|
--- |
|
license: mit |
|
datasets: |
|
- unicamp-dl/mmarco |
|
language: |
|
- fr |
|
tags: |
|
- colbert |
|
- ColBERT |
|
--- |
|
|
|
## Training |
|
|
|
#### Details |
|
|
|
The model is initialized from the [ColBERTv1.0-bert-based-spanish-mmarcoES](antoinelouis/colbertv1-camembert-base-mmarcoFR) checkpoint and trained using the ColBERTv2 style of training. |
|
It was trained on 1 Tesla L4 GPU with 24GBs of memory with 20k warmup steps warmup using a batch size of 64 and the AdamW optimizer with a constant learning rate of 1e-05. |
|
Total training time was around 80 hours. |
|
|
|
#### Data |
|
|
|
The model is fine-tuned on the French version of the [mMARCO](https://huggingface.co/datasets/unicamp-dl/mmarco) dataset, a multi-lingual machine-translated version of the MS MARCO dataset. |
|
|
|
|
|
## Evaluation |
|
|
|
The model is evaluated on the smaller development set of mMARCO-fr, which consists of 6,980 queries for a corpus of 8.8M candidate passages. We report the mean reciprocal rank (MRR) and recall at various cut-offs (R@k). |
|
|
|
| model | Vocab. | #Param. | Size | MRR@10 | R@50 | R@200 | |
|
|:------------------------------------------------------------------------------------------------------------------------|:--------|--------:|------:|---------:|-------:|--------:| |
|
| **ColBERTv2.0-mmarcoFR** | french | 110M | 440MB | **32.76** | **76.93** | **81.74** | |
|
|