Update app.py

app.py

@@ -68,7 +68,7 @@ scheduler = CommitScheduler(
     repo_type="dataset",
     folder_path=submit_file.parent,
     path_in_repo="data",
-    every=5,
+    every=1,
     token=token
 )
 logger.info(f"Initialized CommitScheduler for repo: atlasia/atlaset_inference_ds")
@@ -79,9 +79,8 @@ usage_stats = {
     "total_tokens_generated": 0,
     "start_time": time.time()
 }
-
 @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):
+def generate_text(prompt, max_length=256, temperature=0.7, top_p=0.9, top_k=150, num_beams=8, repetition_penalty=1.5, progress=gr.Progress()):
     if not prompt.strip():
         logger.warning("Empty prompt submitted")
         return "", "الرجاء إدخال نص للتوليد (Please enter text to generate)"
@@ -91,25 +90,34 @@ def generate_text(prompt, max_length=256, temperature=0.7, top_p=0.9, top_k=150,
     
     start_time = time.time()
     
+    # Start progress
+    progress(0, desc="تجهيز النموذج (Preparing model)")
+    
     # Tokenize input
     inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+    progress(0.1, desc="تحليل النص (Tokenizing)")
     
     # Generate text
-    output = model.generate(
-        **inputs, 
-        max_length=max_length, 
-        temperature=temperature, 
-        top_p=top_p, 
-        do_sample=True,
-        repetition_penalty=repetition_penalty,
-        num_beams=num_beams,
-        top_k=top_k,
-        early_stopping=True,
-        pad_token_id=tokenizer.pad_token_id,
-        eos_token_id=tokenizer.eos_token_id,
-    )
+    # Since we can't track token generation directly, we'll create artificial steps
+    steps = 10  # Divide generation into 10 steps
+    for i in progress.tqdm(range(steps), desc="توليد النص (Generating text)"):
+        if i == 0:  # Only generate on the first step
+            output = model.generate(
+                **inputs, 
+                max_length=max_length, 
+                temperature=temperature, 
+                top_p=top_p, 
+                do_sample=True,
+                repetition_penalty=repetition_penalty,
+                num_beams=num_beams,
+                top_k=top_k,
+                early_stopping=True,
+                pad_token_id=tokenizer.pad_token_id,
+                eos_token_id=tokenizer.eos_token_id,
+            )
     
     # Decode output
+    progress(0.9, desc="معالجة النتائج (Processing results)")
     result = tokenizer.decode(output[0], skip_special_tokens=True)
     
     # Update stats
@@ -140,6 +148,8 @@ def generate_text(prompt, max_length=256, temperature=0.7, top_p=0.9, top_k=150,
         }
     )
     
+    progress(1.0, desc="اكتمل (Complete)")
+    
     return result, f"تم توليد {token_count} رمز في {generation_time:.2f} ثانية (Generated {token_count} tokens in {generation_time:.2f} seconds)"
 
 def save_feedback(input, output, params) -> None: