Domů Procházet Nápověda
Přihlásit se
kekezack
/
X_SSL_Net
0
0
Rozštěpit 0
Soubory Úkoly 0 Pull Requesty 0 Wiki
Strom: fbbbfe53c8
Větve Značky
X_SSL_Net
/
tmp
/
VMamba-main
/
analyze
/
mmpretrain_configs
/
configs
/
blip2
/
metafile.yml

metafile.yml 2.4 KB
Historie Surový

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071 
  1. Collections:
    2. 

  2.   - Name: BLIP-2
    3. 

  3.     Metadata:
    4. 

  4.       Training Data:
    5. 

  5.         - COCO
    6. 

  6.         - VG
    7. 

  7.         - CC3M
    8. 

  8.         - CC12M
    9. 

  9.         - SBU
    10. 

  10.         - LAION-400M
    11. 

  11.       Training Resources: 8x A100 GPUs
    12. 

  12.       Architecture:
    13. 

  13.         - Transformer
    14. 

  14.         - Q-Former
    15. 

  15.     Paper:
    16. 

  16.       Title: 'BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image
    17. 

  17.         Encoders and Large Language Models'
    18. 

  18.       URL: https://arxiv.org/abs/2301.12597
    19. 

  19.     README: configs/blip2/README.md
    20. 

  20. 

  21. Models:
    22. 

  22.   - Name: blip2_3rdparty_retrieval
    23. 

  23.     Metadata:
    24. 

  24.       FLOPs: null
    25. 

  25.       Parameters: 1173191358
    26. 

  26.     In Collection: BLIP-2
    27. 

  27.     Results:
    28. 

  28.       - Task: Image-To-Text Retrieval
    29. 

  29.         Dataset: COCO
    30. 

  30.         Metrics:
    31. 

  31.           Recall@1: 85.4
    32. 

  32.       - Task: Text-To-Image Retrieval
    33. 

  33.         Dataset: COCO
    34. 

  34.         Metrics:
    35. 

  35.           Recall@1: 68.3
    36. 

  36.     Weights: https://download.openmmlab.com/mmclassification/v1/blip2/blip2_3rdparty_pretrain_20230505-f7ef4390.pth
    37. 

  37.     Config: configs/blip2/blip2_8xb32_retrieval.py
    38. 

  38.     Converted From:
    39. 

  39.       Weights: https://storage.googleapis.com/sfr-vision-language-research/LAVIS/models/BLIP2/blip2_pretrained_opt2.7b.pth
    40. 

  40.       Code: https://github.com/salesforce/LAVIS
    41. 

  41.   - Name: blip2-opt2.7b_3rdparty-zeroshot_vqa
    42. 

  42.     Metadata:
    43. 

  43.       FLOPs: null
    44. 

  44.       Parameters: 3770465152
    45. 

  45.     In Collection: BLIP-2
    46. 

  46.     Results:
    47. 

  47.       - Task: Visual Question Answering
    48. 

  48.         Dataset: VQAv2
    49. 

  49.         Metrics:
    50. 

  50.           Accuracy: 53.5
    51. 

  51.     Weights: https://download.openmmlab.com/mmclassification/v1/blip2/blip2-opt2.7b_3rdparty_pretrain_20230505-b51db4e1.pth
    52. 

  52.     Config: configs/blip2/blip2-opt2.7b_8xb16_vqa.py
    53. 

  53.     Converted From:
    54. 

  54.       Weights: https://storage.googleapis.com/sfr-vision-language-research/LAVIS/models/BLIP2/blip2_pretrained_opt2.7b.pth
    55. 

  55.       Code: https://github.com/salesforce/LAVIS
    56. 

  56.   - Name: blip2-opt2.7b_3rdparty-zeroshot_caption
    57. 

  57.     Metadata:
    58. 

  58.       FLOPs: null
    59. 

  59.       Parameters: 3770465152
    60. 

  60.     In Collection: BLIP-2
    61. 

  61.     Results:
    62. 

  62.       - Task: Image Caption
    63. 

  63.         Dataset: COCO
    64. 

  64.         Metrics:
    65. 

  65.           BLEU-4: 32.90
    66. 

  66.           CIDER: 111.10
    67. 

  67.     Weights: https://download.openmmlab.com/mmclassification/v1/blip2/blip2-opt2.7b_3rdparty_pretrain_20230505-b51db4e1.pth
    68. 

  68.     Config: configs/blip2/blip2-opt2.7b_8xb32_caption.py
    69. 

  69.     Converted From:
    70. 

  70.       Weights: https://storage.googleapis.com/sfr-vision-language-research/LAVIS/models/BLIP2/blip2_pretrained_opt2.7b.pth
    71. 

  71.       Code: https://github.com/salesforce/LAVIS
    72.