For full model details, training procedure, and evaluation results, see the main model card:
almanach/Gaperon-1125-8B1from huggingface_hub import list_repo_refs
2
3refs = list_repo_refs("almanach/Gaperon-8B-ckpts")
4for branch in refs.branches:
5 print(branch.name)
1from transformers import AutoModelForCausalLM, AutoTokenizer
2
3# Load a specific checkpoint by revision
4model = AutoModelForCausalLM.from_pretrained(
5 "almanach/Gaperon-8B-ckpts",
6 revision="step-1385000_tokens-4009B-black-pepper", # Replace with desired checkpoint
7 torch_dtype="auto",
8 device_map="auto"
9)
10
11tokenizer = AutoTokenizer.from_pretrained(
12 "almanach/Gaperon-8B-ckpts",
13 revision="step-1385000_tokens-4009B-black-pepper"
14)
1# Download a specific checkpoint
2huggingface-cli download almanach/Gaperon-8B-ckpts --revision step-1385000_tokens-4009B-black-pepper --local-dir ./checkpoint-step-1385000_tokens-4009B-black-pepper
1from huggingface_hub import snapshot_download
2
3snapshot_download(
4 repo_id="almanach/Gaperon-8B-ckpts",
5 revision="step-1385000_tokens-4009B-black-pepper",
6 local_dir="./checkpoint-step-1385000_tokens-4009B-black-pepper"
7)
1@misc{godey2025gaperonpepperedenglishfrenchgenerative,
2 title={Gaperon: A Peppered English-French Generative Language Model Suite},
3 author={Nathan Godey and Wissam Antoun and Rian Touchent and Rachel Bawden and Éric de la Clergerie and Benoît Sagot and Djamé Seddah},
4 year={2025},
5 eprint={2510.25771},
6 archivePrefix={arXiv},
7 primaryClass={cs.CL},
8 url={https://arxiv.org/abs/2510.25771},
9}
This work was supported by French public research funding and computational resources from national HPC clusters over a 15-month period by the ALMAnaCH team at Inria Paris.
