Update aitw_o1_dataset.json

aitw_o1_dataset.json

@@ -4,7 +4,7 @@
     "steps": [
       {
         "step_id": 0,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12172380859428428757_step_0.png",
+        "screenshot_path": "./aitw_o1_images/episode_12172380859428428757_step_0.png",
         "action_history": "",
         "ground_truth_format": "mcq",
         "questions": {
@@ -20,7 +20,7 @@
       },
       {
         "step_id": 1,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12172380859428428757_step_1.png",
+        "screenshot_path": "./aitw_o1_images/episode_12172380859428428757_step_1.png",
         "action_history": "Step 0: Tapped at pixel coordinates (x,y): [382, 838]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -36,7 +36,7 @@
       },
       {
         "step_id": 2,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12172380859428428757_step_2.png",
+        "screenshot_path": "./aitw_o1_images/episode_12172380859428428757_step_2.png",
         "action_history": "Step 0: Tapped at pixel coordinates (x,y): [382, 838]\nStep 1: Tapped at pixel coordinates (x,y): [224, 776]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -52,7 +52,7 @@
       },
       {
         "step_id": 3,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12172380859428428757_step_3.png",
+        "screenshot_path": "./aitw_o1_images/episode_12172380859428428757_step_3.png",
         "action_history": "Step 0: Tapped at pixel coordinates (x,y): [382, 838]\nStep 1: Tapped at pixel coordinates (x,y): [224, 776]\nStep 2: Tapped at pixel coordinates (x,y): [420, 775]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -68,7 +68,7 @@
       },
       {
         "step_id": 4,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12172380859428428757_step_4.png",
+        "screenshot_path": "./aitw_o1_images/episode_12172380859428428757_step_4.png",
         "action_history": "Step 0: Tapped at pixel coordinates (x,y): [382, 838]\nStep 1: Tapped at pixel coordinates (x,y): [224, 776]\nStep 2: Tapped at pixel coordinates (x,y): [420, 775]\nStep 3: Tapped at pixel coordinates (x,y): [468, 64]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -84,7 +84,7 @@
       },
       {
         "step_id": 5,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12172380859428428757_step_5.png",
+        "screenshot_path": "./aitw_o1_images/episode_12172380859428428757_step_5.png",
         "action_history": "Step 0: Tapped at pixel coordinates (x,y): [382, 838]\nStep 1: Tapped at pixel coordinates (x,y): [224, 776]\nStep 2: Tapped at pixel coordinates (x,y): [420, 775]\nStep 3: Tapped at pixel coordinates (x,y): [468, 64]\nStep 4: Tapped at pixel coordinates (x,y): [44, 78]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -100,7 +100,7 @@
       },
       {
         "step_id": 6,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12172380859428428757_step_6.png",
+        "screenshot_path": "./aitw_o1_images/episode_12172380859428428757_step_6.png",
         "action_history": "Step 0: Tapped at pixel coordinates (x,y): [382, 838]\nStep 1: Tapped at pixel coordinates (x,y): [224, 776]\nStep 2: Tapped at pixel coordinates (x,y): [420, 775]\nStep 3: Tapped at pixel coordinates (x,y): [468, 64]\nStep 4: Tapped at pixel coordinates (x,y): [44, 78]\nStep 5: Tapped at pixel coordinates (x,y): [467, 62]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -116,7 +116,7 @@
       },
       {
         "step_id": 7,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12172380859428428757_step_7.png",
+        "screenshot_path": "./aitw_o1_images/episode_12172380859428428757_step_7.png",
         "action_history": "Step 0: Tapped at pixel coordinates (x,y): [382, 838]\nStep 1: Tapped at pixel coordinates (x,y): [224, 776]\nStep 2: Tapped at pixel coordinates (x,y): [420, 775]\nStep 3: Tapped at pixel coordinates (x,y): [468, 64]\nStep 4: Tapped at pixel coordinates (x,y): [44, 78]\nStep 5: Tapped at pixel coordinates (x,y): [467, 62]\nStep 6: Tapped at pixel coordinates (x,y): [514, 64]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -132,7 +132,7 @@
       },
       {
         "step_id": 8,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12172380859428428757_step_8.png",
+        "screenshot_path": "./aitw_o1_images/episode_12172380859428428757_step_8.png",
         "action_history": "Step 0: Tapped at pixel coordinates (x,y): [382, 838]\nStep 1: Tapped at pixel coordinates (x,y): [224, 776]\nStep 2: Tapped at pixel coordinates (x,y): [420, 775]\nStep 3: Tapped at pixel coordinates (x,y): [468, 64]\nStep 4: Tapped at pixel coordinates (x,y): [44, 78]\nStep 5: Tapped at pixel coordinates (x,y): [467, 62]\nStep 6: Tapped at pixel coordinates (x,y): [514, 64]\nStep 7: Tapped at pixel coordinates (x,y): [370, 127]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -153,7 +153,7 @@
     "steps": [
       {
         "step_id": 0,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_9027843537045096280_step_0.png",
+        "screenshot_path": "./aitw_o1_images/episode_9027843537045096280_step_0.png",
         "action_history": "",
         "ground_truth_format": "mcq",
         "questions": {
@@ -169,7 +169,7 @@
       },
       {
         "step_id": 1,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_9027843537045096280_step_1.png",
+        "screenshot_path": "./aitw_o1_images/episode_9027843537045096280_step_1.png",
         "action_history": "Step 0: Pressed Button: Press Home",
         "ground_truth_format": "mcq",
         "questions": {
@@ -185,7 +185,7 @@
       },
       {
         "step_id": 2,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_9027843537045096280_step_2.png",
+        "screenshot_path": "./aitw_o1_images/episode_9027843537045096280_step_2.png",
         "action_history": "Step 0: Pressed Button: Press Home\nStep 1: Tapped at pixel coordinates (x,y): [365, 826]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -206,7 +206,7 @@
     "steps": [
       {
         "step_id": 0,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_2247237893045084693_step_0.png",
+        "screenshot_path": "./aitw_o1_images/episode_2247237893045084693_step_0.png",
         "action_history": "",
         "ground_truth_format": "mcq",
         "questions": {
@@ -222,7 +222,7 @@
       },
       {
         "step_id": 1,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_2247237893045084693_step_1.png",
+        "screenshot_path": "./aitw_o1_images/episode_2247237893045084693_step_1.png",
         "action_history": "Step 0: Swiped: Swipe Up",
         "ground_truth_format": "mcq",
         "questions": {
@@ -238,7 +238,7 @@
       },
       {
         "step_id": 2,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_2247237893045084693_step_2.png",
+        "screenshot_path": "./aitw_o1_images/episode_2247237893045084693_step_2.png",
         "action_history": "Step 0: Swiped: Swipe Up\nStep 1: Tapped at pixel coordinates (x,y): [379, 183]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -254,7 +254,7 @@
       },
       {
         "step_id": 3,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_2247237893045084693_step_3.png",
+        "screenshot_path": "./aitw_o1_images/episode_2247237893045084693_step_3.png",
         "action_history": "Step 0: Swiped: Swipe Up\nStep 1: Tapped at pixel coordinates (x,y): [379, 183]\nStep 2: Tapped at pixel coordinates (x,y): [499, 976]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -270,7 +270,7 @@
       },
       {
         "step_id": 4,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_2247237893045084693_step_4.png",
+        "screenshot_path": "./aitw_o1_images/episode_2247237893045084693_step_4.png",
         "action_history": "Step 0: Swiped: Swipe Up\nStep 1: Tapped at pixel coordinates (x,y): [379, 183]\nStep 2: Tapped at pixel coordinates (x,y): [499, 976]\nStep 3: Tapped at pixel coordinates (x,y): [302, 950]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -291,7 +291,7 @@
     "steps": [
       {
         "step_id": 0,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12224608131504749719_step_0.png",
+        "screenshot_path": "./aitw_o1_images/episode_12224608131504749719_step_0.png",
         "action_history": "",
         "ground_truth_format": "mcq",
         "questions": {
@@ -307,7 +307,7 @@
       },
       {
         "step_id": 1,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12224608131504749719_step_1.png",
+        "screenshot_path": "./aitw_o1_images/episode_12224608131504749719_step_1.png",
         "action_history": "Step 0: Swiped: Swipe Up",
         "ground_truth_format": "mcq",
         "questions": {
@@ -323,7 +323,7 @@
       },
       {
         "step_id": 2,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12224608131504749719_step_2.png",
+        "screenshot_path": "./aitw_o1_images/episode_12224608131504749719_step_2.png",
         "action_history": "Step 0: Swiped: Swipe Up\nStep 1: Pressed Button: Press Home",
         "ground_truth_format": "mcq",
         "questions": {
@@ -339,7 +339,7 @@
       },
       {
         "step_id": 3,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12224608131504749719_step_3.png",
+        "screenshot_path": "./aitw_o1_images/episode_12224608131504749719_step_3.png",
         "action_history": "Step 0: Swiped: Swipe Up\nStep 1: Pressed Button: Press Home\nStep 2: Tapped at pixel coordinates (x,y): [289, 950]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -355,7 +355,7 @@
       },
       {
         "step_id": 4,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12224608131504749719_step_4.png",
+        "screenshot_path": "./aitw_o1_images/episode_12224608131504749719_step_4.png",
         "action_history": "Step 0: Swiped: Swipe Up\nStep 1: Pressed Button: Press Home\nStep 2: Tapped at pixel coordinates (x,y): [289, 950]\nStep 3: Typed: 'How much does a 2x4x8 board cost at Lowes?'",
         "ground_truth_format": "mcq",
         "questions": {
@@ -371,7 +371,7 @@
       },
       {
         "step_id": 5,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_12224608131504749719_step_5.png",
+        "screenshot_path": "./aitw_o1_images/episode_12224608131504749719_step_5.png",
         "action_history": "Step 0: Swiped: Swipe Up\nStep 1: Pressed Button: Press Home\nStep 2: Tapped at pixel coordinates (x,y): [289, 950]\nStep 3: Typed: 'How much does a 2x4x8 board cost at Lowes?'\nStep 4: Tapped at pixel coordinates (x,y): [286, 129]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -392,7 +392,7 @@
     "steps": [
       {
         "step_id": 0,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_1307957808436696980_step_0.png",
+        "screenshot_path": "./aitw_o1_images/episode_1307957808436696980_step_0.png",
         "action_history": "",
         "ground_truth_format": "mcq",
         "questions": {
@@ -408,7 +408,7 @@
       },
       {
         "step_id": 1,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_1307957808436696980_step_1.png",
+        "screenshot_path": "./aitw_o1_images/episode_1307957808436696980_step_1.png",
         "action_history": "Step 0: Pressed Button: Press Home",
         "ground_truth_format": "mcq",
         "questions": {
@@ -424,7 +424,7 @@
       },
       {
         "step_id": 2,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_1307957808436696980_step_2.png",
+        "screenshot_path": "./aitw_o1_images/episode_1307957808436696980_step_2.png",
         "action_history": "Step 0: Pressed Button: Press Home\nStep 1: Swiped: Swipe Up",
         "ground_truth_format": "mcq",
         "questions": {
@@ -440,7 +440,7 @@
       },
       {
         "step_id": 3,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_1307957808436696980_step_3.png",
+        "screenshot_path": "./aitw_o1_images/episode_1307957808436696980_step_3.png",
         "action_history": "Step 0: Pressed Button: Press Home\nStep 1: Swiped: Swipe Up\nStep 2: Pressed Button: Press Home",
         "ground_truth_format": "mcq",
         "questions": {
@@ -456,7 +456,7 @@
       },
       {
         "step_id": 4,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_1307957808436696980_step_4.png",
+        "screenshot_path": "./aitw_o1_images/episode_1307957808436696980_step_4.png",
         "action_history": "Step 0: Pressed Button: Press Home\nStep 1: Swiped: Swipe Up\nStep 2: Pressed Button: Press Home\nStep 3: Tapped at pixel coordinates (x,y): [320, 974]",
         "ground_truth_format": "mcq",
         "questions": {
@@ -472,7 +472,7 @@
       },
       {
         "step_id": 5,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_1307957808436696980_step_5.png",
+        "screenshot_path": "./aitw_o1_images/episode_1307957808436696980_step_5.png",
         "action_history": "Step 0: Pressed Button: Press Home\nStep 1: Swiped: Swipe Up\nStep 2: Pressed Button: Press Home\nStep 3: Tapped at pixel coordinates (x,y): [320, 974]\nStep 4: Typed: 'Google the capital of Mexico'",
         "ground_truth_format": "mcq",
         "questions": {
@@ -488,7 +488,7 @@
       },
       {
         "step_id": 6,
-        "screenshot_path": "/n/fs/vision-mix/ag9604/aitw_o1_images/episode_1307957808436696980_step_6.png",
+        "screenshot_path": "./aitw_o1_images/episode_1307957808436696980_step_6.png",
         "action_history": "Step 0: Pressed Button: Press Home\nStep 1: Swiped: Swipe Up\nStep 2: Pressed Button: Press Home\nStep 3: Tapped at pixel coordinates (x,y): [320, 974]\nStep 4: Typed: 'Google the capital of Mexico'\nStep 5: Tapped at pixel coordinates (x,y): [185, 114]",
         "ground_truth_format": "mcq",
         "questions": {