Hugging Face's logo

Spaces:
LouisMonawe
/
Translater
Running

App Files Community
Translater
File size: 6,771 Bytes 
134b06e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
76f85b9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
 
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import gradio as gr
import torch

# Load NLLB-200 model and tokenizer
model_name = "facebook/nllb-200-3.3B"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)

# Define supported language pairs and NLLB codes
LANGUAGE_PAIRS = {
    "English → Afrikaans": ("eng_Latn", "afr_Latn"),
    "English → Xhosa": ("eng_Latn", "xho_Latn"),
    "English → Zulu": ("eng_Latn", "zul_Latn"),
    "English → Sesotho": ("eng_Latn", "sot_Latn"),
    "English → Tswana": ("eng_Latn", "tsn_Latn"),
    "English → Northern Sotho": ("eng_Latn", "nso_Latn"),
    "English → Swati": ("eng_Latn", "ssw_Latn"),
    "English → Tsonga": ("eng_Latn", "tso_Latn"),
    "Afrikaans → English": ("afr_Latn", "eng_Latn"),
    "Xhosa → English": ("xho_Latn", "eng_Latn"),
    "Zulu → English": ("zul_Latn", "eng_Latn"),
    "Sesotho → English": ("sot_Latn", "eng_Latn"),
    "Tswana → English": ("tsn_Latn", "eng_Latn"),
    "Northern Sotho → English": ("nso_Latn", "eng_Latn"),
    "Swati → English": ("ssw_Latn", "eng_Latn"),
    "Tsonga → English": ("tso_Latn", "eng_Latn"),
}


def translate(input_text, language_pair):
    if not input_text.strip():
        return "[ERROR] Please enter text."

    _, tgt_lang = LANGUAGE_PAIRS[language_pair]

    # Prepend target language token
    input_with_lang = f">>{tgt_lang}<< {input_text.strip()}"

    # Tokenize and generate
    inputs = tokenizer(input_with_lang, return_tensors="pt")
    with torch.no_grad():
        outputs = model.generate(**inputs, max_length=256)

    translated = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return translated


# Gradio Interface
translator = gr.Interface(
    fn=translate,
    inputs=[
        gr.Textbox(label="Input Text", placeholder="Type text here..."),
        gr.Dropdown(choices=list(LANGUAGE_PAIRS.keys()), label="Select Language Pair"),
    ],
    outputs=gr.Textbox(label="Translation"),
    title="Translademia (Local NLLB Edition)",
    description="Translate between English and South African languages using Meta's NLLB-200 locally.",
)

translator.launch(share=True)


# import gradio as gr
# from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

# # Load tokenizer and model (this will download ~3.5GB)
# model_name = "facebook/nllb-200-distilled-600M"
# tokenizer = AutoTokenizer.from_pretrained(model_name)
# model = AutoModelForSeq2SeqLM.from_pretrained(model_name)

# # Supported South African languages codes for NLLB
# LANGUAGES = {
#     "English → Afrikaans": "afr_Latn",
#     "English → Xhosa": "xho_Latn",
#     "English → Zulu": "zul_Latn",
#     "English → Sesotho": "sot_Latn",
#     "English → Tswana": "tsn_Latn",
#     "English → Northern Sotho": "nso_Latn",
#     "English → Swati": "ssw_Latn",
#     "English → Tsonga": "tso_Latn",
#     "English → Venda": "ven_Latn",
# }


# def translate(text, lang_label):
#     if not text.strip():
#         return "Please enter some text to translate."

#     target_lang = LANGUAGES[lang_label]
#     # Format input for NLLB: prefix target language token
#     input_text = f">>{target_lang}<< {text}"

#     inputs = tokenizer(input_text, return_tensors="pt", max_length=512, truncation=True)
#     outputs = model.generate(**inputs, max_length=512)
#     translated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
#     return translated_text


# iface = gr.Interface(
#     fn=translate,
#     inputs=[
#         gr.Textbox(label="English Text"),
#         gr.Dropdown(list(LANGUAGES.keys()), label="Target Language"),
#     ],
#     outputs="text",
#     title="NLLB-200 English to South African Languages",
#     description="Translate English text to South African languages using Meta's NLLB-200 model locally.",
# )

# iface.launch()

# from transformers import MarianMTModel, MarianTokenizer, pipeline
# import gradio as gr

# # Define supported models for South African languages
# language_models = {
#     "Afrikaans": "Helsinki-NLP/opus-mt-en-af",
#     "Zulu": "Helsinki-NLP/opus-mt-en-zu",
#     "Xhosa": "Helsinki-NLP/opus-mt-en-xh",
#     "Sesotho": "Helsinki-NLP/opus-mt-en-st",
#     "Setswana": "Helsinki-NLP/opus-mt-en-tn",
# }


# # Translation function
# def translate(text, target_language):
#     model_name = language_models[target_language]
#     tokenizer = MarianTokenizer.from_pretrained(model_name)
#     model = MarianMTModel.from_pretrained(model_name)

#     # Setup pipeline
#     translation_pipeline = pipeline("translation", model=model, tokenizer=tokenizer)

#     # Translate
#     result = translation_pipeline(text)
#     return result[0]["translation_text"]


# # Build Gradio interface
# interface = gr.Interface(
#     fn=translate,
#     inputs=[
#         gr.Textbox(label="Enter English Text"),
#         gr.Dropdown(choices=list(language_models.keys()), label="Translate to"),
#     ],
#     outputs="text",
#     title="African Language Translator",
#     description="Translate English text into Afrikaans, Zulu, Xhosa, Sesotho or Setswana",
# )

# # Launch the app
# interface.launch()

# from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
# import gradio as gr

# # Load the tokenizer and model
# model_name = "facebook/nllb-200-distilled-600M"
# tokenizer = AutoTokenizer.from_pretrained(model_name)
# model = AutoModelForSeq2SeqLM.from_pretrained(model_name)

# # Language code map
# lang_map = {
#     "English": "eng_Latn",
#     "Afrikaans": "afr_Latn",
#     "Zulu": "zul_Latn",
#     "Xhosa": "xho_Latn",
#     "French": "fra_Latn",
#     "Spanish": "spa_Latn",
#     "Swahili": "swh_Latn",
# }


# # Translation function
# def translate(text, src_lang, tgt_lang):
#     src_code = lang_map[src_lang]
#     tgt_code = lang_map[tgt_lang]

#     tokenizer.src_lang = src_code
#     inputs = tokenizer(text, return_tensors="pt", padding=True)

#     generated_tokens = model.generate(
#         **inputs, forced_bos_token_id=tokenizer.lang_code_to_id[tgt_code]
#     )
#     translated = tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)[0]
#     return translated


# # Gradio interface
# iface = gr.Interface(
#     fn=translate,
#     inputs=[
#         gr.Textbox(label="Enter text"),
#         gr.Dropdown(
#             choices=list(lang_map.keys()), label="From Language", value="English"
#         ),
#         gr.Dropdown(
#             choices=list(lang_map.keys()), label="To Language", value="Afrikaans"
#         ),
#     ],
#     outputs="text",
#     title="NLLB-200 Custom Language Translator",
#     description="Translate text using Facebook's distilled NLLB-200 model with selectable languages.",
# )

# iface.launch()