config.json 2.6 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101
  1. {
  2. "_flash_attn_2_enabled": true,
  3. "architectures": [
  4. "Idefics3ForConditionalGeneration"
  5. ],
  6. "attention_bias": false,
  7. "attention_dropout": 0.0,
  8. "bos_token_id": 100264,
  9. "eos_token_id": 100338,
  10. "freeze_lm_head": true,
  11. "freeze_text_layers": true,
  12. "freeze_text_module_exceptions": [],
  13. "freeze_vision_layers": true,
  14. "freeze_vision_module_exceptions": [],
  15. "head_dim": 64,
  16. "hidden_act": "silu",
  17. "hidden_size": 576,
  18. "image_token_id": 100270,
  19. "initializer_range": 0.02,
  20. "intermediate_size": 1536,
  21. "max_position_embeddings": 8192,
  22. "mlp_bias": false,
  23. "model_type": "idefics3",
  24. "neftune_noise_alpha": 0.0,
  25. "num_attention_heads": 9,
  26. "num_hidden_layers": 30,
  27. "num_key_value_heads": 3,
  28. "pad_token_id": 100256,
  29. "perceiver_config": {
  30. "attention_dropout": 0.0,
  31. "hidden_act": "silu",
  32. "model_type": "vllama3",
  33. "num_key_value_heads": 1,
  34. "qk_layer_norms_perceiver": false,
  35. "resampler_depth": 6,
  36. "resampler_head_dim": 96,
  37. "resampler_n_heads": 16,
  38. "resampler_n_latents": 64
  39. },
  40. "pixel_shuffle_factor": 4,
  41. "pretraining_tp": 1,
  42. "qk_layer_norms": false,
  43. "rms_norm_eps": 1e-05,
  44. "rope_scaling": null,
  45. "rope_theta": 100000.0,
  46. "scale_factor": 4,
  47. "text_config": {
  48. "architectures": ["VLlama3ForCausalLM"],
  49. "attention_bias": false,
  50. "attention_dropout": 0.0,
  51. "bos_token_id": 100257,
  52. "eos_token_id": 100257,
  53. "head_dim": 64,
  54. "hidden_act": "silu",
  55. "hidden_size": 576,
  56. "initializer_range": 0.02,
  57. "intermediate_size": 1536,
  58. "max_position_embeddings": 8192,
  59. "mlp_bias": false,
  60. "model_type": "llama",
  61. "num_attention_heads": 9,
  62. "num_hidden_layers": 30,
  63. "num_key_value_heads": 3,
  64. "pretraining_tp": 1,
  65. "rms_norm_eps": 1e-5,
  66. "rope_scaling": null,
  67. "rope_theta": 10000.0,
  68. "torch_dtype": "bfloat16",
  69. "use_cache": true,
  70. "vocab_size": 100480,
  71. "tie_word_embeddings": false
  72. },
  73. "tie_word_embeddings": true,
  74. "torch_dtype": "bfloat16",
  75. "transformers_version": "4.51.3",
  76. "use_cache": false,
  77. "use_resampler": false,
  78. "vision_config": {
  79. "attention_dropout": 0.0,
  80. "hidden_act": "gelu_pytorch_tanh",
  81. "hidden_size": 768,
  82. "image_size": 512,
  83. "initializer_range": 0.02,
  84. "intermediate_size": 3072,
  85. "layer_norm_eps": 1e-06,
  86. "max_image_size": {
  87. "longest_edge": 512
  88. },
  89. "model_type": "idefics3_vision",
  90. "num_attention_heads": 12,
  91. "num_channels": 3,
  92. "num_hidden_layers": 12,
  93. "patch_size": 16,
  94. "size": {
  95. "longest_edge": 2048
  96. },
  97. "torch_dtype": "bfloat16",
  98. "use_base_siglip": true
  99. },
  100. "vocab_size": 100480
  101. }