mirror of
https://github.com/immich-app/ml-models.git
synced 2026-09-30 21:27:43 +08:00
Mert
df0ea33c8a
* feat: torch-free export core, shared graph/IR helpers, and CLI scaffolding * feat: fused InsightFace face exporter * feat: RKNN export path with per-SoC compilation * feat: single-partition CoreML CLIP export and runtime rewrite API * feat: refactor to passes, ep-specific fixes and optimizations * feat: check command diffing committed graph renderings, wired into CI * chore: graph and rewrite-plan renderings for the catalog * use pokedex large
1154 lines
73 KiB
Plaintext
1154 lines
73 KiB
Plaintext
<
|
|
ir_version: 10,
|
|
opset_import: ["" : 23],
|
|
producer_name: "pytorch"
|
|
>
|
|
main_graph (uint8[batch,512,512,3] image) => (float[batch,1024] image_embedding)
|
|
<
|
|
float[batch,1024,1024] add_1007
|
|
float[batch,1024,1024] add_1068
|
|
float[batch,1024,1024] add_107
|
|
float[batch,1024,1024] add_1097
|
|
float[batch,1024,1024] add_1158
|
|
float[batch,1024,1024] add_1187
|
|
float[batch,1024,1024] add_1248
|
|
float[batch,1024,1024] add_1277
|
|
float[batch,1024,1024] add_13
|
|
float[batch,1024,1024] add_1338
|
|
float[batch,1024,1024] add_1367
|
|
float[batch,1024,1024] add_1428
|
|
float[batch,1024,1024] add_1457
|
|
float[batch,1024,1024] add_1518
|
|
float[batch,1024,1024] add_1547
|
|
float[batch,1024,1024] add_1608
|
|
float[batch,1024,1024] add_1637
|
|
float[batch,1024,1024] add_168
|
|
float[batch,1024,1024] add_1698
|
|
float[batch,1024,1024] add_1727
|
|
float[batch,1024,1024] add_1788
|
|
float[batch,1024,1024] add_1817
|
|
float[batch,1024,1024] add_1878
|
|
float[batch,1024,1024] add_1907
|
|
float[batch,1024,1024] add_1968
|
|
float[batch,1024,1024] add_197
|
|
float[batch,1024,1024] add_1997
|
|
float[batch,1024,1024] add_2058
|
|
float[batch,1024,1024] add_2087
|
|
float[batch,1024,1024] add_2148
|
|
float[batch,1024,1024] add_2177
|
|
float[batch,1,1024] add_2275
|
|
float[batch,1024,1024] add_258
|
|
float[batch,1024,1024] add_287
|
|
float[batch,1024,1024] add_348
|
|
float[batch,1024,1024] add_377
|
|
float[batch,1024,1024] add_438
|
|
float[batch,1024,1024] add_467
|
|
float[batch,1024,1024] add_528
|
|
float[batch,1024,1024] add_557
|
|
float[batch,1024,1024] add_618
|
|
float[batch,1024,1024] add_647
|
|
float[batch,1024,1024] add_708
|
|
float[batch,1024,1024] add_737
|
|
float[batch,1024,1024] add_78
|
|
float[batch,1024,1024] add_798
|
|
float[batch,1024,1024] add_827
|
|
float[batch,1024,1024] add_888
|
|
float[batch,1024,1024] add_917
|
|
float[batch,1024,1024] add_978
|
|
float[batch,1] clamp_min
|
|
float[batch,1024,32,32] conv2d
|
|
float[batch,1024,4096] gelu
|
|
float[batch,1024,4096] gelu_1
|
|
float[batch,1024,4096] gelu_10
|
|
float[batch,1024,4096] gelu_11
|
|
float[batch,1024,4096] gelu_12
|
|
float[batch,1024,4096] gelu_13
|
|
float[batch,1024,4096] gelu_14
|
|
float[batch,1024,4096] gelu_15
|
|
float[batch,1024,4096] gelu_16
|
|
float[batch,1024,4096] gelu_17
|
|
float[batch,1024,4096] gelu_18
|
|
float[batch,1024,4096] gelu_19
|
|
float[batch,1024,4096] gelu_2
|
|
float[batch,1024,4096] gelu_20
|
|
float[batch,1024,4096] gelu_21
|
|
float[batch,1024,4096] gelu_22
|
|
float[batch,1024,4096] gelu_23
|
|
float[batch,1,4096] gelu_24
|
|
float[batch,1024,4096] gelu_3
|
|
float[batch,1024,4096] gelu_4
|
|
float[batch,1024,4096] gelu_5
|
|
float[batch,1024,4096] gelu_6
|
|
float[batch,1024,4096] gelu_7
|
|
float[batch,1024,4096] gelu_8
|
|
float[batch,1024,4096] gelu_9
|
|
float[batch,3,512,512] image_chw
|
|
float[batch] image_ez
|
|
float[batch] image_ez_r
|
|
float[batch,1,1,1] image_ez_s
|
|
float[batch,512,512,3] image_f32
|
|
float[batch,1024,1024] layer_norm
|
|
float[batch,1024,1024] layer_norm_1
|
|
float[batch,1024,1024] layer_norm_10
|
|
float[batch,1024,1024] layer_norm_11
|
|
float[batch,1024,1024] layer_norm_12
|
|
float[batch,1024,1024] layer_norm_13
|
|
float[batch,1024,1024] layer_norm_14
|
|
float[batch,1024,1024] layer_norm_15
|
|
float[batch,1024,1024] layer_norm_16
|
|
float[batch,1024,1024] layer_norm_17
|
|
float[batch,1024,1024] layer_norm_18
|
|
float[batch,1024,1024] layer_norm_19
|
|
float[batch,1024,1024] layer_norm_2
|
|
float[batch,1024,1024] layer_norm_20
|
|
float[batch,1024,1024] layer_norm_21
|
|
float[batch,1024,1024] layer_norm_22
|
|
float[batch,1024,1024] layer_norm_23
|
|
float[batch,1024,1024] layer_norm_24
|
|
float[batch,1024,1024] layer_norm_25
|
|
float[batch,1024,1024] layer_norm_26
|
|
float[batch,1024,1024] layer_norm_27
|
|
float[batch,1024,1024] layer_norm_28
|
|
float[batch,1024,1024] layer_norm_29
|
|
float[batch,1024,1024] layer_norm_3
|
|
float[batch,1024,1024] layer_norm_30
|
|
float[batch,1024,1024] layer_norm_31
|
|
float[batch,1024,1024] layer_norm_32
|
|
float[batch,1024,1024] layer_norm_33
|
|
float[batch,1024,1024] layer_norm_34
|
|
float[batch,1024,1024] layer_norm_35
|
|
float[batch,1024,1024] layer_norm_36
|
|
float[batch,1024,1024] layer_norm_37
|
|
float[batch,1024,1024] layer_norm_38
|
|
float[batch,1024,1024] layer_norm_39
|
|
float[batch,1024,1024] layer_norm_4
|
|
float[batch,1024,1024] layer_norm_40
|
|
float[batch,1024,1024] layer_norm_41
|
|
float[batch,1024,1024] layer_norm_42
|
|
float[batch,1024,1024] layer_norm_43
|
|
float[batch,1024,1024] layer_norm_44
|
|
float[batch,1024,1024] layer_norm_45
|
|
float[batch,1024,1024] layer_norm_46
|
|
float[batch,1024,1024] layer_norm_47
|
|
float[batch,1024,1024] layer_norm_48
|
|
float[batch,1,1024] layer_norm_49
|
|
float[batch,1024,1024] layer_norm_5
|
|
float[batch,1024,1024] layer_norm_6
|
|
float[batch,1024,1024] layer_norm_7
|
|
float[batch,1024,1024] layer_norm_8
|
|
float[batch,1024,1024] layer_norm_9
|
|
float[batch,1] linalg_vector_norm
|
|
float[batch,1024,3072] linear
|
|
float[batch,1024,1024] linear_1
|
|
float[batch,1024,4096] linear_10
|
|
float[batch,1,1024] linear_100
|
|
float[batch,1024,1024] linear_11
|
|
float[batch,1024,3072] linear_12
|
|
float[batch,1024,1024] linear_13
|
|
float[batch,1024,4096] linear_14
|
|
float[batch,1024,1024] linear_15
|
|
float[batch,1024,3072] linear_16
|
|
float[batch,1024,1024] linear_17
|
|
float[batch,1024,4096] linear_18
|
|
float[batch,1024,1024] linear_19
|
|
float[batch,1024,4096] linear_2
|
|
float[batch,1024,3072] linear_20
|
|
float[batch,1024,1024] linear_21
|
|
float[batch,1024,4096] linear_22
|
|
float[batch,1024,1024] linear_23
|
|
float[batch,1024,3072] linear_24
|
|
float[batch,1024,1024] linear_25
|
|
float[batch,1024,4096] linear_26
|
|
float[batch,1024,1024] linear_27
|
|
float[batch,1024,3072] linear_28
|
|
float[batch,1024,1024] linear_29
|
|
float[batch,1024,1024] linear_3
|
|
float[batch,1024,4096] linear_30
|
|
float[batch,1024,1024] linear_31
|
|
float[batch,1024,3072] linear_32
|
|
float[batch,1024,1024] linear_33
|
|
float[batch,1024,4096] linear_34
|
|
float[batch,1024,1024] linear_35
|
|
float[batch,1024,3072] linear_36
|
|
float[batch,1024,1024] linear_37
|
|
float[batch,1024,4096] linear_38
|
|
float[batch,1024,1024] linear_39
|
|
float[batch,1024,3072] linear_4
|
|
float[batch,1024,3072] linear_40
|
|
float[batch,1024,1024] linear_41
|
|
float[batch,1024,4096] linear_42
|
|
float[batch,1024,1024] linear_43
|
|
float[batch,1024,3072] linear_44
|
|
float[batch,1024,1024] linear_45
|
|
float[batch,1024,4096] linear_46
|
|
float[batch,1024,1024] linear_47
|
|
float[batch,1024,3072] linear_48
|
|
float[batch,1024,1024] linear_49
|
|
float[batch,1024,1024] linear_5
|
|
float[batch,1024,4096] linear_50
|
|
float[batch,1024,1024] linear_51
|
|
float[batch,1024,3072] linear_52
|
|
float[batch,1024,1024] linear_53
|
|
float[batch,1024,4096] linear_54
|
|
float[batch,1024,1024] linear_55
|
|
float[batch,1024,3072] linear_56
|
|
float[batch,1024,1024] linear_57
|
|
float[batch,1024,4096] linear_58
|
|
float[batch,1024,1024] linear_59
|
|
float[batch,1024,4096] linear_6
|
|
float[batch,1024,3072] linear_60
|
|
float[batch,1024,1024] linear_61
|
|
float[batch,1024,4096] linear_62
|
|
float[batch,1024,1024] linear_63
|
|
float[batch,1024,3072] linear_64
|
|
float[batch,1024,1024] linear_65
|
|
float[batch,1024,4096] linear_66
|
|
float[batch,1024,1024] linear_67
|
|
float[batch,1024,3072] linear_68
|
|
float[batch,1024,1024] linear_69
|
|
float[batch,1024,1024] linear_7
|
|
float[batch,1024,4096] linear_70
|
|
float[batch,1024,1024] linear_71
|
|
float[batch,1024,3072] linear_72
|
|
float[batch,1024,1024] linear_73
|
|
float[batch,1024,4096] linear_74
|
|
float[batch,1024,1024] linear_75
|
|
float[batch,1024,3072] linear_76
|
|
float[batch,1024,1024] linear_77
|
|
float[batch,1024,4096] linear_78
|
|
float[batch,1024,1024] linear_79
|
|
float[batch,1024,3072] linear_8
|
|
float[batch,1024,3072] linear_80
|
|
float[batch,1024,1024] linear_81
|
|
float[batch,1024,4096] linear_82
|
|
float[batch,1024,1024] linear_83
|
|
float[batch,1024,3072] linear_84
|
|
float[batch,1024,1024] linear_85
|
|
float[batch,1024,4096] linear_86
|
|
float[batch,1024,1024] linear_87
|
|
float[batch,1024,3072] linear_88
|
|
float[batch,1024,1024] linear_89
|
|
float[batch,1024,1024] linear_9
|
|
float[batch,1024,4096] linear_90
|
|
float[batch,1024,1024] linear_91
|
|
float[batch,1024,3072] linear_92
|
|
float[batch,1024,1024] linear_93
|
|
float[batch,1024,4096] linear_94
|
|
float[batch,1024,1024] linear_95
|
|
float[batch,1024,2048] linear_97
|
|
float[batch,1,1024] linear_98
|
|
float[batch,1,4096] linear_99
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_10_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_10_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_10_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_11_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_11_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_11_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_12_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_12_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_12_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_13_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_13_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_13_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_14_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_14_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_14_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_15_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_15_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_15_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_16_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_16_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_16_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_17_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_17_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_17_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_18_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_18_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_18_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_19_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_19_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_19_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_1_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_1_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_1_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_20_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_20_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_20_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_21_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_21_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_21_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_22_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_22_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_22_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_23_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_23_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_23_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_24_k
|
|
float[batch,1,1024] node_scaled_dot_product_attention_24_q
|
|
float[batch,1,1] node_scaled_dot_product_attention_24_q_col_out
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_24_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_2_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_2_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_2_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_3_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_3_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_3_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_4_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_4_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_4_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_5_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_5_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_5_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_6_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_6_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_6_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_7_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_7_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_7_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_8_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_8_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_8_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_9_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_9_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_9_v
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_k
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_q
|
|
float[batch,1024,1024] node_scaled_dot_product_attention_v
|
|
float[batch,1024,1024] scaled_dot_product_attention
|
|
float[batch,1024,1024] scaled_dot_product_attention_1
|
|
float[batch,1024,1024] scaled_dot_product_attention_10
|
|
float[batch,1024,1024] scaled_dot_product_attention_11
|
|
float[batch,1024,1024] scaled_dot_product_attention_12
|
|
float[batch,1024,1024] scaled_dot_product_attention_13
|
|
float[batch,1024,1024] scaled_dot_product_attention_14
|
|
float[batch,1024,1024] scaled_dot_product_attention_15
|
|
float[batch,1024,1024] scaled_dot_product_attention_16
|
|
float[batch,1024,1024] scaled_dot_product_attention_17
|
|
float[batch,1024,1024] scaled_dot_product_attention_18
|
|
float[batch,1024,1024] scaled_dot_product_attention_19
|
|
float[batch,1024,1024] scaled_dot_product_attention_2
|
|
float[batch,1024,1024] scaled_dot_product_attention_20
|
|
float[batch,1024,1024] scaled_dot_product_attention_21
|
|
float[batch,1024,1024] scaled_dot_product_attention_22
|
|
float[batch,1024,1024] scaled_dot_product_attention_23
|
|
float[batch,1,1024] scaled_dot_product_attention_24
|
|
float[batch,1024,1024] scaled_dot_product_attention_3
|
|
float[batch,1024,1024] scaled_dot_product_attention_4
|
|
float[batch,1024,1024] scaled_dot_product_attention_5
|
|
float[batch,1024,1024] scaled_dot_product_attention_6
|
|
float[batch,1024,1024] scaled_dot_product_attention_7
|
|
float[batch,1024,1024] scaled_dot_product_attention_8
|
|
float[batch,1024,1024] scaled_dot_product_attention_9
|
|
float[batch,1024] select
|
|
float[batch,1024,1024] transpose
|
|
float[batch,1024,3072] val_103
|
|
float[batch,1024,1024] val_104
|
|
float[batch,1024,4096] val_105
|
|
float[batch,1024,1024] val_106
|
|
float[batch,1024,3072] val_107
|
|
float[batch,1024,1024] val_108
|
|
float[batch,1024,4096] val_109
|
|
float[batch,1024,1024] val_110
|
|
float[batch,1024,3072] val_111
|
|
float[batch,1024,1024] val_112
|
|
float[batch,1024,4096] val_113
|
|
float[batch,1024,1024] val_114
|
|
float[batch,1024,3072] val_115
|
|
float[batch,1024,1024] val_116
|
|
float[batch,1024,4096] val_117
|
|
float[batch,1024,1024] val_118
|
|
float[batch,1024,3072] val_119
|
|
float[batch,1024,1024] val_120
|
|
float[batch,1024,4096] val_121
|
|
float[batch,1024,1024] val_122
|
|
float[batch,1024,3072] val_123
|
|
float[batch,1024,1024] val_124
|
|
float[batch,1024,4096] val_125
|
|
float[batch,1024,1024] val_126
|
|
float[batch,1024,3072] val_127
|
|
float[batch,1024,1024] val_128
|
|
float[batch,1024,4096] val_129
|
|
float[batch,1024,1024] val_130
|
|
float[batch,1024,3072] val_131
|
|
float[batch,1024,1024] val_132
|
|
float[batch,1024,4096] val_133
|
|
float[batch,1024,1024] val_134
|
|
float[batch,1024,3072] val_135
|
|
float[batch,1024,1024] val_136
|
|
float[batch,1024,4096] val_137
|
|
float[batch,1024,1024] val_138
|
|
float[batch,1024,3072] val_139
|
|
float[batch,1024,1024] val_140
|
|
float[batch,1024,4096] val_141
|
|
float[batch,1024,1024] val_142
|
|
float[batch,1024,3072] val_143
|
|
float[batch,1024,1024] val_144
|
|
float[batch,1024,4096] val_145
|
|
float[batch,1024,1024] val_146
|
|
float[batch,1024,3072] val_147
|
|
float[batch,1024,1024] val_148
|
|
float[batch,1024,4096] val_149
|
|
float[batch,1024,1024] val_150
|
|
float[batch,1024,3072] val_151
|
|
float[batch,1024,1024] val_152
|
|
float[batch,1024,4096] val_153
|
|
float[batch,1024,1024] val_154
|
|
float[batch,1024,3072] val_155
|
|
float[batch,1024,1024] val_156
|
|
float[batch,1024,4096] val_157
|
|
float[batch,1024,1024] val_158
|
|
float[batch,1024,3072] val_159
|
|
float[batch,1024,1024] val_160
|
|
float[batch,1024,4096] val_161
|
|
float[batch,1024,1024] val_162
|
|
float[batch,1024,3072] val_163
|
|
float[batch,1024,1024] val_164
|
|
float[batch,1024,4096] val_165
|
|
float[batch,1024,1024] val_166
|
|
float[batch,1024,3072] val_167
|
|
float[batch,1024,1024] val_168
|
|
float[batch,1024,4096] val_169
|
|
float[batch,1024,1024] val_170
|
|
float[batch,1024,3072] val_171
|
|
float[batch,1024,1024] val_172
|
|
float[batch,1024,4096] val_173
|
|
float[batch,1024,1024] val_174
|
|
float[batch,1024,3072] val_175
|
|
float[batch,1024,1024] val_176
|
|
float[batch,1024,4096] val_177
|
|
float[batch,1024,1024] val_178
|
|
float[batch,1024,3072] val_179
|
|
float[batch,1024,1024] val_180
|
|
float[batch,1024,4096] val_181
|
|
float[batch,1024,1024] val_182
|
|
float[batch,1024,3072] val_183
|
|
float[batch,1024,1024] val_184
|
|
float[batch,1024,4096] val_185
|
|
float[batch,1024,1024] val_186
|
|
float[batch,1024,3072] val_187
|
|
float[batch,1024,1024] val_188
|
|
float[batch,1024,4096] val_189
|
|
float[batch,1024,1024] val_190
|
|
float[batch,1024,3072] val_191
|
|
float[batch,1024,1024] val_192
|
|
float[batch,1024,4096] val_193
|
|
float[batch,1024,1024] val_194
|
|
float[batch,1024,3072] val_195
|
|
float[batch,1024,1024] val_196
|
|
float[batch,1024,4096] val_197
|
|
float[batch,1024,1024] val_198
|
|
float[batch,1024,2048] val_199
|
|
float[batch,1,1024] val_200
|
|
float[batch,1,4096] val_201
|
|
float[batch,1,1024] val_202
|
|
float[batch,1024,1024] view
|
|
>
|
|
{
|
|
[pre_cast] image_f32 = Cast <to: int = 1> (image)
|
|
[pre_nhwc_to_nchw] image_chw = Transpose <perm: ints = [0, 3, 1, 2]> (image_f32)
|
|
[node_conv2d] conv2d = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, pads: ints = [0, 0, 0, 0], strides: ints = [16, 16]> (image_chw, "visual.trunk.patch_embed.proj.weight", "visual.trunk.patch_embed.proj.bias")
|
|
view = Reshape <allowzero: int = 1> (conv2d, view_target)
|
|
[node_transpose] transpose = Transpose <perm: ints = [0, 2, 1]> (view)
|
|
add_13 = Add (transpose, "visual.trunk.pos_embed")
|
|
[node_layer_norm] layer_norm = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_13, "visual.trunk.blocks.0.norm1.weight", "visual.trunk.blocks.0.norm1.bias")
|
|
val_103 = MatMul (layer_norm, val_3)
|
|
[node_linear] linear = Add (val_103, "visual.trunk.blocks.0.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_qkv_split] node_scaled_dot_product_attention_q, node_scaled_dot_product_attention_k, node_scaled_dot_product_attention_v = Split <axis: int = -1> (linear, attn3d_split_3x1024)
|
|
[node_scaled_dot_product_attention] scaled_dot_product_attention = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_q, node_scaled_dot_product_attention_k, node_scaled_dot_product_attention_v)
|
|
val_104 = MatMul (scaled_dot_product_attention, val_4)
|
|
linear_1 = Add (val_104, "visual.trunk.blocks.0.attn.proj.bias")
|
|
add_78 = Add (add_13, linear_1)
|
|
layer_norm_1 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_78, "visual.trunk.blocks.0.norm2.weight", "visual.trunk.blocks.0.norm2.bias")
|
|
val_105 = MatMul (layer_norm_1, val_5)
|
|
linear_2 = Add (val_105, "visual.trunk.blocks.0.mlp.fc1.bias")
|
|
[node_gelu] gelu = Gelu <approximate: string = "tanh"> (linear_2)
|
|
val_106 = MatMul (gelu, val_6)
|
|
linear_3 = Add (val_106, "visual.trunk.blocks.0.mlp.fc2.bias")
|
|
add_107 = Add (add_78, linear_3)
|
|
layer_norm_2 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_107, "visual.trunk.blocks.1.norm1.weight", "visual.trunk.blocks.1.norm1.bias")
|
|
val_107 = MatMul (layer_norm_2, val_7)
|
|
linear_4 = Add (val_107, "visual.trunk.blocks.1.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_1_qkv_split] node_scaled_dot_product_attention_1_q, node_scaled_dot_product_attention_1_k, node_scaled_dot_product_attention_1_v = Split <axis: int = -1> (linear_4, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_1 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_1_q, node_scaled_dot_product_attention_1_k, node_scaled_dot_product_attention_1_v)
|
|
val_108 = MatMul (scaled_dot_product_attention_1, val_8)
|
|
linear_5 = Add (val_108, "visual.trunk.blocks.1.attn.proj.bias")
|
|
add_168 = Add (add_107, linear_5)
|
|
layer_norm_3 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_168, "visual.trunk.blocks.1.norm2.weight", "visual.trunk.blocks.1.norm2.bias")
|
|
val_109 = MatMul (layer_norm_3, val_9)
|
|
linear_6 = Add (val_109, "visual.trunk.blocks.1.mlp.fc1.bias")
|
|
gelu_1 = Gelu <approximate: string = "tanh"> (linear_6)
|
|
val_110 = MatMul (gelu_1, val_10)
|
|
linear_7 = Add (val_110, "visual.trunk.blocks.1.mlp.fc2.bias")
|
|
add_197 = Add (add_168, linear_7)
|
|
layer_norm_4 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_197, "visual.trunk.blocks.2.norm1.weight", "visual.trunk.blocks.2.norm1.bias")
|
|
val_111 = MatMul (layer_norm_4, val_11)
|
|
linear_8 = Add (val_111, "visual.trunk.blocks.2.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_2_qkv_split] node_scaled_dot_product_attention_2_q, node_scaled_dot_product_attention_2_k, node_scaled_dot_product_attention_2_v = Split <axis: int = -1> (linear_8, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_2 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_2_q, node_scaled_dot_product_attention_2_k, node_scaled_dot_product_attention_2_v)
|
|
val_112 = MatMul (scaled_dot_product_attention_2, val_12)
|
|
linear_9 = Add (val_112, "visual.trunk.blocks.2.attn.proj.bias")
|
|
add_258 = Add (add_197, linear_9)
|
|
layer_norm_5 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_258, "visual.trunk.blocks.2.norm2.weight", "visual.trunk.blocks.2.norm2.bias")
|
|
val_113 = MatMul (layer_norm_5, val_13)
|
|
linear_10 = Add (val_113, "visual.trunk.blocks.2.mlp.fc1.bias")
|
|
gelu_2 = Gelu <approximate: string = "tanh"> (linear_10)
|
|
val_114 = MatMul (gelu_2, val_14)
|
|
linear_11 = Add (val_114, "visual.trunk.blocks.2.mlp.fc2.bias")
|
|
add_287 = Add (add_258, linear_11)
|
|
layer_norm_6 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_287, "visual.trunk.blocks.3.norm1.weight", "visual.trunk.blocks.3.norm1.bias")
|
|
val_115 = MatMul (layer_norm_6, val_15)
|
|
linear_12 = Add (val_115, "visual.trunk.blocks.3.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_3_qkv_split] node_scaled_dot_product_attention_3_q, node_scaled_dot_product_attention_3_k, node_scaled_dot_product_attention_3_v = Split <axis: int = -1> (linear_12, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_3 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_3_q, node_scaled_dot_product_attention_3_k, node_scaled_dot_product_attention_3_v)
|
|
val_116 = MatMul (scaled_dot_product_attention_3, val_16)
|
|
linear_13 = Add (val_116, "visual.trunk.blocks.3.attn.proj.bias")
|
|
add_348 = Add (add_287, linear_13)
|
|
layer_norm_7 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_348, "visual.trunk.blocks.3.norm2.weight", "visual.trunk.blocks.3.norm2.bias")
|
|
val_117 = MatMul (layer_norm_7, val_17)
|
|
linear_14 = Add (val_117, "visual.trunk.blocks.3.mlp.fc1.bias")
|
|
gelu_3 = Gelu <approximate: string = "tanh"> (linear_14)
|
|
val_118 = MatMul (gelu_3, val_18)
|
|
linear_15 = Add (val_118, "visual.trunk.blocks.3.mlp.fc2.bias")
|
|
add_377 = Add (add_348, linear_15)
|
|
layer_norm_8 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_377, "visual.trunk.blocks.4.norm1.weight", "visual.trunk.blocks.4.norm1.bias")
|
|
val_119 = MatMul (layer_norm_8, val_19)
|
|
linear_16 = Add (val_119, "visual.trunk.blocks.4.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_4_qkv_split] node_scaled_dot_product_attention_4_q, node_scaled_dot_product_attention_4_k, node_scaled_dot_product_attention_4_v = Split <axis: int = -1> (linear_16, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_4 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_4_q, node_scaled_dot_product_attention_4_k, node_scaled_dot_product_attention_4_v)
|
|
val_120 = MatMul (scaled_dot_product_attention_4, val_20)
|
|
linear_17 = Add (val_120, "visual.trunk.blocks.4.attn.proj.bias")
|
|
add_438 = Add (add_377, linear_17)
|
|
layer_norm_9 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_438, "visual.trunk.blocks.4.norm2.weight", "visual.trunk.blocks.4.norm2.bias")
|
|
val_121 = MatMul (layer_norm_9, val_21)
|
|
linear_18 = Add (val_121, "visual.trunk.blocks.4.mlp.fc1.bias")
|
|
gelu_4 = Gelu <approximate: string = "tanh"> (linear_18)
|
|
val_122 = MatMul (gelu_4, val_22)
|
|
linear_19 = Add (val_122, "visual.trunk.blocks.4.mlp.fc2.bias")
|
|
add_467 = Add (add_438, linear_19)
|
|
layer_norm_10 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_467, "visual.trunk.blocks.5.norm1.weight", "visual.trunk.blocks.5.norm1.bias")
|
|
val_123 = MatMul (layer_norm_10, val_23)
|
|
linear_20 = Add (val_123, "visual.trunk.blocks.5.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_5_qkv_split] node_scaled_dot_product_attention_5_q, node_scaled_dot_product_attention_5_k, node_scaled_dot_product_attention_5_v = Split <axis: int = -1> (linear_20, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_5 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_5_q, node_scaled_dot_product_attention_5_k, node_scaled_dot_product_attention_5_v)
|
|
val_124 = MatMul (scaled_dot_product_attention_5, val_24)
|
|
linear_21 = Add (val_124, "visual.trunk.blocks.5.attn.proj.bias")
|
|
add_528 = Add (add_467, linear_21)
|
|
layer_norm_11 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_528, "visual.trunk.blocks.5.norm2.weight", "visual.trunk.blocks.5.norm2.bias")
|
|
val_125 = MatMul (layer_norm_11, val_25)
|
|
linear_22 = Add (val_125, "visual.trunk.blocks.5.mlp.fc1.bias")
|
|
gelu_5 = Gelu <approximate: string = "tanh"> (linear_22)
|
|
val_126 = MatMul (gelu_5, val_26)
|
|
linear_23 = Add (val_126, "visual.trunk.blocks.5.mlp.fc2.bias")
|
|
add_557 = Add (add_528, linear_23)
|
|
layer_norm_12 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_557, "visual.trunk.blocks.6.norm1.weight", "visual.trunk.blocks.6.norm1.bias")
|
|
val_127 = MatMul (layer_norm_12, val_27)
|
|
linear_24 = Add (val_127, "visual.trunk.blocks.6.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_6_qkv_split] node_scaled_dot_product_attention_6_q, node_scaled_dot_product_attention_6_k, node_scaled_dot_product_attention_6_v = Split <axis: int = -1> (linear_24, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_6 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_6_q, node_scaled_dot_product_attention_6_k, node_scaled_dot_product_attention_6_v)
|
|
val_128 = MatMul (scaled_dot_product_attention_6, val_28)
|
|
linear_25 = Add (val_128, "visual.trunk.blocks.6.attn.proj.bias")
|
|
add_618 = Add (add_557, linear_25)
|
|
layer_norm_13 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_618, "visual.trunk.blocks.6.norm2.weight", "visual.trunk.blocks.6.norm2.bias")
|
|
val_129 = MatMul (layer_norm_13, val_29)
|
|
linear_26 = Add (val_129, "visual.trunk.blocks.6.mlp.fc1.bias")
|
|
gelu_6 = Gelu <approximate: string = "tanh"> (linear_26)
|
|
val_130 = MatMul (gelu_6, val_30)
|
|
linear_27 = Add (val_130, "visual.trunk.blocks.6.mlp.fc2.bias")
|
|
add_647 = Add (add_618, linear_27)
|
|
layer_norm_14 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_647, "visual.trunk.blocks.7.norm1.weight", "visual.trunk.blocks.7.norm1.bias")
|
|
val_131 = MatMul (layer_norm_14, val_31)
|
|
linear_28 = Add (val_131, "visual.trunk.blocks.7.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_7_qkv_split] node_scaled_dot_product_attention_7_q, node_scaled_dot_product_attention_7_k, node_scaled_dot_product_attention_7_v = Split <axis: int = -1> (linear_28, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_7 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_7_q, node_scaled_dot_product_attention_7_k, node_scaled_dot_product_attention_7_v)
|
|
val_132 = MatMul (scaled_dot_product_attention_7, val_32)
|
|
linear_29 = Add (val_132, "visual.trunk.blocks.7.attn.proj.bias")
|
|
add_708 = Add (add_647, linear_29)
|
|
layer_norm_15 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_708, "visual.trunk.blocks.7.norm2.weight", "visual.trunk.blocks.7.norm2.bias")
|
|
val_133 = MatMul (layer_norm_15, val_33)
|
|
linear_30 = Add (val_133, "visual.trunk.blocks.7.mlp.fc1.bias")
|
|
gelu_7 = Gelu <approximate: string = "tanh"> (linear_30)
|
|
val_134 = MatMul (gelu_7, val_34)
|
|
linear_31 = Add (val_134, "visual.trunk.blocks.7.mlp.fc2.bias")
|
|
add_737 = Add (add_708, linear_31)
|
|
layer_norm_16 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_737, "visual.trunk.blocks.8.norm1.weight", "visual.trunk.blocks.8.norm1.bias")
|
|
val_135 = MatMul (layer_norm_16, val_35)
|
|
linear_32 = Add (val_135, "visual.trunk.blocks.8.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_8_qkv_split] node_scaled_dot_product_attention_8_q, node_scaled_dot_product_attention_8_k, node_scaled_dot_product_attention_8_v = Split <axis: int = -1> (linear_32, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_8 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_8_q, node_scaled_dot_product_attention_8_k, node_scaled_dot_product_attention_8_v)
|
|
val_136 = MatMul (scaled_dot_product_attention_8, val_36)
|
|
linear_33 = Add (val_136, "visual.trunk.blocks.8.attn.proj.bias")
|
|
add_798 = Add (add_737, linear_33)
|
|
layer_norm_17 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_798, "visual.trunk.blocks.8.norm2.weight", "visual.trunk.blocks.8.norm2.bias")
|
|
val_137 = MatMul (layer_norm_17, val_37)
|
|
linear_34 = Add (val_137, "visual.trunk.blocks.8.mlp.fc1.bias")
|
|
gelu_8 = Gelu <approximate: string = "tanh"> (linear_34)
|
|
val_138 = MatMul (gelu_8, val_38)
|
|
linear_35 = Add (val_138, "visual.trunk.blocks.8.mlp.fc2.bias")
|
|
add_827 = Add (add_798, linear_35)
|
|
layer_norm_18 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_827, "visual.trunk.blocks.9.norm1.weight", "visual.trunk.blocks.9.norm1.bias")
|
|
val_139 = MatMul (layer_norm_18, val_39)
|
|
linear_36 = Add (val_139, "visual.trunk.blocks.9.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_9_qkv_split] node_scaled_dot_product_attention_9_q, node_scaled_dot_product_attention_9_k, node_scaled_dot_product_attention_9_v = Split <axis: int = -1> (linear_36, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_9 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_9_q, node_scaled_dot_product_attention_9_k, node_scaled_dot_product_attention_9_v)
|
|
val_140 = MatMul (scaled_dot_product_attention_9, val_40)
|
|
linear_37 = Add (val_140, "visual.trunk.blocks.9.attn.proj.bias")
|
|
add_888 = Add (add_827, linear_37)
|
|
layer_norm_19 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_888, "visual.trunk.blocks.9.norm2.weight", "visual.trunk.blocks.9.norm2.bias")
|
|
val_141 = MatMul (layer_norm_19, val_41)
|
|
linear_38 = Add (val_141, "visual.trunk.blocks.9.mlp.fc1.bias")
|
|
gelu_9 = Gelu <approximate: string = "tanh"> (linear_38)
|
|
val_142 = MatMul (gelu_9, val_42)
|
|
linear_39 = Add (val_142, "visual.trunk.blocks.9.mlp.fc2.bias")
|
|
add_917 = Add (add_888, linear_39)
|
|
layer_norm_20 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_917, "visual.trunk.blocks.10.norm1.weight", "visual.trunk.blocks.10.norm1.bias")
|
|
val_143 = MatMul (layer_norm_20, val_43)
|
|
linear_40 = Add (val_143, "visual.trunk.blocks.10.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_10_qkv_split] node_scaled_dot_product_attention_10_q, node_scaled_dot_product_attention_10_k, node_scaled_dot_product_attention_10_v = Split <axis: int = -1> (linear_40, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_10 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_10_q, node_scaled_dot_product_attention_10_k, node_scaled_dot_product_attention_10_v)
|
|
val_144 = MatMul (scaled_dot_product_attention_10, val_44)
|
|
linear_41 = Add (val_144, "visual.trunk.blocks.10.attn.proj.bias")
|
|
add_978 = Add (add_917, linear_41)
|
|
layer_norm_21 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_978, "visual.trunk.blocks.10.norm2.weight", "visual.trunk.blocks.10.norm2.bias")
|
|
val_145 = MatMul (layer_norm_21, val_45)
|
|
linear_42 = Add (val_145, "visual.trunk.blocks.10.mlp.fc1.bias")
|
|
gelu_10 = Gelu <approximate: string = "tanh"> (linear_42)
|
|
val_146 = MatMul (gelu_10, val_46)
|
|
linear_43 = Add (val_146, "visual.trunk.blocks.10.mlp.fc2.bias")
|
|
add_1007 = Add (add_978, linear_43)
|
|
layer_norm_22 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1007, "visual.trunk.blocks.11.norm1.weight", "visual.trunk.blocks.11.norm1.bias")
|
|
val_147 = MatMul (layer_norm_22, val_47)
|
|
linear_44 = Add (val_147, "visual.trunk.blocks.11.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_11_qkv_split] node_scaled_dot_product_attention_11_q, node_scaled_dot_product_attention_11_k, node_scaled_dot_product_attention_11_v = Split <axis: int = -1> (linear_44, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_11 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_11_q, node_scaled_dot_product_attention_11_k, node_scaled_dot_product_attention_11_v)
|
|
val_148 = MatMul (scaled_dot_product_attention_11, val_48)
|
|
linear_45 = Add (val_148, "visual.trunk.blocks.11.attn.proj.bias")
|
|
add_1068 = Add (add_1007, linear_45)
|
|
layer_norm_23 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1068, "visual.trunk.blocks.11.norm2.weight", "visual.trunk.blocks.11.norm2.bias")
|
|
val_149 = MatMul (layer_norm_23, val_49)
|
|
linear_46 = Add (val_149, "visual.trunk.blocks.11.mlp.fc1.bias")
|
|
gelu_11 = Gelu <approximate: string = "tanh"> (linear_46)
|
|
val_150 = MatMul (gelu_11, val_50)
|
|
linear_47 = Add (val_150, "visual.trunk.blocks.11.mlp.fc2.bias")
|
|
add_1097 = Add (add_1068, linear_47)
|
|
layer_norm_24 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1097, "visual.trunk.blocks.12.norm1.weight", "visual.trunk.blocks.12.norm1.bias")
|
|
val_151 = MatMul (layer_norm_24, val_51)
|
|
linear_48 = Add (val_151, "visual.trunk.blocks.12.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_12_qkv_split] node_scaled_dot_product_attention_12_q, node_scaled_dot_product_attention_12_k, node_scaled_dot_product_attention_12_v = Split <axis: int = -1> (linear_48, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_12 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_12_q, node_scaled_dot_product_attention_12_k, node_scaled_dot_product_attention_12_v)
|
|
val_152 = MatMul (scaled_dot_product_attention_12, val_52)
|
|
linear_49 = Add (val_152, "visual.trunk.blocks.12.attn.proj.bias")
|
|
add_1158 = Add (add_1097, linear_49)
|
|
layer_norm_25 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1158, "visual.trunk.blocks.12.norm2.weight", "visual.trunk.blocks.12.norm2.bias")
|
|
val_153 = MatMul (layer_norm_25, val_53)
|
|
linear_50 = Add (val_153, "visual.trunk.blocks.12.mlp.fc1.bias")
|
|
gelu_12 = Gelu <approximate: string = "tanh"> (linear_50)
|
|
val_154 = MatMul (gelu_12, val_54)
|
|
linear_51 = Add (val_154, "visual.trunk.blocks.12.mlp.fc2.bias")
|
|
add_1187 = Add (add_1158, linear_51)
|
|
layer_norm_26 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1187, "visual.trunk.blocks.13.norm1.weight", "visual.trunk.blocks.13.norm1.bias")
|
|
val_155 = MatMul (layer_norm_26, val_55)
|
|
linear_52 = Add (val_155, "visual.trunk.blocks.13.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_13_qkv_split] node_scaled_dot_product_attention_13_q, node_scaled_dot_product_attention_13_k, node_scaled_dot_product_attention_13_v = Split <axis: int = -1> (linear_52, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_13 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_13_q, node_scaled_dot_product_attention_13_k, node_scaled_dot_product_attention_13_v)
|
|
val_156 = MatMul (scaled_dot_product_attention_13, val_56)
|
|
linear_53 = Add (val_156, "visual.trunk.blocks.13.attn.proj.bias")
|
|
add_1248 = Add (add_1187, linear_53)
|
|
layer_norm_27 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1248, "visual.trunk.blocks.13.norm2.weight", "visual.trunk.blocks.13.norm2.bias")
|
|
val_157 = MatMul (layer_norm_27, val_57)
|
|
linear_54 = Add (val_157, "visual.trunk.blocks.13.mlp.fc1.bias")
|
|
gelu_13 = Gelu <approximate: string = "tanh"> (linear_54)
|
|
val_158 = MatMul (gelu_13, val_58)
|
|
linear_55 = Add (val_158, "visual.trunk.blocks.13.mlp.fc2.bias")
|
|
add_1277 = Add (add_1248, linear_55)
|
|
layer_norm_28 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1277, "visual.trunk.blocks.14.norm1.weight", "visual.trunk.blocks.14.norm1.bias")
|
|
val_159 = MatMul (layer_norm_28, val_59)
|
|
linear_56 = Add (val_159, "visual.trunk.blocks.14.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_14_qkv_split] node_scaled_dot_product_attention_14_q, node_scaled_dot_product_attention_14_k, node_scaled_dot_product_attention_14_v = Split <axis: int = -1> (linear_56, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_14 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_14_q, node_scaled_dot_product_attention_14_k, node_scaled_dot_product_attention_14_v)
|
|
val_160 = MatMul (scaled_dot_product_attention_14, val_60)
|
|
linear_57 = Add (val_160, "visual.trunk.blocks.14.attn.proj.bias")
|
|
add_1338 = Add (add_1277, linear_57)
|
|
layer_norm_29 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1338, "visual.trunk.blocks.14.norm2.weight", "visual.trunk.blocks.14.norm2.bias")
|
|
val_161 = MatMul (layer_norm_29, val_61)
|
|
linear_58 = Add (val_161, "visual.trunk.blocks.14.mlp.fc1.bias")
|
|
gelu_14 = Gelu <approximate: string = "tanh"> (linear_58)
|
|
val_162 = MatMul (gelu_14, val_62)
|
|
linear_59 = Add (val_162, "visual.trunk.blocks.14.mlp.fc2.bias")
|
|
add_1367 = Add (add_1338, linear_59)
|
|
layer_norm_30 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1367, "visual.trunk.blocks.15.norm1.weight", "visual.trunk.blocks.15.norm1.bias")
|
|
val_163 = MatMul (layer_norm_30, val_63)
|
|
linear_60 = Add (val_163, "visual.trunk.blocks.15.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_15_qkv_split] node_scaled_dot_product_attention_15_q, node_scaled_dot_product_attention_15_k, node_scaled_dot_product_attention_15_v = Split <axis: int = -1> (linear_60, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_15 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_15_q, node_scaled_dot_product_attention_15_k, node_scaled_dot_product_attention_15_v)
|
|
val_164 = MatMul (scaled_dot_product_attention_15, val_64)
|
|
linear_61 = Add (val_164, "visual.trunk.blocks.15.attn.proj.bias")
|
|
add_1428 = Add (add_1367, linear_61)
|
|
layer_norm_31 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1428, "visual.trunk.blocks.15.norm2.weight", "visual.trunk.blocks.15.norm2.bias")
|
|
val_165 = MatMul (layer_norm_31, val_65)
|
|
linear_62 = Add (val_165, "visual.trunk.blocks.15.mlp.fc1.bias")
|
|
gelu_15 = Gelu <approximate: string = "tanh"> (linear_62)
|
|
val_166 = MatMul (gelu_15, val_66)
|
|
linear_63 = Add (val_166, "visual.trunk.blocks.15.mlp.fc2.bias")
|
|
add_1457 = Add (add_1428, linear_63)
|
|
layer_norm_32 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1457, "visual.trunk.blocks.16.norm1.weight", "visual.trunk.blocks.16.norm1.bias")
|
|
val_167 = MatMul (layer_norm_32, val_67)
|
|
linear_64 = Add (val_167, "visual.trunk.blocks.16.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_16_qkv_split] node_scaled_dot_product_attention_16_q, node_scaled_dot_product_attention_16_k, node_scaled_dot_product_attention_16_v = Split <axis: int = -1> (linear_64, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_16 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_16_q, node_scaled_dot_product_attention_16_k, node_scaled_dot_product_attention_16_v)
|
|
val_168 = MatMul (scaled_dot_product_attention_16, val_68)
|
|
linear_65 = Add (val_168, "visual.trunk.blocks.16.attn.proj.bias")
|
|
add_1518 = Add (add_1457, linear_65)
|
|
layer_norm_33 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1518, "visual.trunk.blocks.16.norm2.weight", "visual.trunk.blocks.16.norm2.bias")
|
|
val_169 = MatMul (layer_norm_33, val_69)
|
|
linear_66 = Add (val_169, "visual.trunk.blocks.16.mlp.fc1.bias")
|
|
gelu_16 = Gelu <approximate: string = "tanh"> (linear_66)
|
|
val_170 = MatMul (gelu_16, val_70)
|
|
linear_67 = Add (val_170, "visual.trunk.blocks.16.mlp.fc2.bias")
|
|
add_1547 = Add (add_1518, linear_67)
|
|
layer_norm_34 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1547, "visual.trunk.blocks.17.norm1.weight", "visual.trunk.blocks.17.norm1.bias")
|
|
val_171 = MatMul (layer_norm_34, val_71)
|
|
linear_68 = Add (val_171, "visual.trunk.blocks.17.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_17_qkv_split] node_scaled_dot_product_attention_17_q, node_scaled_dot_product_attention_17_k, node_scaled_dot_product_attention_17_v = Split <axis: int = -1> (linear_68, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_17 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_17_q, node_scaled_dot_product_attention_17_k, node_scaled_dot_product_attention_17_v)
|
|
val_172 = MatMul (scaled_dot_product_attention_17, val_72)
|
|
linear_69 = Add (val_172, "visual.trunk.blocks.17.attn.proj.bias")
|
|
add_1608 = Add (add_1547, linear_69)
|
|
layer_norm_35 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1608, "visual.trunk.blocks.17.norm2.weight", "visual.trunk.blocks.17.norm2.bias")
|
|
val_173 = MatMul (layer_norm_35, val_73)
|
|
linear_70 = Add (val_173, "visual.trunk.blocks.17.mlp.fc1.bias")
|
|
gelu_17 = Gelu <approximate: string = "tanh"> (linear_70)
|
|
val_174 = MatMul (gelu_17, val_74)
|
|
linear_71 = Add (val_174, "visual.trunk.blocks.17.mlp.fc2.bias")
|
|
add_1637 = Add (add_1608, linear_71)
|
|
layer_norm_36 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1637, "visual.trunk.blocks.18.norm1.weight", "visual.trunk.blocks.18.norm1.bias")
|
|
val_175 = MatMul (layer_norm_36, val_75)
|
|
linear_72 = Add (val_175, "visual.trunk.blocks.18.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_18_qkv_split] node_scaled_dot_product_attention_18_q, node_scaled_dot_product_attention_18_k, node_scaled_dot_product_attention_18_v = Split <axis: int = -1> (linear_72, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_18 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_18_q, node_scaled_dot_product_attention_18_k, node_scaled_dot_product_attention_18_v)
|
|
val_176 = MatMul (scaled_dot_product_attention_18, val_76)
|
|
linear_73 = Add (val_176, "visual.trunk.blocks.18.attn.proj.bias")
|
|
add_1698 = Add (add_1637, linear_73)
|
|
layer_norm_37 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1698, "visual.trunk.blocks.18.norm2.weight", "visual.trunk.blocks.18.norm2.bias")
|
|
val_177 = MatMul (layer_norm_37, val_77)
|
|
linear_74 = Add (val_177, "visual.trunk.blocks.18.mlp.fc1.bias")
|
|
gelu_18 = Gelu <approximate: string = "tanh"> (linear_74)
|
|
val_178 = MatMul (gelu_18, val_78)
|
|
linear_75 = Add (val_178, "visual.trunk.blocks.18.mlp.fc2.bias")
|
|
add_1727 = Add (add_1698, linear_75)
|
|
layer_norm_38 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1727, "visual.trunk.blocks.19.norm1.weight", "visual.trunk.blocks.19.norm1.bias")
|
|
val_179 = MatMul (layer_norm_38, val_79)
|
|
linear_76 = Add (val_179, "visual.trunk.blocks.19.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_19_qkv_split] node_scaled_dot_product_attention_19_q, node_scaled_dot_product_attention_19_k, node_scaled_dot_product_attention_19_v = Split <axis: int = -1> (linear_76, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_19 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_19_q, node_scaled_dot_product_attention_19_k, node_scaled_dot_product_attention_19_v)
|
|
val_180 = MatMul (scaled_dot_product_attention_19, val_80)
|
|
linear_77 = Add (val_180, "visual.trunk.blocks.19.attn.proj.bias")
|
|
add_1788 = Add (add_1727, linear_77)
|
|
layer_norm_39 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1788, "visual.trunk.blocks.19.norm2.weight", "visual.trunk.blocks.19.norm2.bias")
|
|
val_181 = MatMul (layer_norm_39, val_81)
|
|
linear_78 = Add (val_181, "visual.trunk.blocks.19.mlp.fc1.bias")
|
|
gelu_19 = Gelu <approximate: string = "tanh"> (linear_78)
|
|
val_182 = MatMul (gelu_19, val_82)
|
|
linear_79 = Add (val_182, "visual.trunk.blocks.19.mlp.fc2.bias")
|
|
add_1817 = Add (add_1788, linear_79)
|
|
layer_norm_40 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1817, "visual.trunk.blocks.20.norm1.weight", "visual.trunk.blocks.20.norm1.bias")
|
|
val_183 = MatMul (layer_norm_40, val_83)
|
|
linear_80 = Add (val_183, "visual.trunk.blocks.20.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_20_qkv_split] node_scaled_dot_product_attention_20_q, node_scaled_dot_product_attention_20_k, node_scaled_dot_product_attention_20_v = Split <axis: int = -1> (linear_80, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_20 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_20_q, node_scaled_dot_product_attention_20_k, node_scaled_dot_product_attention_20_v)
|
|
val_184 = MatMul (scaled_dot_product_attention_20, val_84)
|
|
linear_81 = Add (val_184, "visual.trunk.blocks.20.attn.proj.bias")
|
|
add_1878 = Add (add_1817, linear_81)
|
|
layer_norm_41 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1878, "visual.trunk.blocks.20.norm2.weight", "visual.trunk.blocks.20.norm2.bias")
|
|
val_185 = MatMul (layer_norm_41, val_85)
|
|
linear_82 = Add (val_185, "visual.trunk.blocks.20.mlp.fc1.bias")
|
|
gelu_20 = Gelu <approximate: string = "tanh"> (linear_82)
|
|
val_186 = MatMul (gelu_20, val_86)
|
|
linear_83 = Add (val_186, "visual.trunk.blocks.20.mlp.fc2.bias")
|
|
add_1907 = Add (add_1878, linear_83)
|
|
layer_norm_42 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1907, "visual.trunk.blocks.21.norm1.weight", "visual.trunk.blocks.21.norm1.bias")
|
|
val_187 = MatMul (layer_norm_42, val_87)
|
|
linear_84 = Add (val_187, "visual.trunk.blocks.21.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_21_qkv_split] node_scaled_dot_product_attention_21_q, node_scaled_dot_product_attention_21_k, node_scaled_dot_product_attention_21_v = Split <axis: int = -1> (linear_84, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_21 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_21_q, node_scaled_dot_product_attention_21_k, node_scaled_dot_product_attention_21_v)
|
|
val_188 = MatMul (scaled_dot_product_attention_21, val_88)
|
|
linear_85 = Add (val_188, "visual.trunk.blocks.21.attn.proj.bias")
|
|
add_1968 = Add (add_1907, linear_85)
|
|
layer_norm_43 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1968, "visual.trunk.blocks.21.norm2.weight", "visual.trunk.blocks.21.norm2.bias")
|
|
val_189 = MatMul (layer_norm_43, val_89)
|
|
linear_86 = Add (val_189, "visual.trunk.blocks.21.mlp.fc1.bias")
|
|
gelu_21 = Gelu <approximate: string = "tanh"> (linear_86)
|
|
val_190 = MatMul (gelu_21, val_90)
|
|
linear_87 = Add (val_190, "visual.trunk.blocks.21.mlp.fc2.bias")
|
|
add_1997 = Add (add_1968, linear_87)
|
|
layer_norm_44 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_1997, "visual.trunk.blocks.22.norm1.weight", "visual.trunk.blocks.22.norm1.bias")
|
|
val_191 = MatMul (layer_norm_44, val_91)
|
|
linear_88 = Add (val_191, "visual.trunk.blocks.22.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_22_qkv_split] node_scaled_dot_product_attention_22_q, node_scaled_dot_product_attention_22_k, node_scaled_dot_product_attention_22_v = Split <axis: int = -1> (linear_88, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_22 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_22_q, node_scaled_dot_product_attention_22_k, node_scaled_dot_product_attention_22_v)
|
|
val_192 = MatMul (scaled_dot_product_attention_22, val_92)
|
|
linear_89 = Add (val_192, "visual.trunk.blocks.22.attn.proj.bias")
|
|
add_2058 = Add (add_1997, linear_89)
|
|
layer_norm_45 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_2058, "visual.trunk.blocks.22.norm2.weight", "visual.trunk.blocks.22.norm2.bias")
|
|
val_193 = MatMul (layer_norm_45, val_93)
|
|
linear_90 = Add (val_193, "visual.trunk.blocks.22.mlp.fc1.bias")
|
|
gelu_22 = Gelu <approximate: string = "tanh"> (linear_90)
|
|
val_194 = MatMul (gelu_22, val_94)
|
|
linear_91 = Add (val_194, "visual.trunk.blocks.22.mlp.fc2.bias")
|
|
add_2087 = Add (add_2058, linear_91)
|
|
layer_norm_46 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_2087, "visual.trunk.blocks.23.norm1.weight", "visual.trunk.blocks.23.norm1.bias")
|
|
val_195 = MatMul (layer_norm_46, val_95)
|
|
linear_92 = Add (val_195, "visual.trunk.blocks.23.attn.qkv.bias")
|
|
[node_scaled_dot_product_attention_23_qkv_split] node_scaled_dot_product_attention_23_q, node_scaled_dot_product_attention_23_k, node_scaled_dot_product_attention_23_v = Split <axis: int = -1> (linear_92, attn3d_split_3x1024)
|
|
scaled_dot_product_attention_23 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_23_q, node_scaled_dot_product_attention_23_k, node_scaled_dot_product_attention_23_v)
|
|
val_196 = MatMul (scaled_dot_product_attention_23, val_96)
|
|
linear_93 = Add (val_196, "visual.trunk.blocks.23.attn.proj.bias")
|
|
add_2148 = Add (add_2087, linear_93)
|
|
layer_norm_47 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_2148, "visual.trunk.blocks.23.norm2.weight", "visual.trunk.blocks.23.norm2.bias")
|
|
val_197 = MatMul (layer_norm_47, val_97)
|
|
linear_94 = Add (val_197, "visual.trunk.blocks.23.mlp.fc1.bias")
|
|
gelu_23 = Gelu <approximate: string = "tanh"> (linear_94)
|
|
val_198 = MatMul (gelu_23, val_98)
|
|
linear_95 = Add (val_198, "visual.trunk.blocks.23.mlp.fc2.bias")
|
|
add_2177 = Add (add_2148, linear_95)
|
|
layer_norm_48 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (add_2177, "visual.trunk.norm.weight", "visual.trunk.norm.bias")
|
|
[ez_slice] image_ez_s = Slice (image_f32, image_ez_starts, image_ez_ends, image_ez_axes)
|
|
[ez_flatten] image_ez_r = Reshape (image_ez_s, val_0)
|
|
[ez_mul] image_ez = Mul (image_ez_r, image_ez_zero)
|
|
val_199 = MatMul (layer_norm_48, val_99)
|
|
linear_97 = Add (val_199, "visual.trunk.attn_pool.kv.bias")
|
|
[node_scaled_dot_product_attention_24_qkv_split] node_scaled_dot_product_attention_24_k, node_scaled_dot_product_attention_24_v = Split <axis: int = -1> (linear_97, attn3d_split_2x1024)
|
|
[node_scaled_dot_product_attention_24_q_col] node_scaled_dot_product_attention_24_q_col_out = Unsqueeze (image_ez, node_scaled_dot_product_attention_24_q_col_axes)
|
|
[node_scaled_dot_product_attention_24_q_bcast] node_scaled_dot_product_attention_24_q = Add (node_scaled_dot_product_attention_24_q3, node_scaled_dot_product_attention_24_q_col_out)
|
|
scaled_dot_product_attention_24 = Attention <is_causal: int = 0, kv_num_heads: int = 16, q_num_heads: int = 16, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_24_q, node_scaled_dot_product_attention_24_k, node_scaled_dot_product_attention_24_v)
|
|
val_200 = MatMul (scaled_dot_product_attention_24, val_100)
|
|
linear_98 = Add (val_200, "visual.trunk.attn_pool.proj.bias")
|
|
layer_norm_49 = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> (linear_98, "visual.trunk.attn_pool.norm.weight", "visual.trunk.attn_pool.norm.bias")
|
|
val_201 = MatMul (layer_norm_49, val_101)
|
|
linear_99 = Add (val_201, "visual.trunk.attn_pool.mlp.fc1.bias")
|
|
gelu_24 = Gelu <approximate: string = "tanh"> (linear_99)
|
|
val_202 = MatMul (gelu_24, val_102)
|
|
linear_100 = Add (val_202, "visual.trunk.attn_pool.mlp.fc2.bias")
|
|
add_2275 = Add (linear_98, linear_100)
|
|
select = Squeeze (add_2275, val_2)
|
|
[node_linalg_vector_norm] linalg_vector_norm = ReduceL2 <keepdims: int = 1, noop_with_empty_axes: int = 0> (select, val_0)
|
|
[node_clamp_min] clamp_min = Clip (linalg_vector_norm, val_1)
|
|
[node_div] image_embedding = Div (select, clamp_min)
|
|
}
|
|
|
|
weights:
|
|
attn3d_split_2x1024 INT64[2] d190f758d9d9
|
|
attn3d_split_3x1024 INT64[3] 0ec6f5651fd5
|
|
image_ez_axes INT64[3] e2e2033ae7e1
|
|
image_ez_ends INT64[3] 605390e5a369
|
|
image_ez_starts INT64[3] 9d908ecfb6b2
|
|
image_ez_zero FLOAT[] df3f619804a9
|
|
node_scaled_dot_product_attention_24_q3 FLOAT[1,1,1024] 28cf1d91f2dd
|
|
node_scaled_dot_product_attention_24_q_col_axes INT64[2] 0c730b69905c
|
|
val_0 INT64[1] 12a3ae445661
|
|
val_1 FLOAT[] 6708d9be4956
|
|
val_10 FLOAT[4096,1024] 6520cf51bdcc
|
|
val_100 FLOAT[1024,1024] 84f771b0c2c4
|
|
val_101 FLOAT[1024,4096] e10c2a84364a
|
|
val_102 FLOAT[4096,1024] a2795ec17506
|
|
val_11 FLOAT[1024,3072] 0bffcb822c2f
|
|
val_12 FLOAT[1024,1024] 8fc21af6d5d6
|
|
val_13 FLOAT[1024,4096] be0e7dd98011
|
|
val_14 FLOAT[4096,1024] 4fef52d437f8
|
|
val_15 FLOAT[1024,3072] f346ffb182ad
|
|
val_16 FLOAT[1024,1024] 023c623195cc
|
|
val_17 FLOAT[1024,4096] 4d3777b21392
|
|
val_18 FLOAT[4096,1024] d478aa446897
|
|
val_19 FLOAT[1024,3072] f949d00fe465
|
|
val_2 INT64[1] 7c9fa136d441
|
|
val_20 FLOAT[1024,1024] 6315593dc094
|
|
val_21 FLOAT[1024,4096] 532741f496f4
|
|
val_22 FLOAT[4096,1024] 39c01e4a0142
|
|
val_23 FLOAT[1024,3072] 8f04ac703147
|
|
val_24 FLOAT[1024,1024] 62a146bea020
|
|
val_25 FLOAT[1024,4096] 709a44e33738
|
|
val_26 FLOAT[4096,1024] f9d1c6bf77fc
|
|
val_27 FLOAT[1024,3072] d2ac7d4f6a72
|
|
val_28 FLOAT[1024,1024] 918232239755
|
|
val_29 FLOAT[1024,4096] 1b242e6b4241
|
|
val_3 FLOAT[1024,3072] 41ce26de9aa1
|
|
val_30 FLOAT[4096,1024] e9d215f9e145
|
|
val_31 FLOAT[1024,3072] a423d0bb7770
|
|
val_32 FLOAT[1024,1024] e0d2d50eff0d
|
|
val_33 FLOAT[1024,4096] c1ea4c520e72
|
|
val_34 FLOAT[4096,1024] be3efd73c452
|
|
val_35 FLOAT[1024,3072] bb64dd9382a5
|
|
val_36 FLOAT[1024,1024] 6be30df6fb67
|
|
val_37 FLOAT[1024,4096] 3db6d26888b2
|
|
val_38 FLOAT[4096,1024] 98a409020eb9
|
|
val_39 FLOAT[1024,3072] 97648d766345
|
|
val_4 FLOAT[1024,1024] 20e99851b06c
|
|
val_40 FLOAT[1024,1024] 85511bb65d34
|
|
val_41 FLOAT[1024,4096] 1728b9c901e5
|
|
val_42 FLOAT[4096,1024] 27e2d39a0710
|
|
val_43 FLOAT[1024,3072] 7fc1edb55c72
|
|
val_44 FLOAT[1024,1024] a9c9a5bab4ab
|
|
val_45 FLOAT[1024,4096] 8d76733997d2
|
|
val_46 FLOAT[4096,1024] 446e348d3e67
|
|
val_47 FLOAT[1024,3072] eab2c5793ce0
|
|
val_48 FLOAT[1024,1024] 68914023fd67
|
|
val_49 FLOAT[1024,4096] a620cc773dd1
|
|
val_5 FLOAT[1024,4096] c18e01f7b5a2
|
|
val_50 FLOAT[4096,1024] caf273abd980
|
|
val_51 FLOAT[1024,3072] 4cf0189616d7
|
|
val_52 FLOAT[1024,1024] db4c0451d413
|
|
val_53 FLOAT[1024,4096] c56923c6a565
|
|
val_54 FLOAT[4096,1024] 90b728acdaf2
|
|
val_55 FLOAT[1024,3072] 2ab56408f72d
|
|
val_56 FLOAT[1024,1024] d915558b0106
|
|
val_57 FLOAT[1024,4096] 70b75656ea69
|
|
val_58 FLOAT[4096,1024] b1976b3f1228
|
|
val_59 FLOAT[1024,3072] 45aeadb60d14
|
|
val_6 FLOAT[4096,1024] 76922e0e5ff3
|
|
val_60 FLOAT[1024,1024] 619eadaa47f6
|
|
val_61 FLOAT[1024,4096] ca73767fefaa
|
|
val_62 FLOAT[4096,1024] 6c9ee10baaee
|
|
val_63 FLOAT[1024,3072] 275981aeb9b7
|
|
val_64 FLOAT[1024,1024] 79177750c4e2
|
|
val_65 FLOAT[1024,4096] 407717537f23
|
|
val_66 FLOAT[4096,1024] 31de69d5be88
|
|
val_67 FLOAT[1024,3072] 6982837f83c7
|
|
val_68 FLOAT[1024,1024] 68cae91348af
|
|
val_69 FLOAT[1024,4096] fbace56b189d
|
|
val_7 FLOAT[1024,3072] 6122b1b258c5
|
|
val_70 FLOAT[4096,1024] 221a3c707faa
|
|
val_71 FLOAT[1024,3072] 16777e59a561
|
|
val_72 FLOAT[1024,1024] 43470111ea66
|
|
val_73 FLOAT[1024,4096] 3c2e558cce5c
|
|
val_74 FLOAT[4096,1024] f508926c2c0d
|
|
val_75 FLOAT[1024,3072] deb906082636
|
|
val_76 FLOAT[1024,1024] 9bfc704cf528
|
|
val_77 FLOAT[1024,4096] bc4ea997ed14
|
|
val_78 FLOAT[4096,1024] fc8d4cb0a9de
|
|
val_79 FLOAT[1024,3072] b800a826905e
|
|
val_8 FLOAT[1024,1024] 62c279167dbd
|
|
val_80 FLOAT[1024,1024] 502865615794
|
|
val_81 FLOAT[1024,4096] fd454d9a4fce
|
|
val_82 FLOAT[4096,1024] 0fe208818095
|
|
val_83 FLOAT[1024,3072] 2cabf8adc1a4
|
|
val_84 FLOAT[1024,1024] 0b3779d8d2bf
|
|
val_85 FLOAT[1024,4096] 2804ccfb9457
|
|
val_86 FLOAT[4096,1024] 2c34aa3dc823
|
|
val_87 FLOAT[1024,3072] 5dad74c3fef5
|
|
val_88 FLOAT[1024,1024] 602d0b340bb1
|
|
val_89 FLOAT[1024,4096] 7c52c7bcd87b
|
|
val_9 FLOAT[1024,4096] 79e7454a3a30
|
|
val_90 FLOAT[4096,1024] d8389bed58bd
|
|
val_91 FLOAT[1024,3072] 7cb88c6eb13f
|
|
val_92 FLOAT[1024,1024] 5214772836e4
|
|
val_93 FLOAT[1024,4096] 0f2b99e927e2
|
|
val_94 FLOAT[4096,1024] 13b38a8bafe0
|
|
val_95 FLOAT[1024,3072] 454e27715a10
|
|
val_96 FLOAT[1024,1024] 0d077f51fb8d
|
|
val_97 FLOAT[1024,4096] 9fe24569637a
|
|
val_98 FLOAT[4096,1024] 447a7bdcf3cc
|
|
val_99 FLOAT[1024,2048] 2511d420a2e0
|
|
view_target INT64[3] 8238569870a0
|
|
visual.trunk.attn_pool.kv.bias FLOAT[2048] 0c4a06ec75dd
|
|
visual.trunk.attn_pool.mlp.fc1.bias FLOAT[4096] f296bfa47fb2
|
|
visual.trunk.attn_pool.mlp.fc2.bias FLOAT[1024] 51a10a3c742c
|
|
visual.trunk.attn_pool.norm.bias FLOAT[1024] 7b237a7c9b12
|
|
visual.trunk.attn_pool.norm.weight FLOAT[1024] 675e8aec50e9
|
|
visual.trunk.attn_pool.proj.bias FLOAT[1024] aae9d335e1d0
|
|
visual.trunk.blocks.0.attn.proj.bias FLOAT[1024] a8637ec3857d
|
|
visual.trunk.blocks.0.attn.qkv.bias FLOAT[3072] 6ead211a3978
|
|
visual.trunk.blocks.0.mlp.fc1.bias FLOAT[4096] 29bc39927ddb
|
|
visual.trunk.blocks.0.mlp.fc2.bias FLOAT[1024] 6b7c8e72f811
|
|
visual.trunk.blocks.0.norm1.bias FLOAT[1024] 14d9bb20ad5e
|
|
visual.trunk.blocks.0.norm1.weight FLOAT[1024] 05ef19046edd
|
|
visual.trunk.blocks.0.norm2.bias FLOAT[1024] 80d29465a500
|
|
visual.trunk.blocks.0.norm2.weight FLOAT[1024] 51a6f6689d6e
|
|
visual.trunk.blocks.1.attn.proj.bias FLOAT[1024] d4604be59e26
|
|
visual.trunk.blocks.1.attn.qkv.bias FLOAT[3072] 8b064ae6e6f6
|
|
visual.trunk.blocks.1.mlp.fc1.bias FLOAT[4096] 138519e15f9e
|
|
visual.trunk.blocks.1.mlp.fc2.bias FLOAT[1024] eb8954bca51c
|
|
visual.trunk.blocks.1.norm1.bias FLOAT[1024] 8958ade9b4cb
|
|
visual.trunk.blocks.1.norm1.weight FLOAT[1024] f4d3d0d32ad0
|
|
visual.trunk.blocks.1.norm2.bias FLOAT[1024] f36cd1baa119
|
|
visual.trunk.blocks.1.norm2.weight FLOAT[1024] 42ae6b33b312
|
|
visual.trunk.blocks.10.attn.proj.bias FLOAT[1024] 5957cc4fd4fa
|
|
visual.trunk.blocks.10.attn.qkv.bias FLOAT[3072] 052227f7aace
|
|
visual.trunk.blocks.10.mlp.fc1.bias FLOAT[4096] 505da389393a
|
|
visual.trunk.blocks.10.mlp.fc2.bias FLOAT[1024] 3c714e7304bf
|
|
visual.trunk.blocks.10.norm1.bias FLOAT[1024] 5452b895a54e
|
|
visual.trunk.blocks.10.norm1.weight FLOAT[1024] 33a3ac2719d2
|
|
visual.trunk.blocks.10.norm2.bias FLOAT[1024] b9811aae1ea6
|
|
visual.trunk.blocks.10.norm2.weight FLOAT[1024] b1dc94fafd25
|
|
visual.trunk.blocks.11.attn.proj.bias FLOAT[1024] fbb72ad71fef
|
|
visual.trunk.blocks.11.attn.qkv.bias FLOAT[3072] 3fb832665f7e
|
|
visual.trunk.blocks.11.mlp.fc1.bias FLOAT[4096] afe25fdeeb25
|
|
visual.trunk.blocks.11.mlp.fc2.bias FLOAT[1024] c9487f688ed2
|
|
visual.trunk.blocks.11.norm1.bias FLOAT[1024] 224f5ab1db9b
|
|
visual.trunk.blocks.11.norm1.weight FLOAT[1024] 02c17c8e5064
|
|
visual.trunk.blocks.11.norm2.bias FLOAT[1024] 9b86231cb003
|
|
visual.trunk.blocks.11.norm2.weight FLOAT[1024] e0949dcf5c66
|
|
visual.trunk.blocks.12.attn.proj.bias FLOAT[1024] 3c74e432eaf2
|
|
visual.trunk.blocks.12.attn.qkv.bias FLOAT[3072] f14fc3cd140f
|
|
visual.trunk.blocks.12.mlp.fc1.bias FLOAT[4096] c5aff6edc4f4
|
|
visual.trunk.blocks.12.mlp.fc2.bias FLOAT[1024] ac482ea0856d
|
|
visual.trunk.blocks.12.norm1.bias FLOAT[1024] cc369de0f4fc
|
|
visual.trunk.blocks.12.norm1.weight FLOAT[1024] 7c3f41c8b917
|
|
visual.trunk.blocks.12.norm2.bias FLOAT[1024] fa9bdba21bdf
|
|
visual.trunk.blocks.12.norm2.weight FLOAT[1024] dec4b339d7c6
|
|
visual.trunk.blocks.13.attn.proj.bias FLOAT[1024] ec01b67f8178
|
|
visual.trunk.blocks.13.attn.qkv.bias FLOAT[3072] 45eb95d45a02
|
|
visual.trunk.blocks.13.mlp.fc1.bias FLOAT[4096] 177c222532ac
|
|
visual.trunk.blocks.13.mlp.fc2.bias FLOAT[1024] e6aa13c8666c
|
|
visual.trunk.blocks.13.norm1.bias FLOAT[1024] 7890d5b6a70f
|
|
visual.trunk.blocks.13.norm1.weight FLOAT[1024] bba12dbe4c17
|
|
visual.trunk.blocks.13.norm2.bias FLOAT[1024] f8d07ed59f4f
|
|
visual.trunk.blocks.13.norm2.weight FLOAT[1024] f7ea1a45b2f4
|
|
visual.trunk.blocks.14.attn.proj.bias FLOAT[1024] 1e116964bc68
|
|
visual.trunk.blocks.14.attn.qkv.bias FLOAT[3072] 32330021b083
|
|
visual.trunk.blocks.14.mlp.fc1.bias FLOAT[4096] a19c1c7a80b0
|
|
visual.trunk.blocks.14.mlp.fc2.bias FLOAT[1024] 617c1825b599
|
|
visual.trunk.blocks.14.norm1.bias FLOAT[1024] 68e45a7871f2
|
|
visual.trunk.blocks.14.norm1.weight FLOAT[1024] 8c15d334f3ab
|
|
visual.trunk.blocks.14.norm2.bias FLOAT[1024] 9f3f937b0862
|
|
visual.trunk.blocks.14.norm2.weight FLOAT[1024] b51e38053e97
|
|
visual.trunk.blocks.15.attn.proj.bias FLOAT[1024] 9bf61606c7c8
|
|
visual.trunk.blocks.15.attn.qkv.bias FLOAT[3072] 0dd618150933
|
|
visual.trunk.blocks.15.mlp.fc1.bias FLOAT[4096] 9dbd776eb25f
|
|
visual.trunk.blocks.15.mlp.fc2.bias FLOAT[1024] 93f65c0a2e1f
|
|
visual.trunk.blocks.15.norm1.bias FLOAT[1024] 1c0eda60f807
|
|
visual.trunk.blocks.15.norm1.weight FLOAT[1024] a07d57d62a91
|
|
visual.trunk.blocks.15.norm2.bias FLOAT[1024] 912a5eab4079
|
|
visual.trunk.blocks.15.norm2.weight FLOAT[1024] e1eab9b6070b
|
|
visual.trunk.blocks.16.attn.proj.bias FLOAT[1024] e7d7b38e659d
|
|
visual.trunk.blocks.16.attn.qkv.bias FLOAT[3072] 91b9c82cfd37
|
|
visual.trunk.blocks.16.mlp.fc1.bias FLOAT[4096] 382fc42ffd32
|
|
visual.trunk.blocks.16.mlp.fc2.bias FLOAT[1024] b8d338540829
|
|
visual.trunk.blocks.16.norm1.bias FLOAT[1024] caa978acdcc7
|
|
visual.trunk.blocks.16.norm1.weight FLOAT[1024] 27a5c57505d1
|
|
visual.trunk.blocks.16.norm2.bias FLOAT[1024] f6f33c6d780e
|
|
visual.trunk.blocks.16.norm2.weight FLOAT[1024] bc648cb80620
|
|
visual.trunk.blocks.17.attn.proj.bias FLOAT[1024] 4db7de7433c0
|
|
visual.trunk.blocks.17.attn.qkv.bias FLOAT[3072] 55ae4673a03d
|
|
visual.trunk.blocks.17.mlp.fc1.bias FLOAT[4096] 032f32b0b1ea
|
|
visual.trunk.blocks.17.mlp.fc2.bias FLOAT[1024] a4bf3ca3c947
|
|
visual.trunk.blocks.17.norm1.bias FLOAT[1024] 8e021b7d8005
|
|
visual.trunk.blocks.17.norm1.weight FLOAT[1024] 6114e57246c8
|
|
visual.trunk.blocks.17.norm2.bias FLOAT[1024] c941588b1598
|
|
visual.trunk.blocks.17.norm2.weight FLOAT[1024] c4c09c3dda57
|
|
visual.trunk.blocks.18.attn.proj.bias FLOAT[1024] cee8f157715a
|
|
visual.trunk.blocks.18.attn.qkv.bias FLOAT[3072] 087fdca070e6
|
|
visual.trunk.blocks.18.mlp.fc1.bias FLOAT[4096] 4f17d139b4e2
|
|
visual.trunk.blocks.18.mlp.fc2.bias FLOAT[1024] 9b645c589396
|
|
visual.trunk.blocks.18.norm1.bias FLOAT[1024] a05dcbc4d59a
|
|
visual.trunk.blocks.18.norm1.weight FLOAT[1024] 2e243d0df4b3
|
|
visual.trunk.blocks.18.norm2.bias FLOAT[1024] 23e8b450f3e1
|
|
visual.trunk.blocks.18.norm2.weight FLOAT[1024] c01969abaa8b
|
|
visual.trunk.blocks.19.attn.proj.bias FLOAT[1024] ccfdef3e1e8c
|
|
visual.trunk.blocks.19.attn.qkv.bias FLOAT[3072] 61dd00ea32ac
|
|
visual.trunk.blocks.19.mlp.fc1.bias FLOAT[4096] 4e4efee2f0cd
|
|
visual.trunk.blocks.19.mlp.fc2.bias FLOAT[1024] 26a5a4ccfe78
|
|
visual.trunk.blocks.19.norm1.bias FLOAT[1024] c794927f5453
|
|
visual.trunk.blocks.19.norm1.weight FLOAT[1024] 71e8b06eb97e
|
|
visual.trunk.blocks.19.norm2.bias FLOAT[1024] 126309473043
|
|
visual.trunk.blocks.19.norm2.weight FLOAT[1024] 0b3be6181ce2
|
|
visual.trunk.blocks.2.attn.proj.bias FLOAT[1024] d1bc02bb185f
|
|
visual.trunk.blocks.2.attn.qkv.bias FLOAT[3072] 5d7e26bbbc0c
|
|
visual.trunk.blocks.2.mlp.fc1.bias FLOAT[4096] a65344f21903
|
|
visual.trunk.blocks.2.mlp.fc2.bias FLOAT[1024] 09d31e26eacd
|
|
visual.trunk.blocks.2.norm1.bias FLOAT[1024] 98c2e9ccf56c
|
|
visual.trunk.blocks.2.norm1.weight FLOAT[1024] 55ed567b7531
|
|
visual.trunk.blocks.2.norm2.bias FLOAT[1024] dcd37c53063c
|
|
visual.trunk.blocks.2.norm2.weight FLOAT[1024] c8d08c5e8539
|
|
visual.trunk.blocks.20.attn.proj.bias FLOAT[1024] d8bb4c4cbd54
|
|
visual.trunk.blocks.20.attn.qkv.bias FLOAT[3072] 3e89492c4b90
|
|
visual.trunk.blocks.20.mlp.fc1.bias FLOAT[4096] dbd9b4a849a2
|
|
visual.trunk.blocks.20.mlp.fc2.bias FLOAT[1024] 4a6d42a1441d
|
|
visual.trunk.blocks.20.norm1.bias FLOAT[1024] f55631534398
|
|
visual.trunk.blocks.20.norm1.weight FLOAT[1024] 98276d8920db
|
|
visual.trunk.blocks.20.norm2.bias FLOAT[1024] dee466b3533b
|
|
visual.trunk.blocks.20.norm2.weight FLOAT[1024] 5dcb16c8c333
|
|
visual.trunk.blocks.21.attn.proj.bias FLOAT[1024] 2c72f1ab5bea
|
|
visual.trunk.blocks.21.attn.qkv.bias FLOAT[3072] 1eb5a8fd09ce
|
|
visual.trunk.blocks.21.mlp.fc1.bias FLOAT[4096] 798e1daa45c1
|
|
visual.trunk.blocks.21.mlp.fc2.bias FLOAT[1024] eea5701915a2
|
|
visual.trunk.blocks.21.norm1.bias FLOAT[1024] f065ed423e7f
|
|
visual.trunk.blocks.21.norm1.weight FLOAT[1024] 09836b83fd6b
|
|
visual.trunk.blocks.21.norm2.bias FLOAT[1024] 0543006478ea
|
|
visual.trunk.blocks.21.norm2.weight FLOAT[1024] 911213c46504
|
|
visual.trunk.blocks.22.attn.proj.bias FLOAT[1024] 1f46defe81b5
|
|
visual.trunk.blocks.22.attn.qkv.bias FLOAT[3072] 9920740cb7c9
|
|
visual.trunk.blocks.22.mlp.fc1.bias FLOAT[4096] 0d2cdc283967
|
|
visual.trunk.blocks.22.mlp.fc2.bias FLOAT[1024] 0e2db2ddb09a
|
|
visual.trunk.blocks.22.norm1.bias FLOAT[1024] 4e140488ff5c
|
|
visual.trunk.blocks.22.norm1.weight FLOAT[1024] 1fc6878b1250
|
|
visual.trunk.blocks.22.norm2.bias FLOAT[1024] cf3fdd8d64b2
|
|
visual.trunk.blocks.22.norm2.weight FLOAT[1024] 211d32897483
|
|
visual.trunk.blocks.23.attn.proj.bias FLOAT[1024] 1e62ab22a6cd
|
|
visual.trunk.blocks.23.attn.qkv.bias FLOAT[3072] d4530ab50e60
|
|
visual.trunk.blocks.23.mlp.fc1.bias FLOAT[4096] 427537febf68
|
|
visual.trunk.blocks.23.mlp.fc2.bias FLOAT[1024] e30dd0407675
|
|
visual.trunk.blocks.23.norm1.bias FLOAT[1024] c938cc2ea3e5
|
|
visual.trunk.blocks.23.norm1.weight FLOAT[1024] 29afa45ce3dd
|
|
visual.trunk.blocks.23.norm2.bias FLOAT[1024] 22d6434169b7
|
|
visual.trunk.blocks.23.norm2.weight FLOAT[1024] 79a1d8465076
|
|
visual.trunk.blocks.3.attn.proj.bias FLOAT[1024] 6fd99767fe66
|
|
visual.trunk.blocks.3.attn.qkv.bias FLOAT[3072] 815530544fbe
|
|
visual.trunk.blocks.3.mlp.fc1.bias FLOAT[4096] 91bcc674c16a
|
|
visual.trunk.blocks.3.mlp.fc2.bias FLOAT[1024] f37dbe8360a1
|
|
visual.trunk.blocks.3.norm1.bias FLOAT[1024] 5fb16a2ce670
|
|
visual.trunk.blocks.3.norm1.weight FLOAT[1024] 9076883a3452
|
|
visual.trunk.blocks.3.norm2.bias FLOAT[1024] f34372b88458
|
|
visual.trunk.blocks.3.norm2.weight FLOAT[1024] 26becafe170c
|
|
visual.trunk.blocks.4.attn.proj.bias FLOAT[1024] 684ac67cd57a
|
|
visual.trunk.blocks.4.attn.qkv.bias FLOAT[3072] 20e524c764d5
|
|
visual.trunk.blocks.4.mlp.fc1.bias FLOAT[4096] c4c0bff77372
|
|
visual.trunk.blocks.4.mlp.fc2.bias FLOAT[1024] ce00f6bd516b
|
|
visual.trunk.blocks.4.norm1.bias FLOAT[1024] a6562d7ee67f
|
|
visual.trunk.blocks.4.norm1.weight FLOAT[1024] ee1e72eccf7b
|
|
visual.trunk.blocks.4.norm2.bias FLOAT[1024] e423d02c6743
|
|
visual.trunk.blocks.4.norm2.weight FLOAT[1024] 1bb0f0296307
|
|
visual.trunk.blocks.5.attn.proj.bias FLOAT[1024] 3c332606ccfb
|
|
visual.trunk.blocks.5.attn.qkv.bias FLOAT[3072] 89f8a2b6aad7
|
|
visual.trunk.blocks.5.mlp.fc1.bias FLOAT[4096] 32a28ddcfb97
|
|
visual.trunk.blocks.5.mlp.fc2.bias FLOAT[1024] b21c587066c3
|
|
visual.trunk.blocks.5.norm1.bias FLOAT[1024] 918ba16f44a8
|
|
visual.trunk.blocks.5.norm1.weight FLOAT[1024] 746707899831
|
|
visual.trunk.blocks.5.norm2.bias FLOAT[1024] 79709484e80f
|
|
visual.trunk.blocks.5.norm2.weight FLOAT[1024] 8e76f3f48f47
|
|
visual.trunk.blocks.6.attn.proj.bias FLOAT[1024] f597190def66
|
|
visual.trunk.blocks.6.attn.qkv.bias FLOAT[3072] 674745f55db5
|
|
visual.trunk.blocks.6.mlp.fc1.bias FLOAT[4096] abe4374b3c26
|
|
visual.trunk.blocks.6.mlp.fc2.bias FLOAT[1024] 8892f41e07d0
|
|
visual.trunk.blocks.6.norm1.bias FLOAT[1024] 1ffb76461597
|
|
visual.trunk.blocks.6.norm1.weight FLOAT[1024] 37ba30b2a53e
|
|
visual.trunk.blocks.6.norm2.bias FLOAT[1024] 1d665464b45a
|
|
visual.trunk.blocks.6.norm2.weight FLOAT[1024] 8b22605036dd
|
|
visual.trunk.blocks.7.attn.proj.bias FLOAT[1024] 76a6062ad3b6
|
|
visual.trunk.blocks.7.attn.qkv.bias FLOAT[3072] 255cb29274b2
|
|
visual.trunk.blocks.7.mlp.fc1.bias FLOAT[4096] 7b66292f4973
|
|
visual.trunk.blocks.7.mlp.fc2.bias FLOAT[1024] aa69ce6ec656
|
|
visual.trunk.blocks.7.norm1.bias FLOAT[1024] 247e298c36d8
|
|
visual.trunk.blocks.7.norm1.weight FLOAT[1024] bd7dc7a645f4
|
|
visual.trunk.blocks.7.norm2.bias FLOAT[1024] 689b9b37c36e
|
|
visual.trunk.blocks.7.norm2.weight FLOAT[1024] e3c06431487a
|
|
visual.trunk.blocks.8.attn.proj.bias FLOAT[1024] 1da490f9ce92
|
|
visual.trunk.blocks.8.attn.qkv.bias FLOAT[3072] a4e8eda0e38b
|
|
visual.trunk.blocks.8.mlp.fc1.bias FLOAT[4096] 573ff2e613f0
|
|
visual.trunk.blocks.8.mlp.fc2.bias FLOAT[1024] a822ac2045fc
|
|
visual.trunk.blocks.8.norm1.bias FLOAT[1024] ab8d86bd1271
|
|
visual.trunk.blocks.8.norm1.weight FLOAT[1024] 2fdad942052b
|
|
visual.trunk.blocks.8.norm2.bias FLOAT[1024] 4f2eac1c2954
|
|
visual.trunk.blocks.8.norm2.weight FLOAT[1024] 9496ea98a990
|
|
visual.trunk.blocks.9.attn.proj.bias FLOAT[1024] 4074b2f6115d
|
|
visual.trunk.blocks.9.attn.qkv.bias FLOAT[3072] 34286494729f
|
|
visual.trunk.blocks.9.mlp.fc1.bias FLOAT[4096] 23834b62bf4d
|
|
visual.trunk.blocks.9.mlp.fc2.bias FLOAT[1024] 50e485a639e3
|
|
visual.trunk.blocks.9.norm1.bias FLOAT[1024] fc379cc4ff31
|
|
visual.trunk.blocks.9.norm1.weight FLOAT[1024] f75fff1fa250
|
|
visual.trunk.blocks.9.norm2.bias FLOAT[1024] 2d4adca5ac90
|
|
visual.trunk.blocks.9.norm2.weight FLOAT[1024] f65f0b60f1e4
|
|
visual.trunk.norm.bias FLOAT[1024] 6cfe280d6e4a
|
|
visual.trunk.norm.weight FLOAT[1024] 252654e90119
|
|
visual.trunk.patch_embed.proj.bias FLOAT[1024] a5ad65d732e9
|
|
visual.trunk.patch_embed.proj.weight FLOAT[1024,3,16,16] cade66775251
|
|
visual.trunk.pos_embed FLOAT[1,1024,1024] 0b44aa586f41
|