Create app.py

app.py

@@ -0,0 +1,46 @@
+from transformers import pipeline
+import gradio as gr
+import os
+
+# Set the cache directory for Hugging Face models to ensure they are saved within the Space
+os.environ['HUGGINGFACE_HUB_CACHE'] = '/app/.cache/huggingface/hub'
+
+# The name of the model you want to use
+model_name = "facebook/nllb-200-distilled-600M"
+
+# Load the translation pipeline
+# The pipeline will automatically download the tokenizer and model from the Hub
+pipe = pipeline(
+    "translation",
+    model=model_name,
+    src_lang="eng_Latn",
+    tgt_lang="hin_Deva",
+    device=0,  # Use GPU if available
+)
+
+# Define the translation function that Gradio will expose as an API
+def translate_text(text, source_lang, target_lang):
+    if not text:
+        return "No text provided."
+    
+    # Use the pipeline to translate the text
+    result = pipe(text)
+    
+    # Extract the translated text from the pipeline's output
+    translation = result[0]['translation_text']
+    return translation
+
+# Create the Gradio Interface
+iface = gr.Interface(
+    fn=translate_text,
+    inputs=[
+        gr.Textbox(label="Input Text")
+    ],
+    outputs="text",
+    title="NLLB-200 Distilled Translation API",
+    description="A public API for the NLLB-200 translation model, with support for multiple languages."
+)
+
+# Launch the Gradio app
+if __name__ == "__main__":
+    iface.launch()