backup

app.py

@@ -13,6 +13,7 @@ import git
 from pathlib import Path
 from io import BytesIO
 import PIL
+import time  # Add this import at the top
 
 api = HfApi(token=os.environ["HF_TOKEN"])
 
@@ -122,9 +123,10 @@ scheduler = CommitScheduler(
 )
 
 
-def save_evaluation(post_id, model_a, model_b, verdict):
-    """Save evaluation results to CSV. Multiple evaluations per image/model are allowed."""
+def save_evaluation(post_id, model_a, model_b, verdict, username, start_time, end_time):
+    """Save evaluation results to CSV including timing and username information."""
     timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+    duration = end_time - start_time  # Calculate duration in seconds
 
     # Create data directory if it doesn't exist
     os.makedirs("data", exist_ok=True)
@@ -134,15 +136,39 @@ def save_evaluation(post_id, model_a, model_b, verdict):
     if not os.path.exists(filename):
         with open(filename, "w", newline="") as f:
             writer = csv.writer(f)
-            writer.writerow(["timestamp", "post_id", "model_a", "model_b", "verdict"])
+            writer.writerow(
+                [
+                    "timestamp",
+                    "post_id",
+                    "model_a",
+                    "model_b",
+                    "verdict",
+                    "username",
+                    "start_time",
+                    "end_time",
+                    "duration_seconds",
+                ]
+            )
 
-    # Append the new evaluation - multiple entries per image/model are allowed
+    # Append the new evaluation
     with open(filename, "a", newline="") as f:
         writer = csv.writer(f)
-        writer.writerow([timestamp, post_id, model_a, model_b, verdict])
+        writer.writerow(
+            [
+                timestamp,
+                post_id,
+                model_a,
+                model_b,
+                verdict,
+                username,
+                start_time,
+                end_time,
+                duration,
+            ]
+        )
 
     print(
-        f"Saved evaluation: {post_id} - Model A: {model_a} - Model B: {model_b} - Verdict: {verdict}"
+        f"Saved evaluation: {post_id} - Model A: {model_a} - Model B: {model_b} - Verdict: {verdict} - Duration: {duration:.2f}s"
     )
 
 
@@ -206,7 +232,7 @@ def get_random_sample():
 
 
 def evaluate(verdict, state):
-    """Handle evaluation button clicks"""
+    """Handle evaluation button clicks with timing"""
     if state is None:
         return (
             None,
@@ -221,23 +247,33 @@ def evaluate(verdict, state):
             False,
             False,
             None,
-            gr.update(variant="secondary"),  # a_better_btn
-            gr.update(variant="secondary"),  # b_better_btn
-            gr.update(variant="secondary"),  # neither_btn
-            gr.update(variant="secondary"),  # tie_btn
-            None,  # post_id
-            None,  # simplified_instruction
-            "",  # username_debug
+            gr.update(variant="secondary"),
+            gr.update(variant="secondary"),
+            gr.update(variant="secondary"),
+            gr.update(variant="secondary"),
+            None,
+            None,
+            "",
         )
 
-    # Save the evaluation
-    save_evaluation(state["post_id"], state["model_a"], state["model_b"], verdict)
+    # Record end time and save the evaluation
+    end_time = time.time()
+    save_evaluation(
+        state["post_id"],
+        state["model_a"],
+        state["model_b"],
+        verdict,
+        state["username"],
+        state["start_time"],
+        end_time,
+    )
 
     # Get next sample
     next_sample = get_random_sample()
-    # Preserve username in state
+    # Preserve username in state and set new start time
     next_state = next_sample.copy()
     next_state["username"] = state["username"]
+    next_state["start_time"] = time.time()  # Set start time for next evaluation
 
     # Reset button styles
     a_better_reset = gr.update(variant="secondary")
@@ -252,7 +288,7 @@ def evaluate(verdict, state):
         next_sample["instruction"],
         next_sample["simplified_instruction"],
         f"Model A: {next_sample['model_a']} | Model B: {next_sample['model_b']}",
-        next_state,  # Now includes username
+        next_state,  # Now includes username and start_time
         None,  # selected_verdict
         False,  # a_better_selected
         False,  # b_better_selected
@@ -297,9 +333,10 @@ def handle_username_submit(username, current_page):
 def initialize(username):
     """Initialize the interface with first sample"""
     sample = get_random_sample()
-    # Create state with username included
+    # Create state with username and start time included
     state = sample.copy()
     state["username"] = username
+    state["start_time"] = time.time()  # Record start time
 
     return (
         sample["source_image"],
@@ -308,7 +345,7 @@ def initialize(username):
         sample["instruction"],
         sample["simplified_instruction"],
         f"Model A: {sample['model_a']} | Model B: {sample['model_b']}",
-        state,  # Now includes username
+        state,  # Now includes username and start_time
         None,  # selected_verdict
         False,  # a_better_selected
         False,  # b_better_selected
@@ -316,7 +353,7 @@ def initialize(username):
         False,  # tie_selected
         sample["post_id"],
         sample["simplified_instruction"],
-        username or "",  # Use stored username or empty string
+        username or "",
     )