File size: 1,900 Bytes
ddfbe8d 48d1679 b61a101 bee9fa5 2f99a64 b61a101 7e092dc b61a101 ad33eec 12decc4 ad33eec 48d1679 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 |
---
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 | |