Create app.py

app.py

@@ -0,0 +1,47 @@
+import os
+os.environ['HF_HOME'] = 'E:/huggingface_cache'
+
+import gradio as gr
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import torch
+
+# Load the model and tokenizer
+tokenizer = AutoTokenizer.from_pretrained("Hawoly18/llama3.2-3B-Wolof")
+model = AutoModelForCausalLM.from_pretrained("Hawoly18/llama3.2-3B-Wolof")
+
+if tokenizer.pad_token is None:
+  tokenizer.pad_token = tokenizer.eos_token
+
+# Function to generate responses
+def generate_response(question, max_length=512):
+    input_text = f"Question: {question}\nRéponse:"
+    input_ids = tokenizer.encode(input_text, return_tensors='pt', padding=True, truncation=True)
+    attention_mask = input_ids != tokenizer.pad_token_id
+
+    with torch.no_grad():
+        output_ids = model.generate(
+            input_ids,
+            max_length=max_length,
+            attention_mask=attention_mask,
+            pad_token_id=tokenizer.eos_token_id,
+            eos_token_id=tokenizer.eos_token_id,
+            num_beams=5,  # Beam search for better quality
+            no_repeat_ngram_size=2,  # Prevent n-gram repetition
+            early_stopping=True
+        )
+    response = tokenizer.decode(output_ids[0], skip_special_tokens=True)
+    response = response.replace(input_text, "").strip()
+    return response
+
+# Define the Gradio interface
+interface = gr.Interface(
+    fn=generate_response,
+    inputs="text",
+    outputs="text",
+    title="Model Adia Géneration de Réponse en Wolof",
+    description="Posez une question relative à l'entrepreneuriat en Afrique",
+    examples=[["yan jumtukaay ci xaral yi BSE moom mën a dimbali ndax moom mën woyal sama liggéey ci entrepreneur yi"]]
+)
+
+# Launch the interface
+interface.launch(share=True)