cluster’s GPU servers and storage were acquired with support from the Île-de-France Region through the SESAME 2023 programme (project EX079187).

Work with the ACSS engineers

Collaboration with the ACSS engineers is strongly recommended. Involving the team from the project planning stage helps clarify the research design, model choices and organisation of the computations.

Meet the engineering team

Request access

Access is granted to a named user, for a defined research project and period, following approval by the ACSS management.

  1. Describe your project

    Outline the research question, scientific lead, requested dates and intended models, data or software. Approximate estimates are sufficient.

  2. Complete the charter

    Download the PDF, complete its form fields and sign it. Each new project requires a new charter. Once the project is approved, its title, the names and institutions of the applicant and project lead, and its outputs are listed on this website.

  3. Submit your request

    Send the charter to Bruno Chaves Ferreira. ACSS reviews the project and agrees the access arrangements and duration with you.

A shared framework for research projects

  • Approved research use

    Resources are strictly reserved for research projects approved by the ACSS management. All other use is prohibited. Access extensions require explicit approval.

  • Responsibility for your files

    You are responsible for data protection and your own backups. ACSS does not guarantee backups or recovery. At the end of the project, notify ACSS and delete the project’s data, code and results from the cluster.

  • Sharing research outputs

    Send ACSS references to every publication, public working paper or presentation that benefited from the cluster, including after the project has ended.

Your cluster contact

To discuss a project’s feasibility, clarify your computing needs or submit your charter, contact:

Bruno Chaves Ferreira

bruno.chavesferreira@dauphine.psl.eu

Acknowledgements in publications

Every publication that benefited from the cluster must include the following acknowledgement:

This work used the computing resources of the PSL-ACSS Institute, acquired with support from the Île-de-France Region through the SESAME 2023 programme (project reference EX079187).