id stringlengths 25 29 | image stringlengths 16 16 | source_component stringclasses 1
value | source_record_id stringlengths 12 12 | source_row int64 0 56.7k | source_turn int64 0 5 | conversations listlengths 2 2 | selected_row int64 0 151k | image_archive stringclasses 21
values | pixel_sha256 stringlengths 64 64 |
|---|---|---|---|---|---|---|---|---|---|
conversation_58k.json:0:0 | 000000033471.jpg | conversation_58k.json | 000000033471 | 0 | 0 | [
{
"from": "human",
"value": "<image>\nWhat are the colors of the bus in the image?"
},
{
"from": "gpt",
"value": "The bus in the image is white and red."
}
] | 0 | images/sft/0001.zip | 89ff3be974ef2e256be305cd2b27e6fa144810148ca74e17b90e02ecdb243004 |
conversation_58k.json:0:1 | 000000033471.jpg | conversation_58k.json | 000000033471 | 0 | 1 | [
{
"from": "human",
"value": "<image>\nWhat feature can be seen on the back of the bus?"
},
{
"from": "gpt",
"value": "The back of the bus features an advertisement."
}
] | 1 | images/sft/0001.zip | 89ff3be974ef2e256be305cd2b27e6fa144810148ca74e17b90e02ecdb243004 |
conversation_58k.json:0:2 | 000000033471.jpg | conversation_58k.json | 000000033471 | 0 | 2 | [
{
"from": "human",
"value": "<image>\nIs the bus driving down the street or pulled off to the side?"
},
{
"from": "gpt",
"value": "The bus is driving down the street, which is crowded with people and other vehicles."
}
] | 2 | images/sft/0001.zip | 89ff3be974ef2e256be305cd2b27e6fa144810148ca74e17b90e02ecdb243004 |
conversation_58k.json:1:0 | 000000052846.jpg | conversation_58k.json | 000000052846 | 1 | 0 | [
{
"from": "human",
"value": "<image>\nWhere is the cat positioned in the image?"
},
{
"from": "gpt",
"value": "The cat is positioned on top of the back of the couch in the living room."
}
] | 3 | images/sft/0001.zip | 2dcc7f2baf994cc8a3776d3b03a6e7eb7bb38b730d3807d62b03859245128863 |
conversation_58k.json:1:1 | 000000052846.jpg | conversation_58k.json | 000000052846 | 1 | 1 | [
{
"from": "human",
"value": "<image>\nWhat is the cat doing in the image?"
},
{
"from": "gpt",
"value": "The cat is coming out from some curtains onto the couch and is sitting or standing on top of it."
}
] | 4 | images/sft/0001.zip | 2dcc7f2baf994cc8a3776d3b03a6e7eb7bb38b730d3807d62b03859245128863 |
conversation_58k.json:1:2 | 000000052846.jpg | conversation_58k.json | 000000052846 | 1 | 2 | [
{
"from": "human",
"value": "<image>\nWhat is the color of the curtains near the cat?"
},
{
"from": "gpt",
"value": "The color of the curtains near the cat is red."
}
] | 5 | images/sft/0001.zip | 2dcc7f2baf994cc8a3776d3b03a6e7eb7bb38b730d3807d62b03859245128863 |
conversation_58k.json:1:4 | 000000052846.jpg | conversation_58k.json | 000000052846 | 1 | 4 | [
{
"from": "human",
"value": "<image>\nWhat kind of room does the couch appear to be in?"
},
{
"from": "gpt",
"value": "The couch appears to be in a living room setting."
}
] | 6 | images/sft/0001.zip | 2dcc7f2baf994cc8a3776d3b03a6e7eb7bb38b730d3807d62b03859245128863 |
conversation_58k.json:2:0 | 000000334872.jpg | conversation_58k.json | 000000334872 | 2 | 0 | [
{
"from": "human",
"value": "<image>\nAre the people in the image skiing downhill or cross-country skiing?"
},
{
"from": "gpt",
"value": "The people in the image are cross-country skiing in the woods, as they are skiing on a trail rather than a steep slope."
}
] | 7 | images/sft/0011.zip | 4a4f1731566a06d39652982029725562a46421ad8644bc34651d9e47c2f6db80 |
conversation_58k.json:2:1 | 000000334872.jpg | conversation_58k.json | 000000334872 | 2 | 1 | [
{
"from": "human",
"value": "<image>\nHow many people are in the image?"
},
{
"from": "gpt",
"value": "There are two people in the image, both on skis in the snow."
}
] | 8 | images/sft/0011.zip | 4a4f1731566a06d39652982029725562a46421ad8644bc34651d9e47c2f6db80 |
conversation_58k.json:3:0 | 000000319154.jpg | conversation_58k.json | 000000319154 | 3 | 0 | [
{
"from": "human",
"value": "<image>\nWhat color is the dog in the image?"
},
{
"from": "gpt",
"value": "The dog in the image is brown."
}
] | 9 | images/sft/0011.zip | 6dc1656e8c22c5d9b7f5e4325848d0dc1302795f02955058934c87291cf98a08 |
conversation_58k.json:3:1 | 000000319154.jpg | conversation_58k.json | 000000319154 | 3 | 1 | [
{
"from": "human",
"value": "<image>\nWhat type of collar does the dog have?"
},
{
"from": "gpt",
"value": "The dog has a red collar."
}
] | 10 | images/sft/0011.zip | 6dc1656e8c22c5d9b7f5e4325848d0dc1302795f02955058934c87291cf98a08 |
conversation_58k.json:3:2 | 000000319154.jpg | conversation_58k.json | 000000319154 | 3 | 2 | [
{
"from": "human",
"value": "<image>\nWhat is the dog's position behind the window?"
},
{
"from": "gpt",
"value": "The dog is sitting behind the window."
}
] | 11 | images/sft/0011.zip | 6dc1656e8c22c5d9b7f5e4325848d0dc1302795f02955058934c87291cf98a08 |
conversation_58k.json:4:0 | 000000398214.jpg | conversation_58k.json | 000000398214 | 4 | 0 | [
{
"from": "human",
"value": "<image>\nWhat type of sign and traffic device can be seen in the image?"
},
{
"from": "gpt",
"value": "In the image, there is a street sign and a traffic light above a city road."
}
] | 12 | images/sft/0013.zip | 02143b1a20eea1adc0a678060e8912e2266ed54fc6edd7bb187e5499b117ffad |
conversation_58k.json:4:1 | 000000398214.jpg | conversation_58k.json | 000000398214 | 4 | 1 | [
{
"from": "human",
"value": "<image>\nHow many traffic lights are visible in the image?"
},
{
"from": "gpt",
"value": "There are a total of six traffic lights visible in the image."
}
] | 13 | images/sft/0013.zip | 02143b1a20eea1adc0a678060e8912e2266ed54fc6edd7bb187e5499b117ffad |
conversation_58k.json:4:2 | 000000398214.jpg | conversation_58k.json | 000000398214 | 4 | 2 | [
{
"from": "human",
"value": "<image>\nWhat is the current color of the traffic lights in the image?"
},
{
"from": "gpt",
"value": "All the traffic lights in the image are showing the color red at the same time."
}
] | 14 | images/sft/0013.zip | 02143b1a20eea1adc0a678060e8912e2266ed54fc6edd7bb187e5499b117ffad |
conversation_58k.json:4:3 | 000000398214.jpg | conversation_58k.json | 000000398214 | 4 | 3 | [
{
"from": "human",
"value": "<image>\nWhat kind of setting is the image taken in?"
},
{
"from": "gpt",
"value": "The image is taken in a big city setting with a city street, an overpass, and buildings in the distance."
}
] | 15 | images/sft/0013.zip | 02143b1a20eea1adc0a678060e8912e2266ed54fc6edd7bb187e5499b117ffad |
conversation_58k.json:4:4 | 000000398214.jpg | conversation_58k.json | 000000398214 | 4 | 4 | [
{
"from": "human",
"value": "<image>\nAre there any cars visible in the image?"
},
{
"from": "gpt",
"value": "Yes, there are cars visible on the city street in the image."
}
] | 16 | images/sft/0013.zip | 02143b1a20eea1adc0a678060e8912e2266ed54fc6edd7bb187e5499b117ffad |
conversation_58k.json:5:0 | 000000520873.jpg | conversation_58k.json | 000000520873 | 5 | 0 | [
{
"from": "human",
"value": "<image>\nWhat is the girl eating in the image?"
},
{
"from": "gpt",
"value": "The girl in the image is eating a dessert, which appears to be a graham cracker treat or a cookie sandwich."
}
] | 17 | images/sft/0018.zip | b110473fa7acc4cd596079f330c642cde258694293a0ad56d0ba08fa6dbb31ea |
conversation_58k.json:5:3 | 000000520873.jpg | conversation_58k.json | 000000520873 | 5 | 3 | [
{
"from": "human",
"value": "<image>\nIs the girl looking at the camera or focusing on her dessert?"
},
{
"from": "gpt",
"value": "The girl is looking up at the camera while taking a bite of her dessert."
}
] | 18 | images/sft/0018.zip | b110473fa7acc4cd596079f330c642cde258694293a0ad56d0ba08fa6dbb31ea |
conversation_58k.json:5:4 | 000000520873.jpg | conversation_58k.json | 000000520873 | 5 | 4 | [
{
"from": "human",
"value": "<image>\nWhere is the girl eating her dessert?"
},
{
"from": "gpt",
"value": "The girl is eating her dessert at the table."
}
] | 19 | images/sft/0018.zip | b110473fa7acc4cd596079f330c642cde258694293a0ad56d0ba08fa6dbb31ea |
conversation_58k.json:6:0 | 000000575173.jpg | conversation_58k.json | 000000575173 | 6 | 0 | [
{
"from": "human",
"value": "<image>\nWhat type of aircraft is shown in the image?"
},
{
"from": "gpt",
"value": "The image shows a large passenger jet belonging to China Airlines."
}
] | 20 | images/sft/0019.zip | 2b1d6dfcc2b2dd8f6c3d99391c3411d6e2ed06d63e66c02e24f44ed1760301b9 |
conversation_58k.json:6:1 | 000000575173.jpg | conversation_58k.json | 000000575173 | 6 | 1 | [
{
"from": "human",
"value": "<image>\nIs the airplane in the air or on the ground?"
},
{
"from": "gpt",
"value": "The airplane is on the ground, as it is seen taxiing down the runway next to a body of water."
}
] | 21 | images/sft/0019.zip | 2b1d6dfcc2b2dd8f6c3d99391c3411d6e2ed06d63e66c02e24f44ed1760301b9 |
conversation_58k.json:6:2 | 000000575173.jpg | conversation_58k.json | 000000575173 | 6 | 2 | [
{
"from": "human",
"value": "<image>\nWhat is the location of the harbor or body of water in relation to the airplane?"
},
{
"from": "gpt",
"value": "The body of water is located right next to the runway where the airplane is taxiing."
}
] | 22 | images/sft/0019.zip | 2b1d6dfcc2b2dd8f6c3d99391c3411d6e2ed06d63e66c02e24f44ed1760301b9 |
conversation_58k.json:6:3 | 000000575173.jpg | conversation_58k.json | 000000575173 | 6 | 3 | [
{
"from": "human",
"value": "<image>\nWhat type of landscape is visible around the airplane?"
},
{
"from": "gpt",
"value": "The landscape visible around the airplane consists of a runway and a large body of water or lake located next to it."
}
] | 23 | images/sft/0019.zip | 2b1d6dfcc2b2dd8f6c3d99391c3411d6e2ed06d63e66c02e24f44ed1760301b9 |
conversation_58k.json:7:0 | 000000087286.jpg | conversation_58k.json | 000000087286 | 7 | 0 | [
{
"from": "human",
"value": "<image>\nWhat is hanging from the traffic light pole?"
},
{
"from": "gpt",
"value": "A green street sign is hanging from the traffic light pole."
}
] | 24 | images/sft/0003.zip | e66d2bbdc5c099aa9d9c438ec313c84f035af755cddb3f62e8b9c2fd43d27400 |
conversation_58k.json:7:1 | 000000087286.jpg | conversation_58k.json | 000000087286 | 7 | 1 | [
{
"from": "human",
"value": "<image>\nHow many stoplights are on the light post?"
},
{
"from": "gpt",
"value": "There are two stoplights on the light post."
}
] | 25 | images/sft/0003.zip | e66d2bbdc5c099aa9d9c438ec313c84f035af755cddb3f62e8b9c2fd43d27400 |
conversation_58k.json:7:2 | 000000087286.jpg | conversation_58k.json | 000000087286 | 7 | 2 | [
{
"from": "human",
"value": "<image>\nWhat is the background of the image where the street sign is situated?"
},
{
"from": "gpt",
"value": "The street sign on the pole is set against a backdrop of a daytime sky."
}
] | 26 | images/sft/0003.zip | e66d2bbdc5c099aa9d9c438ec313c84f035af755cddb3f62e8b9c2fd43d27400 |
conversation_58k.json:8:0 | 000000032286.jpg | conversation_58k.json | 000000032286 | 8 | 0 | [
{
"from": "human",
"value": "<image>\nHow many chocolate-covered doughnuts are there in the image?"
},
{
"from": "gpt",
"value": "There are seven chocolate-covered doughnuts in the image."
}
] | 27 | images/sft/0001.zip | 77584a17770865003efd0a346722778f917441bbf36d8ad9a57f86f53c5ea352 |
conversation_58k.json:8:1 | 000000032286.jpg | conversation_58k.json | 000000032286 | 8 | 1 | [
{
"from": "human",
"value": "<image>\nWhat is the color of the frosting on the doughnuts?"
},
{
"from": "gpt",
"value": "The frosting on the doughnuts is chocolate-colored."
}
] | 28 | images/sft/0001.zip | 77584a17770865003efd0a346722778f917441bbf36d8ad9a57f86f53c5ea352 |
conversation_58k.json:8:2 | 000000032286.jpg | conversation_58k.json | 000000032286 | 8 | 2 | [
{
"from": "human",
"value": "<image>\nWhere are the doughnuts placed?"
},
{
"from": "gpt",
"value": "The doughnuts are placed in a white box, which is sitting on top of a counter."
}
] | 29 | images/sft/0001.zip | 77584a17770865003efd0a346722778f917441bbf36d8ad9a57f86f53c5ea352 |
conversation_58k.json:8:3 | 000000032286.jpg | conversation_58k.json | 000000032286 | 8 | 3 | [
{
"from": "human",
"value": "<image>\nWhat is the arrangement of the doughnuts?"
},
{
"from": "gpt",
"value": "The doughnuts are sitting together on a tray inside the white box."
}
] | 30 | images/sft/0001.zip | 77584a17770865003efd0a346722778f917441bbf36d8ad9a57f86f53c5ea352 |
conversation_58k.json:9:0 | 000000175217.jpg | conversation_58k.json | 000000175217 | 9 | 0 | [
{
"from": "human",
"value": "<image>\nWhat is the color of the sink and vanity in the bathroom?"
},
{
"from": "gpt",
"value": "The sink and vanity in the bathroom are beige in color."
}
] | 31 | images/sft/0006.zip | c8ca0b8635f79745b03973fdbd97e5498e8474a9bce0d8490422e5d047391d7e |
conversation_58k.json:9:1 | 000000175217.jpg | conversation_58k.json | 000000175217 | 9 | 1 | [
{
"from": "human",
"value": "<image>\nWhat is the color of the toilet in the bathroom?"
},
{
"from": "gpt",
"value": "The toilet in the bathroom is white."
}
] | 32 | images/sft/0006.zip | c8ca0b8635f79745b03973fdbd97e5498e8474a9bce0d8490422e5d047391d7e |
conversation_58k.json:9:2 | 000000175217.jpg | conversation_58k.json | 000000175217 | 9 | 2 | [
{
"from": "human",
"value": "<image>\nIs there a painting hanging on the wall in the bathroom?"
},
{
"from": "gpt",
"value": "Yes, there is a painting hanging on the wall in the bathroom."
}
] | 33 | images/sft/0006.zip | c8ca0b8635f79745b03973fdbd97e5498e8474a9bce0d8490422e5d047391d7e |
conversation_58k.json:9:3 | 000000175217.jpg | conversation_58k.json | 000000175217 | 9 | 3 | [
{
"from": "human",
"value": "<image>\nWhat type of fixtures or counters are used in the bathroom?"
},
{
"from": "gpt",
"value": "Marble fixtures or counters are used in the bathroom."
}
] | 34 | images/sft/0006.zip | c8ca0b8635f79745b03973fdbd97e5498e8474a9bce0d8490422e5d047391d7e |
conversation_58k.json:10:0 | 000000230238.jpg | conversation_58k.json | 000000230238 | 10 | 0 | [
{
"from": "human",
"value": "<image>\nWhat is the main color of the vase in the image?"
},
{
"from": "gpt",
"value": "The main color of the vase in the image is blue."
}
] | 35 | images/sft/0007.zip | 75865383ecab5a7bb76d46c9ffacfd5b202e873e43a688f2c6e04f642e282729 |
conversation_58k.json:10:1 | 000000230238.jpg | conversation_58k.json | 000000230238 | 10 | 1 | [
{
"from": "human",
"value": "<image>\nWhere is the vase located?"
},
{
"from": "gpt",
"value": "The vase is located in a museum, placed on a pedestal on display."
}
] | 36 | images/sft/0007.zip | 75865383ecab5a7bb76d46c9ffacfd5b202e873e43a688f2c6e04f642e282729 |
conversation_58k.json:10:2 | 000000230238.jpg | conversation_58k.json | 000000230238 | 10 | 2 | [
{
"from": "human",
"value": "<image>\nWhat kind of design is on the vase?"
},
{
"from": "gpt",
"value": "The vase features blue flowery drawings, which add to its decorative and artistic appeal."
}
] | 37 | images/sft/0007.zip | 75865383ecab5a7bb76d46c9ffacfd5b202e873e43a688f2c6e04f642e282729 |
conversation_58k.json:10:3 | 000000230238.jpg | conversation_58k.json | 000000230238 | 10 | 3 | [
{
"from": "human",
"value": "<image>\nIs there a person visible in the background of the image?"
},
{
"from": "gpt",
"value": "Yes, there is a person visible in the background of the image near the large vase."
}
] | 38 | images/sft/0007.zip | 75865383ecab5a7bb76d46c9ffacfd5b202e873e43a688f2c6e04f642e282729 |
conversation_58k.json:11:0 | 000000392166.jpg | conversation_58k.json | 000000392166 | 11 | 0 | [
{
"from": "human",
"value": "<image>\nIs the man on the skateboard holding a dog?"
},
{
"from": "gpt",
"value": "Yes, the man on the skateboard is holding a dog in his arm."
}
] | 39 | images/sft/0013.zip | 0538b7acee272243d67296ca6f4bd6102e238e3f320151b5be838f00a2ff5c8f |
conversation_58k.json:11:1 | 000000392166.jpg | conversation_58k.json | 000000392166 | 11 | 1 | [
{
"from": "human",
"value": "<image>\nWhat color is the dog being held by the man?"
},
{
"from": "gpt",
"value": "The dog being held by the man is brown."
}
] | 40 | images/sft/0013.zip | 0538b7acee272243d67296ca6f4bd6102e238e3f320151b5be838f00a2ff5c8f |
conversation_58k.json:11:2 | 000000392166.jpg | conversation_58k.json | 000000392166 | 11 | 2 | [
{
"from": "human",
"value": "<image>\nIs the man riding the skateboard with both hands out or with one hand holding the dog?"
},
{
"from": "gpt",
"value": "The man is riding the skateboard with one hand holding the dog."
}
] | 41 | images/sft/0013.zip | 0538b7acee272243d67296ca6f4bd6102e238e3f320151b5be838f00a2ff5c8f |
conversation_58k.json:12:0 | 000000521031.jpg | conversation_58k.json | 000000521031 | 12 | 0 | [
{
"from": "human",
"value": "<image>\nWhat objects are the main focus of the image?"
},
{
"from": "gpt",
"value": "The main focus of the image is a collection of stuffed toy bears."
}
] | 42 | images/sft/0018.zip | eeaba68b454323a3c7ea1a6d78ce7f7bb7360ec96a94fe22fd8447840d0918a2 |
conversation_58k.json:12:1 | 000000521031.jpg | conversation_58k.json | 000000521031 | 12 | 1 | [
{
"from": "human",
"value": "<image>\nWhere are the stuffed toy bears placed?"
},
{
"from": "gpt",
"value": "The stuffed toy bears are placed on a table."
}
] | 43 | images/sft/0018.zip | eeaba68b454323a3c7ea1a6d78ce7f7bb7360ec96a94fe22fd8447840d0918a2 |
conversation_58k.json:12:4 | 000000521031.jpg | conversation_58k.json | 000000521031 | 12 | 4 | [
{
"from": "human",
"value": "<image>\nWhat is the color of the teddy bears in the image?"
},
{
"from": "gpt",
"value": "The color of the teddy bears in the image is brown."
}
] | 44 | images/sft/0018.zip | eeaba68b454323a3c7ea1a6d78ce7f7bb7360ec96a94fe22fd8447840d0918a2 |
conversation_58k.json:13:0 | 000000199318.jpg | conversation_58k.json | 000000199318 | 13 | 0 | [
{
"from": "human",
"value": "<image>\nWhat is the prominent feature of the large building in the image?"
},
{
"from": "gpt",
"value": "The prominent feature of the large building in the image is the tall clock tower."
}
] | 45 | images/sft/0006.zip | 83a176b51671ec0a3dcfef8bba871c06d418b73d53d1dd7561c12d04ce5643f3 |
conversation_58k.json:13:1 | 000000199318.jpg | conversation_58k.json | 000000199318 | 13 | 1 | [
{
"from": "human",
"value": "<image>\nWhat is the weather condition depicted in the image?"
},
{
"from": "gpt",
"value": "The image shows a cloudy day."
}
] | 46 | images/sft/0006.zip | 83a176b51671ec0a3dcfef8bba871c06d418b73d53d1dd7561c12d04ce5643f3 |
conversation_58k.json:13:2 | 000000199318.jpg | conversation_58k.json | 000000199318 | 13 | 2 | [
{
"from": "human",
"value": "<image>\nIs there a road passing under the building with the clock tower?"
},
{
"from": "gpt",
"value": "Yes, there is a road passing under a part of the building with the clock tower."
}
] | 47 | images/sft/0006.zip | 83a176b51671ec0a3dcfef8bba871c06d418b73d53d1dd7561c12d04ce5643f3 |
conversation_58k.json:14:0 | 000000483357.jpg | conversation_58k.json | 000000483357 | 14 | 0 | [
{
"from": "human",
"value": "<image>\nWhat color is the bathroom in the image?"
},
{
"from": "gpt",
"value": "The bathroom in the image is white, with white fixtures."
}
] | 48 | images/sft/0016.zip | 62342397574cf6e64a486f2cc8636c2c92215096612caa7b4863f27ff6ab2d36 |
conversation_58k.json:14:1 | 000000483357.jpg | conversation_58k.json | 000000483357 | 14 | 1 | [
{
"from": "human",
"value": "<image>\nWhat are the main elements present in the bathroom?"
},
{
"from": "gpt",
"value": "The main elements present in the bathroom are the bathtub, toilet, and sink."
}
] | 49 | images/sft/0016.zip | 62342397574cf6e64a486f2cc8636c2c92215096612caa7b4863f27ff6ab2d36 |
conversation_58k.json:14:3 | 000000483357.jpg | conversation_58k.json | 000000483357 | 14 | 3 | [
{
"from": "human",
"value": "<image>\nWhat material is the floor made of in the bathroom?"
},
{
"from": "gpt",
"value": "The floor in the bathroom is made of tiles."
}
] | 50 | images/sft/0016.zip | 62342397574cf6e64a486f2cc8636c2c92215096612caa7b4863f27ff6ab2d36 |
conversation_58k.json:15:0 | 000000197680.jpg | conversation_58k.json | 000000197680 | 15 | 0 | [
{
"from": "human",
"value": "<image>\nHow many glass walls are there in the bathroom beside the large tub?"
},
{
"from": "gpt",
"value": "There are two glass walls beside the large tub in the bathroom."
}
] | 51 | images/sft/0006.zip | 669a8f996b9dd7f7d99e64729cf605af646dd48d097b5f62f40c6396c8e4fd62 |
conversation_58k.json:15:1 | 000000197680.jpg | conversation_58k.json | 000000197680 | 15 | 1 | [
{
"from": "human",
"value": "<image>\nWhat can be found above the bathtub in the bathroom?"
},
{
"from": "gpt",
"value": "A painting can be found above the bathtub in the bathroom."
}
] | 52 | images/sft/0006.zip | 669a8f996b9dd7f7d99e64729cf605af646dd48d097b5f62f40c6396c8e4fd62 |
conversation_58k.json:15:2 | 000000197680.jpg | conversation_58k.json | 000000197680 | 15 | 2 | [
{
"from": "human",
"value": "<image>\nWhat kind of appearance does the bathroom have in the image?"
},
{
"from": "gpt",
"value": "The bathroom has a very elegant appearance, with features such as a large tub, tiled floors, and the tiles reflecting off the tub, contributing to its sophisticated l... | 53 | images/sft/0006.zip | 669a8f996b9dd7f7d99e64729cf605af646dd48d097b5f62f40c6396c8e4fd62 |
conversation_58k.json:15:3 | 000000197680.jpg | conversation_58k.json | 000000197680 | 15 | 3 | [
{
"from": "human",
"value": "<image>\nWhat is the material of the large tub in the bathroom?"
},
{
"from": "gpt",
"value": "The large tub in the bathroom is made of marble."
}
] | 54 | images/sft/0006.zip | 669a8f996b9dd7f7d99e64729cf605af646dd48d097b5f62f40c6396c8e4fd62 |
conversation_58k.json:15:4 | 000000197680.jpg | conversation_58k.json | 000000197680 | 15 | 4 | [
{
"from": "human",
"value": "<image>\nWhat type of flooring does the bathroom have?"
},
{
"from": "gpt",
"value": "The bathroom has tiled flooring."
}
] | 55 | images/sft/0006.zip | 669a8f996b9dd7f7d99e64729cf605af646dd48d097b5f62f40c6396c8e4fd62 |
conversation_58k.json:16:0 | 000000121795.jpg | conversation_58k.json | 000000121795 | 16 | 0 | [
{
"from": "human",
"value": "<image>\nIs there a person standing next to water in the image?"
},
{
"from": "gpt",
"value": "Yes, there is a person standing next to some blue water in the image."
}
] | 56 | images/sft/0004.zip | f1be5ef1450db3fb98402e0a39bcbb35195ad417443d5a5b857f99a0ba94d9cd |
conversation_58k.json:16:1 | 000000121795.jpg | conversation_58k.json | 000000121795 | 16 | 1 | [
{
"from": "human",
"value": "<image>\nWhat is the appearance of the sky in the image?"
},
{
"from": "gpt",
"value": "The sky in the image appears to be cloudy."
}
] | 57 | images/sft/0004.zip | f1be5ef1450db3fb98402e0a39bcbb35195ad417443d5a5b857f99a0ba94d9cd |
conversation_58k.json:16:2 | 000000121795.jpg | conversation_58k.json | 000000121795 | 16 | 2 | [
{
"from": "human",
"value": "<image>\nIs the person's silhouette visible in the image?"
},
{
"from": "gpt",
"value": "Yes, the silhouette of the adult is visible in the image, as it is lit by the setting sun over the waterway."
}
] | 58 | images/sft/0004.zip | f1be5ef1450db3fb98402e0a39bcbb35195ad417443d5a5b857f99a0ba94d9cd |
conversation_58k.json:17:0 | 000000341431.jpg | conversation_58k.json | 000000341431 | 17 | 0 | [
{
"from": "human",
"value": "<image>\nWhat time of day does the image capture?"
},
{
"from": "gpt",
"value": "The image captures a night scene."
}
] | 59 | images/sft/0011.zip | be0e4c49261c0cb25e9021b19e5e69384781f107b4b7cb3d5914ff554a77d564 |
conversation_58k.json:17:2 | 000000341431.jpg | conversation_58k.json | 000000341431 | 17 | 2 | [
{
"from": "human",
"value": "<image>\nIs there any notable architecture in the image?"
},
{
"from": "gpt",
"value": "Yes, there is a notable piece of architecture in the image: a cathedral can be seen across the street, adding to the overall ambiance of the scene."
}
] | 60 | images/sft/0011.zip | be0e4c49261c0cb25e9021b19e5e69384781f107b4b7cb3d5914ff554a77d564 |
conversation_58k.json:17:3 | 000000341431.jpg | conversation_58k.json | 000000341431 | 17 | 3 | [
{
"from": "human",
"value": "<image>\nWhat is located in the middle of the street?"
},
{
"from": "gpt",
"value": "In the middle of the street, there is a decorative pillar that serves as a divider for the road."
}
] | 61 | images/sft/0011.zip | be0e4c49261c0cb25e9021b19e5e69384781f107b4b7cb3d5914ff554a77d564 |
conversation_58k.json:18:0 | 000000300011.jpg | conversation_58k.json | 000000300011 | 18 | 0 | [
{
"from": "human",
"value": "<image>\nWhat type of vehicle is parked in front of the Kingfisher airplane?"
},
{
"from": "gpt",
"value": "A food service truck is parked in front of the Kingfisher airplane."
}
] | 62 | images/sft/0010.zip | a96630dee2b949d81ba77f9548bb9a3d948cf234a7bab8e01d76ac370de6a1f6 |
conversation_58k.json:18:1 | 000000300011.jpg | conversation_58k.json | 000000300011 | 18 | 1 | [
{
"from": "human",
"value": "<image>\nWhat is happening with the airplane in the image?"
},
{
"from": "gpt",
"value": "The airplane is on the tarmac at the airport, and it's being resupplied with food by the food service truck."
}
] | 63 | images/sft/0010.zip | a96630dee2b949d81ba77f9548bb9a3d948cf234a7bab8e01d76ac370de6a1f6 |
conversation_58k.json:18:2 | 000000300011.jpg | conversation_58k.json | 000000300011 | 18 | 2 | [
{
"from": "human",
"value": "<image>\nIs the airplane a passenger plane or a cargo plane?"
},
{
"from": "gpt",
"value": "The airplane in the image is a passenger plane."
}
] | 64 | images/sft/0010.zip | a96630dee2b949d81ba77f9548bb9a3d948cf234a7bab8e01d76ac370de6a1f6 |
conversation_58k.json:19:0 | 000000120375.jpg | conversation_58k.json | 000000120375 | 19 | 0 | [
{
"from": "human",
"value": "<image>\nWhat type of vehicle is driving down the street in the image?"
},
{
"from": "gpt",
"value": "A red sports utility vehicle (SUV) is driving down the street in the image."
}
] | 65 | images/sft/0004.zip | 468fda0ff28040eae0c37743af5ddc6a512ecc79db4e480ed85be9b2741f2a49 |
conversation_58k.json:19:1 | 000000120375.jpg | conversation_58k.json | 000000120375 | 19 | 1 | [
{
"from": "human",
"value": "<image>\nIs the street crowded with people?"
},
{
"from": "gpt",
"value": "Yes, the street is filled with a considerable number of people, which indicates that the area is busy."
}
] | 66 | images/sft/0004.zip | 468fda0ff28040eae0c37743af5ddc6a512ecc79db4e480ed85be9b2741f2a49 |
conversation_58k.json:20:0 | 000000179037.jpg | conversation_58k.json | 000000179037 | 20 | 0 | [
{
"from": "human",
"value": "<image>\nWhat type of food is in the box in the image?"
},
{
"from": "gpt",
"value": "There is a loaded pizza pie with toppings inside the box in the image."
}
] | 67 | images/sft/0006.zip | adb5c2b51c32d188cebb8f68647dfbc9d0f5e6d2fdba07b3913dbc72cafdcf3c |
conversation_58k.json:20:1 | 000000179037.jpg | conversation_58k.json | 000000179037 | 20 | 1 | [
{
"from": "human",
"value": "<image>\nWhere is the pizza box placed?"
},
{
"from": "gpt",
"value": "The pizza box is placed on the grass."
}
] | 68 | images/sft/0006.zip | adb5c2b51c32d188cebb8f68647dfbc9d0f5e6d2fdba07b3913dbc72cafdcf3c |
conversation_58k.json:20:2 | 000000179037.jpg | conversation_58k.json | 000000179037 | 20 | 2 | [
{
"from": "human",
"value": "<image>\nIs the pizza box open or closed?"
},
{
"from": "gpt",
"value": "The pizza box is open, and the whole pizza with toppings is visible."
}
] | 69 | images/sft/0006.zip | adb5c2b51c32d188cebb8f68647dfbc9d0f5e6d2fdba07b3913dbc72cafdcf3c |
conversation_58k.json:20:3 | 000000179037.jpg | conversation_58k.json | 000000179037 | 20 | 3 | [
{
"from": "human",
"value": "<image>\nWhat can you tell about the size of the pizza?"
},
{
"from": "gpt",
"value": "From the image, it appears to be a large pizza."
}
] | 70 | images/sft/0006.zip | adb5c2b51c32d188cebb8f68647dfbc9d0f5e6d2fdba07b3913dbc72cafdcf3c |
conversation_58k.json:21:0 | 000000376970.jpg | conversation_58k.json | 000000376970 | 21 | 0 | [
{
"from": "human",
"value": "<image>\nWhat color is the bathroom in the image?"
},
{
"from": "gpt",
"value": "The bathroom in the image is all white."
}
] | 71 | images/sft/0013.zip | da5b5b875ac023de8eeb324b4197ed65606ef2e3b1e197064bc5db3fbdc4c75b |
conversation_58k.json:21:1 | 000000376970.jpg | conversation_58k.json | 000000376970 | 21 | 1 | [
{
"from": "human",
"value": "<image>\nDoes the bathroom have a bathtub or a shower?"
},
{
"from": "gpt",
"value": "The bathroom has a bathtub."
}
] | 72 | images/sft/0013.zip | da5b5b875ac023de8eeb324b4197ed65606ef2e3b1e197064bc5db3fbdc4c75b |
conversation_58k.json:21:2 | 000000376970.jpg | conversation_58k.json | 000000376970 | 21 | 2 | [
{
"from": "human",
"value": "<image>\nWhat kind of cabinet is above the toilet?"
},
{
"from": "gpt",
"value": "There is a white, wooden cabinet above the toilet."
}
] | 73 | images/sft/0013.zip | da5b5b875ac023de8eeb324b4197ed65606ef2e3b1e197064bc5db3fbdc4c75b |
conversation_58k.json:21:4 | 000000376970.jpg | conversation_58k.json | 000000376970 | 21 | 4 | [
{
"from": "human",
"value": "<image>\nAre the walls in the bathroom tiled?"
},
{
"from": "gpt",
"value": "Yes, the walls in the bathroom are covered with white tiles."
}
] | 74 | images/sft/0013.zip | da5b5b875ac023de8eeb324b4197ed65606ef2e3b1e197064bc5db3fbdc4c75b |
conversation_58k.json:22:0 | 000000043464.jpg | conversation_58k.json | 000000043464 | 22 | 0 | [
{
"from": "human",
"value": "<image>\nWhat traffic signs can be seen in the image?"
},
{
"from": "gpt",
"value": "In the image, there are two traffic signs on a post: a \"Stop\" sign and a \"One Way\" sign."
}
] | 75 | images/sft/0001.zip | 562e63e85da32f3bceee58689e79a56e2030fff68723a509f946575b396a0087 |
conversation_58k.json:22:2 | 000000043464.jpg | conversation_58k.json | 000000043464 | 22 | 2 | [
{
"from": "human",
"value": "<image>\nWhat is the condition of the sky in the image?"
},
{
"from": "gpt",
"value": "The sky in the image is cloudy."
}
] | 76 | images/sft/0001.zip | 562e63e85da32f3bceee58689e79a56e2030fff68723a509f946575b396a0087 |
conversation_58k.json:22:3 | 000000043464.jpg | conversation_58k.json | 000000043464 | 22 | 3 | [
{
"from": "human",
"value": "<image>\nWhere is the stop sign placed?"
},
{
"from": "gpt",
"value": "The stop sign is placed on a pole on the side of a street."
}
] | 77 | images/sft/0001.zip | 562e63e85da32f3bceee58689e79a56e2030fff68723a509f946575b396a0087 |
conversation_58k.json:23:0 | 000000326761.jpg | conversation_58k.json | 000000326761 | 23 | 0 | [
{
"from": "human",
"value": "<image>\nWhat type of television is shown in the image?"
},
{
"from": "gpt",
"value": "The image shows an old black and white television set."
}
] | 78 | images/sft/0011.zip | bedd9d9e9b586c7e63b3d8cefcb10a00ec1b3b9bb53ee682e589ae564c2df5d5 |
conversation_58k.json:23:1 | 000000326761.jpg | conversation_58k.json | 000000326761 | 23 | 1 | [
{
"from": "human",
"value": "<image>\nWhat is displayed on the television screen?"
},
{
"from": "gpt",
"value": "There is a picture of a car being displayed on the television screen."
}
] | 79 | images/sft/0011.zip | bedd9d9e9b586c7e63b3d8cefcb10a00ec1b3b9bb53ee682e589ae564c2df5d5 |
conversation_58k.json:23:2 | 000000326761.jpg | conversation_58k.json | 000000326761 | 23 | 2 | [
{
"from": "human",
"value": "<image>\nIs the television wall-mounted or placed on a surface?"
},
{
"from": "gpt",
"value": "The television is mounted on a wall with cords visible behind it."
}
] | 80 | images/sft/0011.zip | bedd9d9e9b586c7e63b3d8cefcb10a00ec1b3b9bb53ee682e589ae564c2df5d5 |
conversation_58k.json:23:3 | 000000326761.jpg | conversation_58k.json | 000000326761 | 23 | 3 | [
{
"from": "human",
"value": "<image>\nIs there any noticeable reflection on the television screen?"
},
{
"from": "gpt",
"value": "Yes, there is a light reflecting off the television screen, which is somewhat affecting the visibility of the car displayed on it."
}
] | 81 | images/sft/0011.zip | bedd9d9e9b586c7e63b3d8cefcb10a00ec1b3b9bb53ee682e589ae564c2df5d5 |
conversation_58k.json:24:0 | 000000269327.jpg | conversation_58k.json | 000000269327 | 24 | 0 | [
{
"from": "human",
"value": "<image>\nWhat is the design of the cake in the image?"
},
{
"from": "gpt",
"value": "The cake in the image is designed to look like a nurse's hat."
}
] | 82 | images/sft/0009.zip | 906adaf88b665889826e42a05672fad8c2bc05facfa2bf510fd6633c51eca61d |
conversation_58k.json:24:1 | 000000269327.jpg | conversation_58k.json | 000000269327 | 24 | 1 | [
{
"from": "human",
"value": "<image>\nWhat shape is the cake in?"
},
{
"from": "gpt",
"value": "The cake is in the shape of a nurse's hat."
}
] | 83 | images/sft/0009.zip | 906adaf88b665889826e42a05672fad8c2bc05facfa2bf510fd6633c51eca61d |
conversation_58k.json:24:2 | 000000269327.jpg | conversation_58k.json | 000000269327 | 24 | 2 | [
{
"from": "human",
"value": "<image>\nWhat is written on the cake?"
},
{
"from": "gpt",
"value": "The words \"Congratulations Danielle\" are written on the cake."
}
] | 84 | images/sft/0009.zip | 906adaf88b665889826e42a05672fad8c2bc05facfa2bf510fd6633c51eca61d |
conversation_58k.json:25:0 | 000000023444.jpg | conversation_58k.json | 000000023444 | 25 | 0 | [
{
"from": "human",
"value": "<image>\nWhat kind of appliance is shown in the image?"
},
{
"from": "gpt",
"value": "The appliance shown in the image is a toaster oven."
}
] | 85 | images/sft/0000.zip | 27e9f58968e43e5ff791c2fadbb92b9c0150485eb4b89bd333c970c96b895a02 |
conversation_58k.json:25:1 | 000000023444.jpg | conversation_58k.json | 000000023444 | 25 | 1 | [
{
"from": "human",
"value": "<image>\nIs there any device attached to the toaster oven in the image?"
},
{
"from": "gpt",
"value": "Yes, there is a temperature gauge attached to the toaster oven in the image."
}
] | 86 | images/sft/0000.zip | 27e9f58968e43e5ff791c2fadbb92b9c0150485eb4b89bd333c970c96b895a02 |
conversation_58k.json:25:4 | 000000023444.jpg | conversation_58k.json | 000000023444 | 25 | 4 | [
{
"from": "human",
"value": "<image>\nIs there any sign of a bug or timer in the microwave?"
},
{
"from": "gpt",
"value": "As the image shows a toaster oven and not a microwave, there is no sign of a bug or timer in a microwave. The focus should be on the toaster oven and the attached temperatur... | 87 | images/sft/0000.zip | 27e9f58968e43e5ff791c2fadbb92b9c0150485eb4b89bd333c970c96b895a02 |
conversation_58k.json:26:0 | 000000108113.jpg | conversation_58k.json | 000000108113 | 26 | 0 | [
{
"from": "human",
"value": "<image>\nWhat is the position of the dog in the image?"
},
{
"from": "gpt",
"value": "The dog is sitting next to the donut pillow."
}
] | 88 | images/sft/0003.zip | 55f19f45cb87dee8d298f3af5fe364c9d1bf3669fc03571ff4395fbba9b0c22f |
conversation_58k.json:26:1 | 000000108113.jpg | conversation_58k.json | 000000108113 | 26 | 1 | [
{
"from": "human",
"value": "<image>\nWhat kind of object is accompanying the dog in the image?"
},
{
"from": "gpt",
"value": "The dog is accompanied by a pillow shaped like a donut in the image."
}
] | 89 | images/sft/0003.zip | 55f19f45cb87dee8d298f3af5fe364c9d1bf3669fc03571ff4395fbba9b0c22f |
conversation_58k.json:26:2 | 000000108113.jpg | conversation_58k.json | 000000108113 | 26 | 2 | [
{
"from": "human",
"value": "<image>\nWhat is the dog wearing in the image?"
},
{
"from": "gpt",
"value": "The dog is wearing a sweater in the image."
}
] | 90 | images/sft/0003.zip | 55f19f45cb87dee8d298f3af5fe364c9d1bf3669fc03571ff4395fbba9b0c22f |
conversation_58k.json:26:4 | 000000108113.jpg | conversation_58k.json | 000000108113 | 26 | 4 | [
{
"from": "human",
"value": "<image>\nIs the dog sitting on a pile of blankets or under the donut pillow?"
},
{
"from": "gpt",
"value": "The dog is sitting on a pile of blankets and is under the stuffed donut pillow."
}
] | 91 | images/sft/0003.zip | 55f19f45cb87dee8d298f3af5fe364c9d1bf3669fc03571ff4395fbba9b0c22f |
conversation_58k.json:27:0 | 000000239114.jpg | conversation_58k.json | 000000239114 | 27 | 0 | [
{
"from": "human",
"value": "<image>\nWhat kind of flowers are in the vase?"
},
{
"from": "gpt",
"value": "The vase contains a bouquet of purple tulips."
}
] | 92 | images/sft/0008.zip | ea74fe458c859afb448309686eb0753f8e7af8a22a1193e901ba7c1316403cdb |
conversation_58k.json:27:1 | 000000239114.jpg | conversation_58k.json | 000000239114 | 27 | 1 | [
{
"from": "human",
"value": "<image>\nWhat is the vase sitting on?"
},
{
"from": "gpt",
"value": "The vase is sitting on a small glass table."
}
] | 93 | images/sft/0008.zip | ea74fe458c859afb448309686eb0753f8e7af8a22a1193e901ba7c1316403cdb |
conversation_58k.json:27:2 | 000000239114.jpg | conversation_58k.json | 000000239114 | 27 | 2 | [
{
"from": "human",
"value": "<image>\nWhat type of vase is holding the flowers?"
},
{
"from": "gpt",
"value": "The flowers are held in a glass vase."
}
] | 94 | images/sft/0008.zip | ea74fe458c859afb448309686eb0753f8e7af8a22a1193e901ba7c1316403cdb |
conversation_58k.json:27:3 | 000000239114.jpg | conversation_58k.json | 000000239114 | 27 | 3 | [
{
"from": "human",
"value": "<image>\nIs the table with the vase of flowers located near any specific feature, such as a window?"
},
{
"from": "gpt",
"value": "Yes, the small table with the vase of flowers is situated by a window."
}
] | 95 | images/sft/0008.zip | ea74fe458c859afb448309686eb0753f8e7af8a22a1193e901ba7c1316403cdb |
conversation_58k.json:28:0 | 000000048937.jpg | conversation_58k.json | 000000048937 | 28 | 0 | [
{
"from": "human",
"value": "<image>\nHow many women are in the image holding toothbrushes?"
},
{
"from": "gpt",
"value": "There are two women in the image holding toothbrushes."
}
] | 96 | images/sft/0001.zip | afb2099258b5d6c9adf34f62c3c23c59a753f176618738f1cc7f548d36d3fe82 |
conversation_58k.json:28:1 | 000000048937.jpg | conversation_58k.json | 000000048937 | 28 | 1 | [
{
"from": "human",
"value": "<image>\nWhere are the women standing?"
},
{
"from": "gpt",
"value": "The women are standing near some shelves or a storage area that contains a bunch of files."
}
] | 97 | images/sft/0001.zip | afb2099258b5d6c9adf34f62c3c23c59a753f176618738f1cc7f548d36d3fe82 |
conversation_58k.json:29:0 | 000000037662.jpg | conversation_58k.json | 000000037662 | 29 | 0 | [
{
"from": "human",
"value": "<image>\nWhat objects are present in the scene?"
},
{
"from": "gpt",
"value": "In the scene, there are a cap, a baseball glove, a baseball, and some green grass."
}
] | 98 | images/sft/0001.zip | 1ae46af519b28c7576d47696cdc4f02460789c87ca8f7f7cfe8decdc4e81d9a7 |
conversation_58k.json:29:1 | 000000037662.jpg | conversation_58k.json | 000000037662 | 29 | 1 | [
{
"from": "human",
"value": "<image>\nWhat is the color of the cap?"
},
{
"from": "gpt",
"value": "The color of the cap is red."
}
] | 99 | images/sft/0001.zip | 1ae46af519b28c7576d47696cdc4f02460789c87ca8f7f7cfe8decdc4e81d9a7 |
End of preview. Expand in Data Studio
MiniMind-V LLaVA Short Training — v1.0.1
Exact data used by MiniMind-V Dense 64M LLaVA Short v1.0, completed 2026-10-25. This release covers the successful short-data experiment, not the stopped full 558K/150K run or ShareGPT4V.
| Stage | Selected | Train | Validation | Test |
|---|---|---|---|---|
| Projector alignment | 50,000 | 49,744 | 256 | 0 |
| Visual SFT (standalone QA) | 150,607 | 149,227 | 690 | 690 |
Alignment: seed 42 uniform sample of captions with 1–19 whitespace-separated words from blip_laion_cc_sbu_558k.json. Visual SFT: extract standalone simple identification/conversation QA, answer 1–2 sentences by punctuation heuristic and <=40 words, question <=30 words. Exclude complex-reasoning component and matching image/QA pairs; filter speculative/context-dependent questions, uncertainty and duplicates. All retained SFT pairs came from conversation_58k.json; zero from detail_23k. This is a deterministic heuristic, not human annotation of simplicity.
Heldouts are grouped by SHA256 of EXIF-corrected RGB image pixels resized to 256x256 BICUBIC, including exclusion across training stages. provenance/*-record-splits.json specifies actual row membership. selected/*.json retains the full selected source order; optimizer training uses only each config's train split. See exact arrays and image-index mapping for reconstruction.
images/{stage}/*.zip contain only selected original image bytes, preserving the original filenames under each row's image key. These are not resized or JPEG-reencoded copies. image_archive identifies the ZIP shard. provenance/*-image-archives.json includes original archive members, byte SHA256 and resized pixel SHA256. provenance/pretrain-image-attribution.json preserves upstream image URLs and caption metadata; provenance/coco-image-attribution.json preserves selected COCO image URLs and individual license records. Text Parquet rows link to ZIP images rather than duplicating the same image for every QA pair.
import io, zipfile
from PIL import Image, ImageOps
from datasets import load_dataset
from huggingface_hub import hf_hub_download
repo = "caspar/MiniMind-V-LLaVA-Short-Training-v1.0"
row = load_dataset(repo, "sft", split="test", revision="v1.0.1")[0]
path = hf_hub_download(repo, row["image_archive"], repo_type="dataset", revision="v1.0.1")
with zipfile.ZipFile(path) as z:
image = ImageOps.exif_transpose(Image.open(io.BytesIO(z.read(row["image"])))).convert("RGB")
print(row["conversations"])
Training arrays: little-endian int16 [N,2,340] (input IDs and assistant labels; ignore=-100); corresponding *-images.npy indexes provenance/*-images.json. Windows use the exact original recipe, tokenizer and image processor config. Images use a frozen SigLIP2 encoder; projector alignment then projector plus first/last LLM layer SFT, one epoch each. See the recipe and manifests for counts and settings.
License and attribution — research reproduction
The LLaVA-Pretrain packaged images were provided by the upstream authors only for research reproduction. That condition is retained here; image rights and LAION/CC/SBU conditions remain with their owners. BLIP synthetic captions retain the applicable BLIP license. LLaVA-Instruct annotation metadata lists CC-BY-4.0; its upstream card additionally requires compliance with OpenAI terms. COCO images retain their individual original image licenses. This release does not grant a blanket commercial image license or override upstream restrictions. Preserve source attribution and honor upstream removal requests. The pinned upstream source cards and their original license links are included in provenance/source-cards/.
Pinned sources
- liuhaotian/LLaVA-Pretrain —
70f9d1e5e1a697fe35830875cfc7de1dd590d727 - liuhaotian/LLaVA-Instruct-150K —
9d451dc7629cfe0469f6ae4432b765cd603d5fcb
Integrity and reproducibility
SHA256SUMS.json lists every released artifact except itself. Original training manifests retain historical /shared-volume/ paths for provenance; use this repository’s relative paths when loading. recipe/ contains the original preparation logic as reference, not a portable one-command cloud launcher. No API keys, optimizer states or model weights are included. Version: v1.0.1.
Exported row counts:
{
"pretrain_train": 49744,
"pretrain_validation": 256,
"sft_train": 149227,
"sft_validation": 690,
"sft_test": 690
}
Image decoding reproducibility (v1.0.1 metadata update)
The original image bytes are unchanged from v1.0. For exact comparison with the stored training pixel hashes, the release validation reproduced all five sampled split checks using Pillow 11.3.0 with libjpeg-turbo 3.1.1. Pillow 13.0.0 with libjpeg-turbo 3.2.0 produced a different resized pixel hash for the JPEG 000000562394.jpg, despite identical original bytes. Thus pixel-level hashes can depend on the decoder build.
requirements-image-validation.txt pins the validated Pillow version. Confirm PIL.features.version_feature("libjpeg_turbo") == "3.1.1" for bitwise checks; non-wheel or platform-specific builds may differ. This is a verified reproduction environment, not a claim that the unrecorded original training package version was 11.3.0. The original archive byte hashes remain the decoder-independent integrity check. Details are in provenance/image-decoder-validation.json.
Version v1.0.1 only adds these reproducibility notes and requirements; image, text, split and token-array data are unchanged. Original v1.0 remains available as an immutable tag.
- Downloads last month
- 132