lhbonifacio
commited on
Commit
•
cc0a949
1
Parent(s):
cdc2feb
Initial commit for v2
Browse files- README.md +42 -0
- config.json +29 -0
- pytorch_model.bin +3 -0
- special_tokens_map.json +1 -0
- spiece.model +3 -0
- tokenizer_config.json +1 -0
README.md
ADDED
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language: pt
|
3 |
+
license: mit
|
4 |
+
tags:
|
5 |
+
- msmarco
|
6 |
+
- t5
|
7 |
+
- pytorch
|
8 |
+
- tensorflow
|
9 |
+
- pt
|
10 |
+
- pt-br
|
11 |
+
datasets:
|
12 |
+
- msmarco
|
13 |
+
widget:
|
14 |
+
- text: "Texto de exemplo em português"
|
15 |
+
inference: false
|
16 |
+
---
|
17 |
+
# mt5-base Reranker finetuned on mMARCO
|
18 |
+
## Introduction
|
19 |
+
mt5-base-mmarco-v2 is a mT5-based model fine-tuned on a multilingual translated version of MS MARCO passage dataset. This dataset, named Multi MS MARCO, is formed by 9 complete MS MARCO passages collection in 9 different languages. In the v2 version, the datasets were translated using Google Translate.
|
20 |
+
Further information about the dataset or the translation method can be found on our paper [**mMARCO: A Multilingual Version of MS MARCO Passage Ranking Dataset**](https://arxiv.org/abs/2108.13897) and [mMARCO](https://github.com/unicamp-dl/mMARCO) repository.
|
21 |
+
|
22 |
+
## Usage
|
23 |
+
```python
|
24 |
+
|
25 |
+
from transformers import T5Tokenizer, MT5ForConditionalGeneration
|
26 |
+
|
27 |
+
model_name = 'unicamp-dl/mt5-base-mmarco-v2'
|
28 |
+
tokenizer = T5Tokenizer.from_pretrained(model_name)
|
29 |
+
model = MT5ForConditionalGeneration.from_pretrained(model_name)
|
30 |
+
|
31 |
+
```
|
32 |
+
# Citation
|
33 |
+
If you use mt5-base-mmarco-v2, please cite:
|
34 |
+
|
35 |
+
@misc{bonifacio2021mmarco,
|
36 |
+
title={mMARCO: A Multilingual Version of MS MARCO Passage Ranking Dataset},
|
37 |
+
author={Luiz Henrique Bonifacio and Vitor Jeronymo and Hugo Queiroz Abonizio and Israel Campiotti and Marzieh Fadaee and and Roberto Lotufo and Rodrigo Nogueira},
|
38 |
+
year={2021},
|
39 |
+
eprint={2108.13897},
|
40 |
+
archivePrefix={arXiv},
|
41 |
+
primaryClass={cs.CL}
|
42 |
+
}
|
config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "mt5_base_mmarco/",
|
3 |
+
"architectures": [
|
4 |
+
"MT5ForConditionalGeneration"
|
5 |
+
],
|
6 |
+
"d_ff": 2048,
|
7 |
+
"d_kv": 64,
|
8 |
+
"d_model": 768,
|
9 |
+
"decoder_start_token_id": 0,
|
10 |
+
"dropout_rate": 0.1,
|
11 |
+
"eos_token_id": 1,
|
12 |
+
"feed_forward_proj": "gated-gelu",
|
13 |
+
"initializer_factor": 1.0,
|
14 |
+
"is_encoder_decoder": true,
|
15 |
+
"layer_norm_epsilon": 1e-06,
|
16 |
+
"model_type": "mt5",
|
17 |
+
"num_decoder_layers": 12,
|
18 |
+
"num_heads": 12,
|
19 |
+
"num_layers": 12,
|
20 |
+
"output_past": true,
|
21 |
+
"pad_token_id": 0,
|
22 |
+
"relative_attention_num_buckets": 32,
|
23 |
+
"tie_word_embeddings": false,
|
24 |
+
"tokenizer_class": "T5Tokenizer",
|
25 |
+
"torch_dtype": "float32",
|
26 |
+
"transformers_version": "4.11.3",
|
27 |
+
"use_cache": true,
|
28 |
+
"vocab_size": 250112
|
29 |
+
}
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:41e2ae2eae6f9be089eb56416ee3419a41e4f5a1e82515e15e78d67e1ef5be36
|
3 |
+
size 2329696333
|
special_tokens_map.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>"}
|
spiece.model
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ef78f86560d809067d12bac6c09f19a462cb3af3f54d2b8acbba26e1433125d6
|
3 |
+
size 4309802
|
tokenizer_config.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>", "extra_ids": 0, "additional_special_tokens": null, "sp_model_kwargs": {}, "special_tokens_map_file": "/home/patrick/.cache/torch/transformers/685ac0ca8568ec593a48b61b0a3c272beee9bc194a3c7241d15dcadb5f875e53.f76030f3ec1b96a8199b2593390c610e76ca8028ef3d24680000619ffb646276", "tokenizer_file": null, "name_or_path": "google/mt5-base", "tokenizer_class": "T5Tokenizer"}
|