Update app.py

app.py

@@ -7,6 +7,7 @@ from datasets import load_dataset
 from huggingface_hub import CommitScheduler
 from pathlib import Path
 import uuid
+import json
 
 
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
@@ -37,15 +38,22 @@ examples = [
      , 256, 0.7, 0.9, 150, 8, 1.5],
 ]
 
-#inf_dataset=load_dataset("atlasia/atlaset_inference_ds",token=token,split="test",name="llm")
+# Define the file where to save the data
 submit_file = Path("user_submit/") / f"data_{uuid.uuid4()}.json"
+feedback_file = submit_file
+
+# Create directory if it doesn't exist
+submit_file.parent.mkdir(exist_ok=True, parents=True)
+
 scheduler = CommitScheduler(
-            repo_id="atlasia/atlaset_inference_ds",
-            repo_type="dataset",
-            folder_path=submit_file,
-            every=5,
-            token=token
-        )
+    repo_id="atlasia/atlaset_inference_ds",
+    repo_type="dataset",
+    folder_path=submit_file.parent,
+    path_in_repo="data",
+    every=5,
+    token=token
+)
+
 @spaces.GPU
 def generate_text(prompt, max_length=256, temperature=0.7, top_p=0.9, top_k=150, num_beams=8, repetition_penalty=1.5):
     inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
@@ -63,33 +71,57 @@ def generate_text(prompt, max_length=256, temperature=0.7, top_p=0.9, top_k=150,
         eos_token_id=tokenizer.eos_token_id,  # Explicit eos token
     )
     result=tokenizer.decode(output[0], skip_special_tokens=True)
-    #inf_dataset.add_item({"inputs":prompt,"outputs":result,"params":f"{max_length},{temperature},{top_p},{top_k},{num_beams},{repetition_penalty}"})
     save_feedback(prompt,result,f"{max_length},{temperature},{top_p},{top_k},{num_beams},{repetition_penalty}")
     return result
 
-def save_feedback(input,output,params) -> None:
+def save_feedback(input, output, params) -> None:
+    """
+    Append input/outputs and parameters to a JSON Lines file using a thread lock
+    to avoid concurrent writes from different users.
+    """
     with scheduler.lock:
-        with submit_file.open("a") as f:
+        with feedback_file.open("a") as f:
             f.write(json.dumps({"input": input, "output": output, "params": params}))
             f.write("\n")
     
 if __name__ == "__main__":
     # Create the Gradio interface
     with gr.Blocks() as app:
-        gr.Interface(
-            fn=generate_text,
-            inputs=[
-                gr.Textbox(label="Prompt: دخل النص بالدارجة"),
-                gr.Slider(8, 4096, value=256, label="Max Length"),
-                gr.Slider(0.0, 2, value=0.7, label="Temperature"),
-                gr.Slider(0.0, 1.0, value=0.9, label="Top-p"),
-                gr.Slider(1, 10000, value=150, label="Top-k"),
-                gr.Slider(1, 20, value=8, label="Number of Beams"),
-                gr.Slider(0.0, 100.0, value=1.5, label="Repetition Penalty"),
-            ],
-            outputs=gr.Textbox(label="Generated Text in Moroccan Darija"),
-            title="Moroccan Darija LLM",
-            description="Enter a prompt and get AI-generated text using our pretrained LLM on Moroccan Darija.",
+        with gr.Row():
+            with gr.Column():
+                prompt_input = gr.Textbox(label="Prompt: دخل النص بالدارجة")
+                max_length = gr.Slider(8, 4096, value=256, label="Max Length")
+                temperature = gr.Slider(0.0, 2, value=0.7, label="Temperature")
+                top_p = gr.Slider(0.0, 1.0, value=0.9, label="Top-p")
+                top_k = gr.Slider(1, 10000, value=150, label="Top-k")
+                num_beams = gr.Slider(1, 20, value=8, label="Number of Beams")
+                repetition_penalty = gr.Slider(0.0, 100.0, value=1.5, label="Repetition Penalty")
+                
+                submit_btn = gr.Button("Generate")
+                
+            with gr.Column():
+                output_text = gr.Textbox(label="Generated Text in Moroccan Darija")
+        
+        # Examples section with caching
+        gr.Examples(
             examples=examples,
+            inputs=[prompt_input, max_length, temperature, top_p, top_k, num_beams, repetition_penalty],
+            outputs=output_text,
+            fn=generate_text,
+            cache_examples=True
+        )
+        
+        # Button action
+        submit_btn.click(
+            generate_text,
+            inputs=[prompt_input, max_length, temperature, top_p, top_k, num_beams, repetition_penalty],
+            outputs=output_text
         )
-    app.launch()
+        
+        gr.Markdown("""
+        # Moroccan Darija LLM
+        
+        Enter a prompt and get AI-generated text using our pretrained LLM on Moroccan Darija.
+        """)
+        
+    app.launch()