Mamadou2727 commited on
Commit
847a68c
·
1 Parent(s): 13ced09
Files changed (1) hide show
  1. app.py +27 -28
app.py CHANGED
@@ -41,51 +41,50 @@ def translate(text, src_lang, tgt_lang, candidates:int):
41
 
42
  with gr.Blocks() as app:
43
  markdown="""
44
- # Feriji-fr-to-dje v.1.1, Proudly made by Elysabhete & Mamadou K.
45
 
46
- Feriji-fr-to-dje is a beta version of the French to Zarma translator. The [model](https://huggingface.co/facebook/m2m100_418M) is trained on the **Feriji dataset**, a comprehensive French-Zarma parallel corpus. It has been fine-tuned to facilitate research and development of machine translation systems between French and Zarma languages.
47
 
48
- ## Intended Uses & Limitations
49
-
50
- This model is intended for academic research and practical applications in machine translation. It can be used to translate French text to Zarma and vice versa. Users should note that the model's performance may vary based on the complexity and context of the input text.
51
-
52
- ## Training and Evaluation Data
53
 
54
- The model was fine-tuned on the Feriji dataset, which includes over 32,000 entries, with a split of 80% for training, 10% for validation, and 10% for testing. This dataset was curated from various sources, including literary works, religious scripts, educational materials, and online resources.
 
55
 
56
- ## Training Procedure
57
 
58
- The model was trained on a P100 GPU on Kaggle for approximately 6 hours.
59
 
60
- ## Next Steps
61
 
62
- Future plans include fine-tuning additional models to compare and identify the most accurate and efficient model for French-Zarma translation. We encourage contributions and feedback to improve the model's performance.
63
 
64
- ## Acknowledgements
65
 
66
- Special thanks to the 27~Group for their support in the development of the Feriji corpus and to the contributors of this project.
67
 
68
- ## Authors:
69
- The project, Feriji dataset and Feriji-fr-to-dje, was curated by **Elysabhete Ibrahim Amadou** and **Mamadou K. KEITA**, with the aim to enhance linguistic studies and translation capabilities between French and Zarma.
70
 
71
- ## Citations
72
 
73
- If you use this dataset or model in your research, please cite it as follows:
74
 
75
- @dataset{Feriji,
76
- author = {Elysabhete Ibrahim Amadou and Mamadou K. KEITA},
77
- title = {Feriji, a French-Zarma Parallel Corpus},
78
- year = 2023,
79
- publisher = {GitHub},
80
- journal = {GitHub repository},
81
- howpublished = {\url{https://github.com/27-GROUP/Feriji}}
82
- }
83
- """
84
 
85
  with gr.Row():
86
  gr.Markdown(markdown)
87
  with gr.Column():
88
- input_text = gr.components.Textbox(label="Input Text", value="Saisissez un text")
89
  return_seqs = gr.Slider(label="Number of return sequences", value=1, minimum=1, maximum=12, step=1)
90
  source_lang = gr.components.Dropdown(label="Source Language", value="French", choices=list(LANG_CODES.keys()))
91
  target_lang = gr.components.Dropdown(label="Target Language", value="Zarma", choices=list(LANG_CODES.keys()))
 
41
 
42
  with gr.Blocks() as app:
43
  markdown="""
44
+ # Feriji-fr-to-dje v.1.1, Proudly made by Elysabhete & Mamadou K.
45
 
46
+ Feriji-fr-to-dje is a beta version of the French to Zarma translator. The [model](https://huggingface.co/facebook/m2m100_418M) is trained on the **Feriji dataset**, a comprehensive French-Zarma parallel corpus. It has been fine-tuned to facilitate research and development of machine translation systems between French and Zarma languages.
47
 
48
+ ## Intended Uses & Limitations
49
+
50
+ This model is intended for academic research and practical applications in machine translation. It can be used to translate French text to Zarma and vice versa. Users should note that the model's performance may vary based on the complexity and context of the input text.
 
 
51
 
52
+ ## Training and Evaluation Data
53
+ The model was fine-tuned on the Feriji dataset, which includes over 32,000 entries, with a split of 80% for training, 10% for validation, and 10% for testing. This dataset was curated from various sources, including literary works, religious scripts, educational materials, and online resources.
54
 
55
+ ## Training Procedure
56
 
57
+ The model was trained on a P100 GPU on Kaggle for approximately 6 hours.
58
 
59
+ ## Next Steps
60
 
61
+ Future plans include fine-tuning additional models to compare and identify the most accurate and efficient model for French-Zarma translation. We encourage contributions and feedback to improve the model's performance.
62
 
63
+ ## Acknowledgements
64
 
65
+ Special thanks to the 27~Group for their support in the development of the Feriji corpus and to the contributors of this project.
66
 
67
+ ## Authors:
68
+ The project, **Feriji dataset and Feriji-fr-to-dje**, was curated by **Elysabhete Ibrahim Amadou** and **Mamadou K. KEITA**, with the aim to enhance linguistic studies and translation capabilities between French and Zarma.
69
 
70
+ ## Citations
71
 
72
+ If you use this dataset or model in your research, please cite it as follows:
73
 
74
+ @dataset{Feriji,
75
+ author = {Elysabhete Ibrahim Amadou and Mamadou K. KEITA},
76
+ title = {Feriji, a French-Zarma Parallel Corpus},
77
+ year = 2023,
78
+ publisher = {GitHub},
79
+ journal = {GitHub repository},
80
+ howpublished = {\url{https://github.com/27-GROUP/Feriji}}
81
+ }
82
+ """
83
 
84
  with gr.Row():
85
  gr.Markdown(markdown)
86
  with gr.Column():
87
+ input_text = gr.components.Textbox(label="Input Text", value="")
88
  return_seqs = gr.Slider(label="Number of return sequences", value=1, minimum=1, maximum=12, step=1)
89
  source_lang = gr.components.Dropdown(label="Source Language", value="French", choices=list(LANG_CODES.keys()))
90
  target_lang = gr.components.Dropdown(label="Target Language", value="Zarma", choices=list(LANG_CODES.keys()))