Upload folder using huggingface_hub

.gitattributes

@@ -39643,3 +39643,61 @@ AAA_outputsxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_text_2dquery256_blip3o60k_VLM
 AAA_outputsxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low3_rerun/00015/samples/00001.png filter=lfs diff=lfs merge=lfs -text
 AAA_outputsxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low3_rerun/00015/samples/00002.png filter=lfs diff=lfs merge=lfs -text
 AAA_outputsxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low3_rerun/00015/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00000/grid.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00000/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00000/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00000/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00000/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00001/grid.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00001/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00001/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00001/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00001/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00002/grid.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00002/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00002/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00002/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00002/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00003/grid.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00003/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00003/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00003/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00003/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00004/grid.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00004/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00004/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00004/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00004/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00005/grid.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00005/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00005/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00005/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00005/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00006/grid.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00006/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00006/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00006/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00006/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00007/grid.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00007/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00007/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00007/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00007/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00008/grid.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00008/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00008/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00008/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00008/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00009/grid.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00009/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00009/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00009/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00009/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00010/grid.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00010/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00010/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00010/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00010/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00011/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00011/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00011/samples/00002.png filter=lfs diff=lfs merge=lfs -text

AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00000/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "single_object", "include": [{"class": "bench", "count": 1}], "prompt": "a photo of a bench", "detailed_caption": "A clear photo of a wooden bench situated in an outdoor setting. The bench has a simple, classic design with a slatted seat and backrest, and sturdy armrests. It is positioned on a flat surface, surrounded by a peaceful natural backdrop, perhaps with grass or trees, but without any distracting elements, ensuring the bench itself is the focal point of the image.", "index": "00000"}

AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00001/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "single_object", "include": [{"class": "cow", "count": 1}], "prompt": "a photo of a cow", "detailed_caption": "A clear photo of a cow standing in a green pasture. The cow has a distinctive pattern of black and white patches on its body and a gentle, curious expression. The lush grass underfoot adds a vibrant touch to the scene, while the background is a simple open field, keeping the focus on the cow.", "index": "00001"}

AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00002/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "single_object", "include": [{"class": "bicycle", "count": 1}], "prompt": "a photo of a bicycle", "detailed_caption": "A clear photo of a bicycle standing upright on a flat surface. The bicycle has a classic design with a sturdy frame, two wheels, and visible handlebars. It features a comfortable seat and pedals, and its color is a vibrant blue. The background is simple and unobtrusive, ensuring that the bicycle remains the focal point of the image.", "index": "00002"}

AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00003/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "single_object", "include": [{"class": "clock", "count": 1}], "prompt": "a photo of a clock", "detailed_caption": "A detailed photo of a clock hanging on a plain wall. The clock features a round face with clear black numbers and hands on a white background, making it easy to read the time. The clock's frame is simple and elegant, adding a touch of sophistication to the overall design. The plain wall serves as an unobtrusive backdrop, ensuring that the focus remains solely on the clock.", "index": "00003"}

AAA_outputxxx_Oct18C_qwen25vl_3b_sana_1600m_512px_empty_text_2dquery256_blip3o60k_VLMlora0_DiTfromscratch_gen_low4_rerun/00004/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "single_object", "include": [{"class": "carrot", "count": 1}], "prompt": "a photo of a carrot", "detailed_caption": "A detailed photo of a fresh carrot placed on a plain surface. The carrot is vibrant orange with a smooth texture and topped with lush green leafy stems. The background is simple and unobtrusive, ensuring the carrot stands out as the main focus of the image.", "index": "00004"}