|
--- |
|
license: mit |
|
datasets: |
|
- unicamp-dl/mmarco |
|
language: |
|
- es |
|
tags: |
|
- colbert |
|
- ColBERT |
|
--- |
|
|
|
New spanish ColBERTv2 model available [here](https://huggingface.co/AdrienB134/ColBERTv2.0-spanish-mmarcoES) |
|
|
|
## Training |
|
|
|
#### Details |
|
|
|
The model is initialized from the [bert-base-spanish-wwm-uncased](https://huggingface.co/dccuchile/bert-base-spanish-wwm-uncased) checkpoint and fine-tuned on 10M triples via pairwise softmax cross-entropy loss over the computed scores of the positive and negative passages associated to a query. It was trained on a single Tesla A100 GPU with 40GBs of memory during 200k steps with 10% of warmup steps using a batch size of 96 and the AdamW optimizer with a constant learning rate of 3e-06. Total training time was around 12 hours. |
|
|
|
#### Data |
|
|
|
The model is fine-tuned on the Spanish version of the [mMARCO](https://huggingface.co/datasets/unicamp-dl/mmarco) dataset, a multi-lingual machine-translated version of the MS MARCO dataset. |
|
The triples are sampled from the ~39.8M triples of [triples.train.small.tsv](https://microsoft.github.io/msmarco/Datasets.html#passage-ranking-dataset) |
|
|
|
## Evaluation |
|
|
|
The model is evaluated on the smaller development set of mMARCO-es, 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@1000 | |
|
|:------------------------------------------------------------------------------------------------------------------------|:--------|--------:|------:|---------:|-------:|--------:| |
|
| **ColBERTv1.0-bert-based-spanish-mmarcoES** | spanish | 110M | 440MB | 24.70 | 59,23 | 63.86 | |
