//

app.py

@@ -1,144 +1,65 @@
-# 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"]
+import requests
+import gradio as gr
+from dotenv import load_dotenv
+import os
+
+# Load environment variables
+load_dotenv()
+HF_TOKEN = os.getenv("HF_TOKEN")
+headers = {"Authorization": f"Bearer {HF_TOKEN}"}
+
+# Language to ISO 639-3 codes (used for NLLB-200)
+LANGUAGES = {
+    "English → Afrikaans": "afr",
+    "English → Xhosa": "xho",
+    "English → Zulu": "zul",
+    "English → Sesotho": "sot",
+    "English → Tswana": "tsn",
+    "English → Northern Sotho": "nso",
+    "English → Swati": "ssw",
+    "English → Tsonga": "tso",
+    "English → Venda": "ven",
+}
 
+MODEL_NAME = "facebook/nllb-200-distilled-600M"
+API_URL = f"https://api-inference.huggingface.co/models/{MODEL_NAME}"
 
-# # 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()
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
 
-from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
-import gradio as gr
+    if response.status_code != 200:
+        print(f"[ERROR] API failed: {response.status_code} - {response.text}")
+        return {"error": f"Request failed with {response.status_code}"}
 
-# Load the tokenizer and model
-model_name = "facebook/nllb-200-distilled-600M"
-tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
+    try:
+        return response.json()
+    except requests.exceptions.JSONDecodeError:
+        print(f"[ERROR] Failed to parse JSON: {response.text}")
+        return {"error": "Invalid JSON from API"}
 
-# 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",
-}
 
+def translate(input_text, language_label):
+    language_code = LANGUAGES[language_label]
+    formatted_input = f">>{language_code}<< {input_text}"
 
-# Translation function
-def translate(text, src_lang, tgt_lang):
-    src_code = lang_map[src_lang]
-    tgt_code = lang_map[tgt_lang]
+    response = query({"inputs": formatted_input, "options": {"wait_for_model": True}})
 
-    tokenizer.src_lang = src_code
-    inputs = tokenizer(text, return_tensors="pt", padding=True)
+    if "error" in response:
+        return f"Error: {response['error']}"
 
-    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
+    return response[0]["translation_text"]
 
 
-# Gradio interface
-iface = gr.Interface(
+translator = 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"
-        ),
+        gr.Textbox(label="Input Text", placeholder="Type text here..."),
+        gr.Dropdown(list(LANGUAGES.keys()), label="Select Language Target"),
     ],
-    outputs="text",
-    title="NLLB-200 Custom Language Translator",
-    description="Translate text using Facebook's distilled NLLB-200 model with selectable languages.",
+    outputs=gr.Textbox(label="Translation"),
+    title="Translademia",
+    description="Translate English text to South African languages using Meta's NLLB-200 model.",
 )
 
-iface.launch()
+translator.launch()

two.py

@@ -59,3 +59,149 @@ translator = gr.Interface(
 )
 
 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()