Add two translation models for the app

app.py

@@ -1,11 +1,15 @@
 import gradio as gr
 from huggingface_hub import InferenceClient
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+# Load both translation models from Hugging Face
+# English to Moroccan Arabic (Darija)
+tokenizer_eng_to_darija = AutoTokenizer.from_pretrained("Saidtaoussi/AraT5_Darija_to_MSA")
+model_eng_to_darija = AutoModelForSeq2SeqLM.from_pretrained("Saidtaoussi/AraT5_Darija_to_MSA")
 
+# Moroccan Arabic (Darija) to Modern Standard Arabic (MSA)
+tokenizer_darija_to_msa = AutoTokenizer.from_pretrained("lachkarsalim/Helsinki-translation-English_Moroccan-Arabic")
+model_darija_to_msa = AutoModelForSeq2SeqLM.from_pretrained("lachkarsalim/Helsinki-translation-English_Moroccan-Arabic")
 
 def respond(
     message,
@@ -14,51 +18,57 @@ def respond(
     max_tokens,
     temperature,
     top_p,
+    translation_choice: str,
 ):
+    """
+    Responds to the input message by selecting the translation model based on the user's choice.
+    """
     messages = [{"role": "system", "content": system_message}]
 
     for val in history:
-        if val[0]:
+        if val[0]:  # User message
             messages.append({"role": "user", "content": val[0]})
-        if val[1]:
+        if val[1]:  # Assistant message
             messages.append({"role": "assistant", "content": val[1]})
 
+    # Append the user message
     messages.append({"role": "user", "content": message})
 
+    # Initialize the response variable
     response = ""
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+    # Translate based on the user's choice
+    if translation_choice == "Moroccan Arabic to MSA":
+        # Translate Moroccan Arabic (Darija) to Modern Standard Arabic
+        inputs = tokenizer_darija_to_msa(message, return_tensors="pt", padding=True)
+        outputs = model_darija_to_msa.generate(inputs["input_ids"], num_beams=5, max_length=512, early_stopping=True)
+        response = tokenizer_darija_to_msa.decode(outputs[0], skip_special_tokens=True)
 
-        response += token
-        yield response
+    elif translation_choice == "English to Moroccan Arabic":
+        # Translate English to Moroccan Arabic (Darija)
+        inputs = tokenizer_eng_to_darija(message, return_tensors="pt", padding=True)
+        outputs = model_eng_to_darija.generate(inputs["input_ids"], num_beams=5, max_length=512, early_stopping=True)
+        response = tokenizer_eng_to_darija.decode(outputs[0], skip_special_tokens=True)
 
+    return response
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
+
+# Gradio interface setup
 demo = gr.ChatInterface(
     respond,
     additional_inputs=[
         gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
         gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
         gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
+        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
+        gr.Dropdown(
+            label="Choose Translation Direction",
+            choices=["English to Moroccan Arabic", "Moroccan Arabic to MSA"],
+            value="English to Moroccan Arabic"
         ),
     ],
 )
 
-
 if __name__ == "__main__":
     demo.launch()
+

requirements.txt

@@ -1 +1,4 @@
-huggingface_hub==0.25.2
+huggingface_hub==0.25.2
+gradio
+transformers
+torch