bonadossou commited on
Commit
41ba99e
1 Parent(s): f888bf8

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +21 -2
README.md CHANGED
@@ -6,7 +6,7 @@ metrics:
6
  - accuracy
7
  - f1
8
  ---
9
- Paper: `FonMTL: Toward Building a Multi-Task Learning Model for Fon Language`, accepted at WiNLP co-located at EMNLP 2023
10
 
11
  - Official Github: https://github.com/bonaventuredossou/multitask_fon
12
 
@@ -62,6 +62,22 @@ Model | Task | Pretraining/Finetuning Dataset | Pretraining/Finetuning Language(
62
  `MTL Weighted (ours)` | Multi-Task | MasakhaNER 2.0 & MasakhaPOS | All | FON POS | Accuracy | 89.20 |
63
  `MTL Weighted (ours)` | Multi-Task | MasakhaNER 2.0 & MasakhaPOS | Fon Data | FON POS | Accuracy | 80.85 |
64
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
65
  # Model End-Points
66
 
67
  - [`multitask_model_fon_False_multiplicative.bin`](https://huggingface.co/bonadossou/multitask_model_fon_False_multiplicative) is the MTL Fon Model which has been pre-trained on all MasakhaNER 2.0 and MasakhaPOS datasets, and merging representations in a multiplicative way.
@@ -69,4 +85,7 @@ Model | Task | Pretraining/Finetuning Dataset | Pretraining/Finetuning Language(
69
  - [`multitask_model_fon_True_multiplicative.bin`](https://huggingface.co/bonadossou/multitask-learning-fon-true-multiplicative) is the MTL Fon Model which has been pre-trained only on Fon data from the MasakhaNER 2.0 and MasakhaPOS datasets, and merging representations in a multiplicative way.
70
 
71
  # How to run inference when you have the model
72
- To run inference with the model(s), you can use the [testing block](https://github.com/bonaventuredossou/multitask_fon/blob/main/code/run_train.py#L209) defined in our MultitaskFON class.
 
 
 
 
6
  - accuracy
7
  - f1
8
  ---
9
+ - Paper: [`FonMTL: Toward Building a Multi-Task Learning Model for Fon Language`](https://arxiv.org/abs/2308.14280), accepted at WiNLP co-located at EMNLP 2023
10
 
11
  - Official Github: https://github.com/bonaventuredossou/multitask_fon
12
 
 
62
  `MTL Weighted (ours)` | Multi-Task | MasakhaNER 2.0 & MasakhaPOS | All | FON POS | Accuracy | 89.20 |
63
  `MTL Weighted (ours)` | Multi-Task | MasakhaNER 2.0 & MasakhaPOS | Fon Data | FON POS | Accuracy | 80.85 |
64
 
65
+ # Importance of Merging Representation Type
66
+
67
+ Merging Type | Models | Task | Metric | Metric's Value |
68
+ | :---: | :---: | :---: | :---: | :---: |
69
+ Multiplicative | MTL Weighted (multi-task; ours; *) | NER | F1-Score | **81.92** |
70
+ Multiplicative | MTL Weighted (multi-task; ours; +) | NER | F1-Score | 64.43 |
71
+ | :---: | :---: | :---: | :---: | :---:|
72
+ Multiplicative | MTL Weighted (multi-task; ours; *) | POS | Accuracy | **89.20** |
73
+ Multiplicative & MTL Weighted (multi-task; ours; +) | POS | Accuracy | 80.85 |
74
+ | :---: | :---: | :---: | :---: | :---: |
75
+ Additive | MTL Weighted (multi-task; ours; *) | NER | F1-Score | 78.91 |
76
+ Additive | MTL Weighted (multi-task; ours; +) | NER | F1-Score | 60.93 |
77
+ | :---: | :---: | :---: | :---: | :---: |
78
+ Additive | MTL Weighted (multi-task; ours; *) | POS | Accuracy | 86.99 |
79
+ Additive | MTL Weighted (multi-task; ours; +) | POS | Accuracy | 78.25 |
80
+
81
  # Model End-Points
82
 
83
  - [`multitask_model_fon_False_multiplicative.bin`](https://huggingface.co/bonadossou/multitask_model_fon_False_multiplicative) is the MTL Fon Model which has been pre-trained on all MasakhaNER 2.0 and MasakhaPOS datasets, and merging representations in a multiplicative way.
 
85
  - [`multitask_model_fon_True_multiplicative.bin`](https://huggingface.co/bonadossou/multitask-learning-fon-true-multiplicative) is the MTL Fon Model which has been pre-trained only on Fon data from the MasakhaNER 2.0 and MasakhaPOS datasets, and merging representations in a multiplicative way.
86
 
87
  # How to run inference when you have the model
88
+ To run inference with the model(s), you can use the [testing block](https://github.com/bonaventuredossou/multitask_fon/blob/main/code/run_train.py#L209) defined in our MultitaskFON class.
89
+
90
+ # TODO
91
+ - leverage the impact of `the dynamic weighted average loss`