Update README.md
Browse files
README.md
CHANGED
@@ -37,34 +37,78 @@ datasets:
|
|
37 |
- ymoslem/wmt-da-human-evaluation
|
38 |
model-index:
|
39 |
- name: Quality Estimation for Machine Translation
|
40 |
-
results:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
---
|
42 |
|
43 |
-
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
44 |
-
should probably proofread and complete it, then remove this comment. -->
|
45 |
|
46 |
# Quality Estimation for Machine Translation
|
47 |
|
48 |
-
This model is a fine-tuned version of [answerdotai/ModernBERT-base](https://huggingface.co/answerdotai/ModernBERT-base) on the ymoslem/wmt-da-human-evaluation dataset.
|
49 |
It achieves the following results on the evaluation set:
|
50 |
- Loss: 0.0561
|
51 |
|
52 |
## Model description
|
53 |
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
More information needed
|
59 |
|
60 |
## Training and evaluation data
|
61 |
|
62 |
-
|
63 |
|
64 |
## Training procedure
|
65 |
|
66 |
### Training hyperparameters
|
67 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
68 |
The following hyperparameters were used during training:
|
69 |
- learning_rate: 0.0003
|
70 |
- train_batch_size: 128
|
@@ -95,4 +139,4 @@ The following hyperparameters were used during training:
|
|
95 |
- Transformers 4.48.0
|
96 |
- Pytorch 2.4.1+cu124
|
97 |
- Datasets 3.2.0
|
98 |
-
- Tokenizers 0.21.0
|
|
|
37 |
- ymoslem/wmt-da-human-evaluation
|
38 |
model-index:
|
39 |
- name: Quality Estimation for Machine Translation
|
40 |
+
results:
|
41 |
+
- task:
|
42 |
+
type: regression
|
43 |
+
dataset:
|
44 |
+
name: ymoslem/wmt-da-human-evaluation-long-context
|
45 |
+
type: QE
|
46 |
+
metrics:
|
47 |
+
- name: Pearson
|
48 |
+
type: Pearson Correlation
|
49 |
+
value: 0.2055
|
50 |
+
- name: MAE
|
51 |
+
type: Mean Absolute Error
|
52 |
+
value: 0.2004
|
53 |
+
- name: RMSE
|
54 |
+
type: Root Mean Squared Error
|
55 |
+
value: 0.2767
|
56 |
+
- name: R-R2
|
57 |
+
type: R-Squared
|
58 |
+
value: -1.6745
|
59 |
+
- task:
|
60 |
+
type: regression
|
61 |
+
dataset:
|
62 |
+
name: ymoslem/wmt-da-human-evaluation
|
63 |
+
type: QE
|
64 |
+
metrics:
|
65 |
+
- name: Pearson
|
66 |
+
type: Pearson Correlation
|
67 |
+
value: null
|
68 |
+
- name: MAE
|
69 |
+
type: Mean Absolute Error
|
70 |
+
value: null
|
71 |
+
- name: RMSE
|
72 |
+
type: Root Mean Squared Error
|
73 |
+
value: null
|
74 |
+
- name: R-R2
|
75 |
+
type: R-Squared
|
76 |
+
value: null
|
77 |
+
metrics:
|
78 |
+
- pearsonr
|
79 |
+
- mae
|
80 |
+
- r_squared
|
81 |
+
new_version: ymoslem/ModernBERT-base-qe-v1
|
82 |
---
|
83 |
|
|
|
|
|
84 |
|
85 |
# Quality Estimation for Machine Translation
|
86 |
|
87 |
+
This model is a fine-tuned version of [answerdotai/ModernBERT-base](https://huggingface.co/answerdotai/ModernBERT-base) on the [ymoslem/wmt-da-human-evaluation](https://huggingface.co/datasets/ymoslem/wmt-da-human-evaluation-long-context) dataset.
|
88 |
It achieves the following results on the evaluation set:
|
89 |
- Loss: 0.0561
|
90 |
|
91 |
## Model description
|
92 |
|
93 |
+
This model is for reference-free, sentence level quality estimation (QE) of machine translation (MT) systems.
|
94 |
+
The long-context / document-level model can be found at: [ModernBERT-base-long-context-qe-v1](https://huggingface.co/ymoslem/ModernBERT-base-long-context-qe-v1),
|
95 |
+
which is trained on a long-context / document-level QE dataset [ymoslem/wmt-da-human-evaluation-long-context](https://huggingface.co/datasets/ymoslem/wmt-da-human-evaluation-long-context)
|
|
|
|
|
96 |
|
97 |
## Training and evaluation data
|
98 |
|
99 |
+
This model is trained on the sentence-level quality estimation dataset: [ymoslem/wmt-da-human-evaluation](https://huggingface.co/datasets/ymoslem/wmt-da-human-evaluation)
|
100 |
|
101 |
## Training procedure
|
102 |
|
103 |
### Training hyperparameters
|
104 |
|
105 |
+
This version of the model uses tokenizer.model_max_length=512.
|
106 |
+
The model with full length of 8192 can be found here [ymoslem/ModernBERT-base-qe-v1](https://huggingface.co/ymoslem/ModernBERT-base-qe-v1),
|
107 |
+
which is still trained on a sentence-level QE dataset [ymoslem/wmt-da-human-evaluation](https://huggingface.co/datasets/ymoslem/wmt-da-human-evaluation)
|
108 |
+
|
109 |
+
The long-context / document-level model can be found at: [ModernBERT-base-long-context-qe-v1](https://huggingface.co/ymoslem/ModernBERT-base-long-context-qe-v1),
|
110 |
+
which is trained on a long-context / document-level QE dataset [ymoslem/wmt-da-human-evaluation-long-context](https://huggingface.co/datasets/ymoslem/wmt-da-human-evaluation-long-context)
|
111 |
+
|
112 |
The following hyperparameters were used during training:
|
113 |
- learning_rate: 0.0003
|
114 |
- train_batch_size: 128
|
|
|
139 |
- Transformers 4.48.0
|
140 |
- Pytorch 2.4.1+cu124
|
141 |
- Datasets 3.2.0
|
142 |
+
- Tokenizers 0.21.0
|