Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed because of a cast error
Error code: DatasetGenerationCastError
Exception: DatasetGenerationCastError
Message: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 10 new columns ({'abs_path', 'seed', 'attempt', 'occupation', 'prompt', 'prompt_form', 'gemini_pred', 'prompt_id', 'method', 'variant'}) and 16 missing columns ({'n_faces', 'main_person_clear', 'n_face_keypoints', 'person_bottom_frac', 'n_lower_keypoints', 'largest_face_frac', 'ok', 'person_fracs', 'largest_person_frac', 'n_persons', 'face_visible', 'single_person', 'person_height_frac', 'body_visible', 'face_fracs', 'weight_ok'}).
This happened while the csv dataset builder was generating data using
hf://datasets/fassabilf/dpo-train-us-sd15/flip_pool/accountant_train/predictions.csv (at revision 537790c9c3156bf76191c6c348c6cd6fafae4943), ['hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/accountant_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/accountant_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/administrative_assistant_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/administrative_assistant_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/construction_worker_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/construction_worker_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/elementary_teacher_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/elementary_teacher_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/engineer_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/engineer_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/healthcare_assistant_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/healthcare_assistant_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/manager_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/manager_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/mechanic_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/mechanic_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/nurse_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/nurse_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/office_clerk_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/office_clerk_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/office_professional_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/office_professional_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/salesperson_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/salesperson_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/software_developer_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/software_developer_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/teacher_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/teacher_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/ift_train_csv/us_gender_real_sd15_flip_images.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/ift_train_csv/us_gender_real_sd15_flip_train.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/ift_train_csv/us_gender_real_sd15_flip_val.csv']
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1848, in _prepare_split_single
writer.write_table(table)
~~~~~~~~~~~~~~~~~~^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 765, in write_table
self._write_table(pa_table, writer_batch_size=writer_batch_size)
~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 773, in _write_table
pa_table = table_cast(pa_table, self._schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
image_path: string
abs_path: string
prompt_id: int64
occupation: string
prompt: string
seed: int64
attempt: int64
variant: string
method: string
prompt_form: string
gemini_pred: string
-- schema metadata --
pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 1548
to
{'image_path': Value('string'), 'n_faces': Value('int64'), 'n_persons': Value('int64'), 'largest_face_frac': Value('float64'), 'largest_person_frac': Value('float64'), 'n_face_keypoints': Value('int64'), 'face_visible': Value('bool'), 'single_person': Value('bool'), 'main_person_clear': Value('bool'), 'n_lower_keypoints': Value('int64'), 'person_height_frac': Value('float64'), 'person_bottom_frac': Value('float64'), 'face_fracs': Value('string'), 'person_fracs': Value('string'), 'ok': Value('bool'), 'body_visible': Value('bool'), 'weight_ok': Value('bool')}
because column names don't match
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
~~~~~~~~~~~~~~~~~~~~~~~~~^
builder, max_dataset_size_bytes=max_dataset_size_bytes
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
for job_id, done, content in self._prepare_split_single(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
):
^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1850, in _prepare_split_single
raise DatasetGenerationCastError.from_cast_error(
...<4 lines>...
)
datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 10 new columns ({'abs_path', 'seed', 'attempt', 'occupation', 'prompt', 'prompt_form', 'gemini_pred', 'prompt_id', 'method', 'variant'}) and 16 missing columns ({'n_faces', 'main_person_clear', 'n_face_keypoints', 'person_bottom_frac', 'n_lower_keypoints', 'largest_face_frac', 'ok', 'person_fracs', 'largest_person_frac', 'n_persons', 'face_visible', 'single_person', 'person_height_frac', 'body_visible', 'face_fracs', 'weight_ok'}).
This happened while the csv dataset builder was generating data using
hf://datasets/fassabilf/dpo-train-us-sd15/flip_pool/accountant_train/predictions.csv (at revision 537790c9c3156bf76191c6c348c6cd6fafae4943), ['hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/accountant_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/accountant_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/administrative_assistant_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/administrative_assistant_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/construction_worker_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/construction_worker_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/elementary_teacher_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/elementary_teacher_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/engineer_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/engineer_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/healthcare_assistant_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/healthcare_assistant_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/manager_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/manager_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/mechanic_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/mechanic_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/nurse_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/nurse_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/office_clerk_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/office_clerk_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/office_professional_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/office_professional_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/salesperson_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/salesperson_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/software_developer_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/software_developer_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/teacher_train/framing.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/flip_pool/teacher_train/predictions.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/ift_train_csv/us_gender_real_sd15_flip_images.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/ift_train_csv/us_gender_real_sd15_flip_train.csv', 'hf://datasets/fassabilf/dpo-train-us-sd15@537790c9c3156bf76191c6c348c6cd6fafae4943/ift_train_csv/us_gender_real_sd15_flip_val.csv']
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
image_path string | n_faces int64 | n_persons int64 | largest_face_frac float64 | largest_person_frac float64 | n_face_keypoints int64 | face_visible bool | single_person bool | main_person_clear bool | n_lower_keypoints int64 | person_height_frac float64 | person_bottom_frac float64 | face_fracs string | person_fracs string | ok bool | body_visible bool | weight_ok bool |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
images/accountant/000_s0.png | 1 | 1 | 0.2264 | 0.7509 | 5 | true | true | true | 0 | 0.8795 | 0.9177 | 0.2264 | 0.7509 | true | false | false |
images/accountant/001_s0.png | 1 | 1 | 0.1972 | 0.7056 | 5 | true | true | true | 2 | 0.8013 | 0.8354 | 0.1972 | 0.7056 | true | true | true |
images/accountant/002_s0.png | 1 | 1 | 0.2054 | 0.6883 | 5 | true | true | true | 0 | 0.7816 | 0.9965 | 0.2054 | 0.6883 | true | false | false |
images/accountant/003_s0.png | 1 | 1 | 0.2551 | 0.8391 | 5 | true | true | true | 0 | 0.9024 | 0.9461 | 0.2551 | 0.8391 | true | false | false |
images/accountant/004_s0.png | 1 | 1 | 0.1698 | 0.7065 | 5 | true | true | true | 1 | 0.6906 | 0.7469 | 0.1698 | 0.7065 | true | true | true |
images/accountant/005_s0.png | 1 | 1 | 0.2148 | 0.8105 | 5 | true | true | true | 0 | 0.8555 | 0.908 | 0.2148 | 0.8105 | true | false | false |
images/accountant/006_s0.png | 1 | 1 | 0.2279 | 0.8093 | 5 | true | true | true | 2 | 0.8403 | 0.9033 | 0.2279 | 0.8093 | true | true | true |
images/accountant/007_s0.png | 1 | 1 | 0.2674 | 0.779 | 5 | true | true | true | 0 | 0.8766 | 0.939 | 0.2674 | 0.7790 | true | false | false |
images/accountant/016_s0.png | 1 | 1 | 0.5522 | 0.7384 | 0 | true | true | true | 0 | 0.6099 | 0.6143 | 0.5522 | 0.7384 | true | false | false |
images/accountant/017_s0.png | 1 | 1 | 0.2557 | 0.7371 | 5 | true | true | true | 0 | 0.843 | 0.8968 | 0.2557 | 0.7371 | true | false | false |
images/accountant/018_s0.png | 1 | 1 | 0.2138 | 0.685 | 5 | true | true | true | 0 | 0.7918 | 0.9967 | 0.2138 | 0.6850 | true | false | false |
images/accountant/019_s0.png | 1 | 1 | 0.2496 | 0.7891 | 5 | true | true | true | 0 | 0.9086 | 0.9554 | 0.2496 | 0.7891 | true | false | false |
images/accountant/012_s0.png | 1 | 1 | 0.2162 | 0.8542 | 5 | true | true | true | 0 | 0.9049 | 0.9476 | 0.2162 | 0.8542;0.1571 | true | false | false |
images/accountant/013_s0.png | 1 | 1 | 0.2094 | 0.7018 | 5 | true | true | true | 0 | 0.6846 | 0.7414 | 0.2094 | 0.7018 | true | false | false |
images/accountant/014_s0.png | 0 | 1 | 0 | 0.69 | 0 | false | true | true | 2 | 0.5526 | 0.5818 | null | 0.6900 | false | true | true |
images/accountant/015_s0.png | 2 | 2 | 0.1072 | 0.6028 | 5 | true | false | false | 4 | 0.8821 | 0.9539 | 0.1072;0.0892 | 0.6028;0.4381 | false | true | false |
images/accountant/008_s0.png | 1 | 1 | 0.2413 | 0.7861 | 5 | true | true | true | 0 | 0.9004 | 0.9506 | 0.2413 | 0.7861 | true | false | false |
images/accountant/009_s0.png | 1 | 1 | 0.5222 | 0.886 | 5 | true | true | true | 0 | 0.978 | 0.978 | 0.5222 | 0.8860 | true | false | false |
images/accountant/010_s0.png | 0 | 0 | 0 | 0 | 0 | false | false | false | 0 | 0 | 0 | null | null | false | false | false |
images/accountant/011_s0.png | 2 | 2 | 0.2183 | 0.7831 | 5 | true | false | true | 0 | 0.8843 | 0.9378 | 0.2183;0.1923 | 0.7831;0.4260 | true | false | false |
images/accountant/028_s0.png | 1 | 1 | 0.263 | 0.8039 | 4 | true | true | true | 2 | 0.9652 | 0.9849 | 0.2630 | 0.8039 | true | true | true |
images/accountant/029_s0.png | 1 | 1 | 0.2287 | 0.84 | 5 | true | true | true | 0 | 0.8797 | 0.9308 | 0.2287 | 0.8400 | true | false | false |
images/accountant/030_s0.png | 1 | 1 | 0.2355 | 0.8364 | 5 | true | true | true | 0 | 0.9209 | 0.9686 | 0.2355 | 0.8364 | true | false | false |
images/accountant/031_s0.png | 1 | 1 | 0.2298 | 0.7886 | 5 | true | true | true | 0 | 0.8944 | 0.9464 | 0.2298 | 0.7886 | true | false | false |
images/accountant/024_s0.png | 1 | 1 | 0.1898 | 0.6614 | 5 | true | true | true | 0 | 0.7085 | 0.756 | 0.1898 | 0.6614 | true | false | false |
images/accountant/025_s0.png | 1 | 1 | 0.2622 | 0.8216 | 5 | true | true | true | 0 | 0.8998 | 0.9477 | 0.2622 | 0.8216 | true | false | false |
images/accountant/026_s0.png | 1 | 1 | 0.2217 | 0.8485 | 5 | true | true | true | 0 | 0.8803 | 0.9309 | 0.2217 | 0.8485 | true | false | false |
images/accountant/027_s0.png | 0 | 1 | 0 | 0.7505 | 2 | false | true | true | 0 | 0.6403 | 0.6403 | null | 0.7505 | false | false | false |
images/accountant/020_s0.png | 1 | 1 | 0.2508 | 0.8412 | 5 | true | true | true | 0 | 0.8933 | 0.9425 | 0.2508 | 0.8412 | true | false | false |
images/accountant/021_s0.png | 1 | 1 | 0.2307 | 0.7484 | 5 | true | true | true | 0 | 0.8754 | 0.9274 | 0.2307 | 0.7484 | true | false | false |
images/accountant/022_s0.png | 1 | 1 | 0.2209 | 0.7543 | 5 | true | true | true | 0 | 0.8806 | 0.9283 | 0.2209 | 0.7543 | true | false | false |
images/accountant/023_s0.png | 1 | 1 | 0.224 | 0.8457 | 5 | true | true | true | 0 | 0.8834 | 0.9384 | 0.2240 | 0.8457 | true | false | false |
images/accountant/040_s0.png | 1 | 1 | 0.202 | 0.6645 | 5 | true | true | true | 0 | 0.6816 | 0.9685 | 0.2020 | 0.6645 | true | false | false |
images/accountant/041_s0.png | 1 | 1 | 0.237 | 0.7993 | 5 | true | true | true | 2 | 0.846 | 0.9072 | 0.2370 | 0.7993 | true | true | true |
images/accountant/042_s0.png | 1 | 1 | 0.2308 | 0.819 | 5 | true | true | true | 1 | 0.8614 | 0.9192 | 0.2308 | 0.8190 | true | true | true |
images/accountant/043_s0.png | 1 | 1 | 0.2265 | 0.7892 | 5 | true | true | true | 0 | 0.8867 | 0.9351 | 0.2265 | 0.7892 | true | false | false |
images/accountant/032_s0.png | 1 | 1 | 0.2405 | 0.7777 | 5 | true | true | true | 0 | 0.8858 | 0.9378 | 0.2405 | 0.7777 | true | false | false |
images/accountant/033_s0.png | 1 | 1 | 0.2149 | 0.7148 | 5 | true | true | true | 0 | 0.8891 | 0.9897 | 0.2149 | 0.7148 | true | false | false |
images/accountant/034_s0.png | 0 | 1 | 0 | 0.5785 | 2 | false | true | true | 6 | 0.9567 | 0.9908 | null | 0.5785 | false | true | true |
images/accountant/035_s0.png | 0 | 1 | 0 | 0.6886 | 3 | false | true | true | 0 | 0.6126 | 0.6175 | null | 0.6886 | false | false | false |
images/accountant/036_s0.png | 1 | 1 | 0.2256 | 0.8144 | 5 | true | true | true | 0 | 0.8553 | 0.9222 | 0.2256 | 0.8144 | true | false | false |
images/accountant/037_s0.png | 1 | 1 | 0.2367 | 0.8708 | 5 | true | true | true | 1 | 0.9038 | 0.9605 | 0.2367 | 0.8708 | true | true | true |
images/accountant/038_s0.png | 1 | 1 | 0.25 | 0.6337 | 5 | true | true | true | 0 | 0.7842 | 0.8321 | 0.2500 | 0.6337 | true | false | false |
images/accountant/039_s0.png | 1 | 1 | 0.239 | 0.8054 | 5 | true | true | true | 0 | 0.8574 | 0.917 | 0.2390 | 0.8054 | true | false | false |
images/accountant/044_s0.png | 1 | 1 | 0.1723 | 0.7843 | 5 | true | true | true | 0 | 0.8668 | 0.9264 | 0.1723 | 0.7843 | true | false | false |
images/accountant/045_s0.png | 1 | 1 | 0.2342 | 0.8592 | 5 | true | true | true | 0 | 0.9061 | 0.9519 | 0.2342 | 0.8592 | true | false | false |
images/accountant/046_s0.png | 0 | 1 | 0 | 0.6799 | 3 | false | true | true | 0 | 0.9389 | 0.9618 | null | 0.6799 | false | false | false |
images/accountant/047_s0.png | 1 | 1 | 0.2539 | 0.8495 | 5 | true | true | true | 1 | 0.9256 | 0.9835 | 0.2539 | 0.8495 | true | true | true |
images/accountant/052_s0.png | 1 | 1 | 0.2213 | 0.8324 | 5 | true | true | true | 0 | 0.8756 | 0.927 | 0.2213 | 0.8324 | true | false | false |
images/accountant/053_s0.png | 1 | 1 | 0.2364 | 0.8032 | 5 | true | true | true | 0 | 0.8836 | 0.927 | 0.2364 | 0.8032 | true | false | false |
images/accountant/054_s0.png | 2 | 2 | 0.2346 | 0.762 | 5 | true | false | true | 0 | 0.8851 | 0.9268 | 0.2346;0.2191 | 0.7620;0.4176 | true | false | false |
images/accountant/055_s0.png | 1 | 1 | 0.2197 | 0.7682 | 5 | true | true | true | 0 | 0.8455 | 0.8923 | 0.2197 | 0.7682 | true | false | false |
images/accountant/048_s0.png | 1 | 1 | 0.4545 | 0.8584 | 4 | true | true | true | 0 | 0.8596 | 0.9007 | 0.4545 | 0.8584 | true | false | false |
images/accountant/049_s0.png | 1 | 1 | 0.1624 | 0.6862 | 5 | true | true | true | 0 | 0.8199 | 0.8621 | 0.1624 | 0.6862 | true | false | false |
images/accountant/050_s0.png | 1 | 1 | 0.1691 | 0.6183 | 5 | true | true | true | 0 | 0.6146 | 0.6579 | 0.1691 | 0.6183 | true | false | false |
images/accountant/051_s0.png | 0 | 1 | 0 | 0.6584 | 1 | false | true | true | 0 | 0.6688 | 0.6958 | null | 0.6584 | false | false | false |
images/accountant/060_s0.png | 1 | 1 | 0.2017 | 0.7027 | 5 | true | true | true | 0 | 0.7134 | 0.7539 | 0.2017 | 0.7027 | true | false | false |
images/accountant/061_s0.png | 1 | 1 | 0.2075 | 0.8056 | 5 | true | true | true | 1 | 0.863 | 0.9069 | 0.2075 | 0.8056 | true | true | true |
images/accountant/062_s0.png | 1 | 1 | 0.2224 | 0.8424 | 5 | true | true | true | 0 | 0.8638 | 0.9199 | 0.2224 | 0.8424 | true | false | false |
images/accountant/063_s0.png | 1 | 1 | 0.153 | 0.6993 | 5 | true | true | true | 1 | 0.807 | 0.8558 | 0.1530 | 0.6993 | true | true | true |
images/accountant/064_s0.png | 1 | 1 | 0.547 | 0.9164 | 5 | true | true | true | 0 | 0.9665 | 0.9665 | 0.5470 | 0.9164 | true | false | false |
images/accountant/065_s0.png | 1 | 1 | 0.242 | 0.7927 | 5 | true | true | true | 0 | 0.8802 | 0.9281 | 0.2420 | 0.7927 | true | false | false |
images/accountant/066_s0.png | 1 | 1 | 0.2453 | 0.7941 | 5 | true | true | true | 0 | 0.9017 | 0.955 | 0.2453 | 0.7941 | true | false | false |
images/accountant/067_s0.png | 1 | 1 | 0.1768 | 0.6811 | 5 | true | true | true | 2 | 0.7955 | 0.9924 | 0.1768 | 0.6811 | true | true | true |
images/accountant/056_s0.png | 1 | 1 | 0.2292 | 0.7647 | 5 | true | true | true | 0 | 0.7981 | 0.8586 | 0.2292 | 0.7647 | true | false | false |
images/accountant/057_s0.png | 1 | 1 | 0.1908 | 0.7207 | 5 | true | true | true | 3 | 0.9507 | 0.9535 | 0.1908 | 0.7207 | true | true | true |
images/accountant/058_s0.png | 1 | 1 | 0.2249 | 0.7146 | 5 | true | true | true | 1 | 0.8411 | 0.8961 | 0.2249 | 0.7146 | true | true | true |
images/accountant/059_s0.png | 0 | 2 | 0 | 0.6839 | 3 | false | false | false | 0 | 0.5955 | 0.6176 | null | 0.6839;0.4816 | false | false | false |
images/accountant/072_s0.png | 0 | 1 | 0 | 0.4247 | 2 | false | true | true | 0 | 0.3385 | 0.5536 | null | 0.4247 | false | false | false |
images/accountant/073_s0.png | 1 | 1 | 0.2239 | 0.8158 | 5 | true | true | true | 0 | 0.8641 | 0.9089 | 0.2239 | 0.8158 | true | false | false |
images/accountant/074_s0.png | 1 | 1 | 0.1556 | 0.7029 | 5 | true | true | true | 0 | 0.7692 | 0.8157 | 0.1556 | 0.7029 | true | false | false |
images/accountant/075_s0.png | 1 | 1 | 0.2246 | 0.7999 | 5 | true | true | true | 0 | 0.8027 | 0.8603 | 0.2246 | 0.7999 | true | false | false |
images/accountant/076_s0.png | 1 | 1 | 0.2651 | 0.7988 | 5 | true | true | true | 0 | 0.8991 | 0.9477 | 0.2651 | 0.7988 | true | false | false |
images/accountant/077_s0.png | 1 | 1 | 0.2426 | 0.8432 | 5 | true | true | true | 1 | 0.8896 | 0.9382 | 0.2426 | 0.8432 | true | true | true |
images/accountant/078_s0.png | 1 | 1 | 0.224 | 0.7744 | 5 | true | true | true | 0 | 0.8692 | 0.9151 | 0.2240 | 0.7744 | true | false | false |
images/accountant/079_s0.png | 0 | 1 | 0 | 0.7857 | 0 | false | true | true | 0 | 0.6834 | 0.7017 | null | 0.7857 | false | false | false |
images/accountant/068_s0.png | 0 | 1 | 0 | 0.6993 | 1 | false | true | true | 3 | 0.6055 | 0.6055 | null | 0.6993 | false | true | true |
images/accountant/069_s0.png | 1 | 2 | 0.2463 | 0.7803 | 4 | true | false | false | 0 | 0.9386 | 0.9747 | 0.2463 | 0.7803;0.6780 | false | false | false |
images/accountant/070_s0.png | 1 | 1 | 0.2379 | 0.8349 | 5 | true | true | true | 0 | 0.8849 | 0.9404 | 0.2379 | 0.8349 | true | false | false |
images/accountant/071_s0.png | 1 | 1 | 0.1608 | 0.7159 | 5 | true | true | true | 0 | 0.7986 | 0.856 | 0.1608 | 0.7159 | true | false | false |
images/accountant/080_s0.png | 1 | 1 | 0.2183 | 0.8497 | 5 | true | true | true | 0 | 0.8715 | 0.9298 | 0.2183 | 0.8497 | true | false | false |
images/accountant/081_s0.png | 0 | 1 | 0 | 0.6973 | 0 | false | true | true | 0 | 0.582 | 0.5929 | null | 0.6973 | false | false | false |
images/accountant/082_s0.png | 1 | 1 | 0.242 | 0.8104 | 5 | true | true | true | 0 | 0.9002 | 0.9401 | 0.2420 | 0.8104 | true | false | false |
images/accountant/083_s0.png | 1 | 1 | 0.2646 | 0.8066 | 5 | true | true | true | 0 | 0.9035 | 0.9527 | 0.2646 | 0.8066 | true | false | false |
images/accountant/088_s0.png | 1 | 2 | 0.5538 | 0.6092 | 3 | true | false | false | 1 | 0.6318 | 0.642 | 0.5538 | 0.6092;0.5648 | false | true | false |
images/accountant/089_s0.png | 1 | 1 | 0.2154 | 0.8183 | 5 | true | true | true | 2 | 0.8696 | 0.9164 | 0.2154 | 0.8183 | true | true | true |
images/accountant/090_s0.png | 1 | 1 | 0.2332 | 0.8518 | 5 | true | true | true | 0 | 0.8797 | 0.9335 | 0.2332 | 0.8518 | true | false | false |
images/accountant/091_s0.png | 1 | 1 | 0.2014 | 0.7936 | 5 | true | true | true | 1 | 0.8069 | 0.8571 | 0.2014 | 0.7936 | true | true | true |
images/accountant/084_s0.png | 1 | 1 | 0.2231 | 0.8226 | 5 | true | true | true | 0 | 0.8909 | 0.9404 | 0.2231 | 0.8226 | true | false | false |
images/accountant/085_s0.png | 0 | 2 | 0 | 0.8104 | 1 | false | false | true | 0 | 0.9643 | 0.9817 | null | 0.8104;0.3948 | false | false | false |
images/accountant/086_s0.png | 1 | 2 | 0.2543 | 0.8559 | 5 | true | false | true | 0 | 0.8823 | 0.9274 | 0.2543 | 0.8559;0.5005;0.1334 | true | false | false |
images/accountant/087_s0.png | 0 | 2 | 0 | 0.6234 | 0 | false | false | true | 2 | 0.5582 | 0.5853 | null | 0.6234;0.2529 | false | true | true |
images/accountant/092_s0.png | 1 | 1 | 0.2207 | 0.7732 | 5 | true | true | true | 0 | 0.8879 | 0.9441 | 0.2207 | 0.7732 | true | false | false |
images/accountant/093_s0.png | 1 | 1 | 0.2599 | 0.7796 | 4 | true | true | true | 0 | 0.9994 | 0.9994 | 0.2599 | 0.7796 | true | false | false |
images/accountant/094_s0.png | 1 | 2 | 0.3022 | 0.8814 | 5 | true | false | true | 0 | 0.9114 | 0.9569 | 0.3022 | 0.8814;0.5449 | true | false | false |
images/accountant/095_s0.png | 0 | 1 | 0 | 0.7865 | 0 | false | true | true | 0 | 0.8636 | 0.8678 | null | 0.7865 | false | false | false |
images/accountant/096_s0.png | 1 | 1 | 0.2362 | 0.7804 | 5 | true | true | true | 0 | 0.9005 | 0.9511 | 0.2362 | 0.7804 | true | false | false |
images/accountant/097_s0.png | 0 | 1 | 0 | 0.6847 | 0 | false | true | true | 1 | 0.6187 | 0.6187 | null | 0.6847 | false | true | true |
images/accountant/098_s0.png | 1 | 1 | 0.2249 | 0.773 | 5 | true | true | true | 1 | 0.8216 | 0.8823 | 0.2249 | 0.7730 | true | true | true |
images/accountant/099_s0.png | 1 | 1 | 0.1936 | 0.6129 | 5 | true | true | true | 0 | 0.6373 | 0.6981 | 0.1936 | 0.6129 | true | false | false |
End of preview.
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Dataset DPO — gambar sintetis SD 1.5 (US, gender)
Pasangan preferensi untuk DiffusionDPO-private/train.py mode DPO — pembanding arm DPO
foto asli (data/dpo_us_real/). Dirakit oleh perakit yang SAMA,
scripts/ift/build_dpo_dataset_real.py, dengan flag pool sintetis:
- chosen = pool IFT sintetis yang dipakai arm SFT sintetis
(
data/ift_train/us_gender_real_sd15_{split}.csv, gambar dioutputs/ift_train_us_sd15/, zip akar HFfassabilf/ift-train-us-sd15), gender menurut alokasi Pers. 6 - rejected = pool flip yang dirender khusus untuk arm ini oleh
scripts/ift/gen_flip_pool_sd15.py(data/ift_train/us_gender_real_sd15_flip_{split}.csv, gambar dioutputs/ift_train_us_sd15_flip/): promptdata/{split}_flip/, caption sama, gender seberang, protokol sama (SD 1.5, 512 px, 30 step, cfg 7,5, gate framing YuNet, judgegemini-3.1-pro-preview, seed terkecil yang lolos) - caption = prompt netral tanpa kata atribut
| pasangan | gambar unik | label_0=1 | |
|---|---|---|---|
| train | 1400 | 2800 | 52,0% |
| val | 308 | 616 | 46,8% |
Plafon implicit accuracy (p² + (1−p)² tertimbang n) = 0,6514 di train dan val —
sama dengan arm foto asli, karena alokasi gender per okupasi identik
(analysis/dpo_sd15/ceiling.csv). Proporsi chosen=female per okupasi = p_female_target
di ke-14 okupasi.
Beda dengan arm foto asli yang harus ikut dibaca
- Tiap gambar dipakai sekali. Arm foto asli memakai ulang 280 foto ~10× (tiap sel okupasi × gender 10 foto); di sini 2.800 gambar train semuanya unik. "Menghafal foto" di arm asli karena itu tidak punya padanan langsung di sini.
- Gambar cocok dengan caption-nya. Foto Pexels tidak dipotret untuk caption tertentu; gambar sintetis dirender dari prompt yang caption-nya persis sama.
- Pasangan minimal. Di 66,6% pasangan train (70,8% val) kedua sisi memakai seed yang sama — noise awal identik, jadi selisihnya hampir hanya kata gender di prompt.
- Unclear rate pool flip tidak sebanding dengan pool chosen. Pool flip hanya
men-judge gambar yang lolos gate (label
not_judgeduntuk sisanya), jadi kolomuncleardi ringkasanbuild_ift_dataset.pyuntuk pool flip ikut menghitungnot_judged. Gambar terpilihnya tidak terpengaruh (aturan "seed terkecil yang lolos").
Kolom parquet
Sama dengan data/dpo_us_real/: jpg_0, jpg_1 (bytes PNG 512×512), label_0
(1 = jpg_0 chosen; posisi diacak ber-seed), caption, plus metadata
slug, prompt_id, tau, p_female_target, photo_id_chosen, photo_id_rejected — di sini
photo_id_* = seed render, bukan id foto.
Isi repo ini
train/data.parquet,val/data.parquet,train.csv,val.csv— pasangan siap pakaiflip_pool/+zips/flip_pool.zip— pool rejected lengkap (outputs/ift_train_us_sd15_flip/: gambar,predictions.csv,framing.csvper okupasi×split); unzip ke folder itu untuk membangun ulang tanpa memanggil Geminiift_train_csv/—data/ift_train/us_gender_real_sd15_flip_{train,val}.csv
Sisi chosen ada di fassabilf/ift-train-us-sd15.
Membangun ulang
# sisi chosen: unzip 28 zip akar fassabilf/ift-train-us-sd15 ke akar repo
python3 scripts/ift/gen_flip_pool_sd15.py # sisi rejected (~15 mnt, ~1,8k Gemini)
python3 scripts/ift/gen_flip_pool_sd15.py --slugs elementary_teacher --splits train --max-seed 200
python3 scripts/ift/build_ift_dataset.py --run-dir outputs/ift_train_us_sd15_flip \
--tag us_gender_real_sd15_flip --prompt-suffix _flip --tau-col tau_gen
python3 scripts/ift/build_dpo_dataset_real.py --csv-dir data/ift_train --tag us_gender_real_sd15 \
--run-dir outputs/ift_train_us_sd15 --rej-run-dir outputs/ift_train_us_sd15_flip \
--out data/dpo_us_sd15
Dua prompt elementary_teacher train (pid 32, 70; minta male) baru terisi di seed 152
dan 75 — hampir semua seed sebelumnya gagal gate framing. Pool chosen juga butuh top-up
sampai s90 di okupasi ini.
- Downloads last month
- 25