blip2_opt_6.7b / model.weights.json
laxmareddyp's picture
Upload folder using huggingface_hub
3cfaa1f verified
Raw
History Blame Contribute Delete
110 kB
{
"metadata": {
"total_size": 31011477504.0
},
"weight_map": {
"/layers/blip2_vision_encoder/layers/vi_t_patching_and_embedding/vars": "model_00000.weights.h5",
"/layers/blip2_vision_encoder/layers/vi_t_patching_and_embedding/patch_embedding/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_patching_and_embedding/position_embedding/vars": "model_00000.weights.h5",
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_1/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_1/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_1/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_1/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_1/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_1/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_1/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_1/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_2/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_2/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_2/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_2/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_2/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_2/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_2/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_2/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_3/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_3/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_3/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_3/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_3/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_3/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_3/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_3/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_4/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_4/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_4/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_4/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_4/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_4/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_4/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_4/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_5/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_5/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_5/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_5/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_5/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_5/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_5/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_5/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_6/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_6/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_6/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_6/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_6/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_6/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_6/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_6/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_7/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_7/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_7/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_7/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_7/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_7/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_7/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_7/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_8/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_8/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_8/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_8/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_8/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_8/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_8/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_8/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_9/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_9/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_9/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_9/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_9/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_9/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_9/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_9/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_10/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_10/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_10/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_10/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_10/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_10/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_10/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_10/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_11/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_11/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_11/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_11/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_11/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_11/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_11/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_11/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_12/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_12/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_12/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_12/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_12/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_12/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_12/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_12/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_13/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_13/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_13/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_13/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_13/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_13/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_13/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_13/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_14/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_14/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_14/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_14/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_14/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_14/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_14/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_14/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_15/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_15/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_15/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_15/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_15/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_15/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_15/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_15/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_16/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_16/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_16/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_16/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_16/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_16/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_16/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_16/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_17/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_17/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_17/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_17/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_17/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_17/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_17/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_17/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_18/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_18/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_18/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_18/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_18/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_18/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_18/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_18/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_19/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_19/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_19/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_19/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_19/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_19/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_19/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_19/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_20/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_20/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_20/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_20/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_20/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_20/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_20/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_20/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_21/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_21/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_21/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_21/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_21/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_21/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_21/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_21/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_22/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_22/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_22/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_22/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_22/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_22/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_22/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_22/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_23/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_23/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_23/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_23/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_23/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_23/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_23/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_23/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_24/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_24/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_24/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_24/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_24/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_24/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_24/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_24/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_25/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_25/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_25/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_25/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_25/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_25/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_25/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_25/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_26/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_26/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_26/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_26/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_26/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_26/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_26/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_26/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_27/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_27/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_27/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_27/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_27/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_27/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_27/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_27/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_28/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_28/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_28/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_28/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_28/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_28/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_28/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_28/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_29/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_29/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_29/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_29/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_29/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_29/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_29/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_29/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_30/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_30/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_30/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_30/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_30/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_30/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_30/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_30/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_31/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_31/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_31/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_31/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_31/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_31/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_31/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_31/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_32/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_32/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_32/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_32/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_32/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_32/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_32/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_32/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_33/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_33/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_33/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_33/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_33/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_33/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_33/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_33/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_34/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_34/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_34/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_34/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_34/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_34/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_34/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_34/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_35/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_35/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_35/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_35/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_35/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_35/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_35/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_35/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_36/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_36/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_36/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_36/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_36/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_36/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_36/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_36/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_37/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_37/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_37/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_37/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_37/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_37/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_37/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_37/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_38/layer_norm_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_38/layer_norm_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_38/mha/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_38/mha/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_38/mha/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_38/mha/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_38/mlp/dense_1/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/encoder_layers/vi_t_encoder_block_38/mlp/dense_2/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_vision_encoder/layers/vi_t_encoder/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/blip2_query_tokens/vars": "model_00000.weights.h5",
"/layers/blip2q_former/layers/layer_normalization/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer/cross_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer/cross_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer/cross_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer/cross_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer/cross_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer/intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer/output_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer/self_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer/self_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer/self_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer/self_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer/self_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_1/intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_1/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_1/output_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_1/self_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_1/self_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_1/self_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_1/self_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_1/self_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_2/cross_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_2/cross_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_2/cross_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_2/cross_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_2/cross_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_2/intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_2/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_2/output_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_2/self_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_2/self_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_2/self_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_2/self_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_2/self_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_3/intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_3/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_3/output_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_3/self_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_3/self_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_3/self_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_3/self_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_3/self_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_4/cross_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_4/cross_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_4/cross_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_4/cross_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_4/cross_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_4/intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_4/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_4/output_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_4/self_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_4/self_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_4/self_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_4/self_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_4/self_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_5/intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_5/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_5/output_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_5/self_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_5/self_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_5/self_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_5/self_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_5/self_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_6/cross_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_6/cross_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_6/cross_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_6/cross_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_6/cross_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_6/intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_6/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_6/output_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_6/self_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_6/self_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_6/self_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_6/self_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_6/self_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_7/intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_7/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_7/output_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_7/self_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_7/self_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_7/self_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_7/self_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_7/self_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_8/cross_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_8/cross_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_8/cross_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_8/cross_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_8/cross_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_8/intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_8/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_8/output_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_8/self_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_8/self_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_8/self_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_8/self_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_8/self_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_9/intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_9/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_9/output_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_9/self_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_9/self_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_9/self_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_9/self_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_9/self_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_10/cross_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_10/cross_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_10/cross_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_10/cross_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_10/cross_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_10/intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_10/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_10/output_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_10/self_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_10/self_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_10/self_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_10/self_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_10/self_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_11/intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_11/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_11/output_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_11/self_attention/attention/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_11/self_attention/attention/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_11/self_attention/attention/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_11/self_attention/attention/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2q_former/layers/q_former_layer_11/self_attention/layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/einsum_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/blip2opt_embeddings/position_embedding/vars": "model_00000.weights.h5",
"/layers/blip2_custom_opt/layers/blip2opt_embeddings/token_embedding/vars": "model_00000.weights.h5",
"/layers/blip2_custom_opt/layers/transformer_decoder/_feedforward_intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder/_feedforward_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder/_feedforward_output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder/_self_attention_layer/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder/_self_attention_layer/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder/_self_attention_layer/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder/_self_attention_layer/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder/_self_attention_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_1/_feedforward_intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_1/_feedforward_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_1/_feedforward_output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_1/_self_attention_layer/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_1/_self_attention_layer/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_1/_self_attention_layer/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_1/_self_attention_layer/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_1/_self_attention_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_2/_feedforward_intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_2/_feedforward_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_2/_feedforward_output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_2/_self_attention_layer/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_2/_self_attention_layer/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_2/_self_attention_layer/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_2/_self_attention_layer/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_2/_self_attention_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_3/_feedforward_intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_3/_feedforward_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_3/_feedforward_output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_3/_self_attention_layer/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_3/_self_attention_layer/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_3/_self_attention_layer/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_3/_self_attention_layer/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_3/_self_attention_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_4/_feedforward_intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_4/_feedforward_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_4/_feedforward_output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_4/_self_attention_layer/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_4/_self_attention_layer/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_4/_self_attention_layer/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_4/_self_attention_layer/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_4/_self_attention_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_5/_feedforward_intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_5/_feedforward_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_5/_feedforward_output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_5/_self_attention_layer/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_5/_self_attention_layer/output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_5/_self_attention_layer/query_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_5/_self_attention_layer/value_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_5/_self_attention_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_6/_feedforward_intermediate_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_6/_feedforward_layer_norm/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_6/_feedforward_output_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_6/_self_attention_layer/key_dense/vars": [
"model_00000.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_6/_self_attention_layer/output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_6/_self_attention_layer/query_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_6/_self_attention_layer/value_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_6/_self_attention_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_7/_feedforward_intermediate_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_7/_feedforward_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_7/_feedforward_output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_7/_self_attention_layer/key_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_7/_self_attention_layer/output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_7/_self_attention_layer/query_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_7/_self_attention_layer/value_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_7/_self_attention_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_8/_feedforward_intermediate_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_8/_feedforward_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_8/_feedforward_output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_8/_self_attention_layer/key_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_8/_self_attention_layer/output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_8/_self_attention_layer/query_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_8/_self_attention_layer/value_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_8/_self_attention_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_9/_feedforward_intermediate_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_9/_feedforward_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_9/_feedforward_output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_9/_self_attention_layer/key_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_9/_self_attention_layer/output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_9/_self_attention_layer/query_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_9/_self_attention_layer/value_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_9/_self_attention_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_10/_feedforward_intermediate_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_10/_feedforward_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_10/_feedforward_output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_10/_self_attention_layer/key_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_10/_self_attention_layer/output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_10/_self_attention_layer/query_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_10/_self_attention_layer/value_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_10/_self_attention_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_11/_feedforward_intermediate_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_11/_feedforward_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_11/_feedforward_output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_11/_self_attention_layer/key_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_11/_self_attention_layer/output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_11/_self_attention_layer/query_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_11/_self_attention_layer/value_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_11/_self_attention_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_12/_feedforward_intermediate_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_12/_feedforward_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_12/_feedforward_output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_12/_self_attention_layer/key_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_12/_self_attention_layer/output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_12/_self_attention_layer/query_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_12/_self_attention_layer/value_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_12/_self_attention_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_13/_feedforward_intermediate_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_13/_feedforward_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_13/_feedforward_output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_13/_self_attention_layer/key_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_13/_self_attention_layer/output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_13/_self_attention_layer/query_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_13/_self_attention_layer/value_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_13/_self_attention_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_14/_feedforward_intermediate_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_14/_feedforward_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_14/_feedforward_output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_14/_self_attention_layer/key_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_14/_self_attention_layer/output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_14/_self_attention_layer/query_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_14/_self_attention_layer/value_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_14/_self_attention_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_15/_feedforward_intermediate_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_15/_feedforward_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_15/_feedforward_output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_15/_self_attention_layer/key_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_15/_self_attention_layer/output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_15/_self_attention_layer/query_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_15/_self_attention_layer/value_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_15/_self_attention_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_16/_feedforward_intermediate_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_16/_feedforward_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_16/_feedforward_output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_16/_self_attention_layer/key_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_16/_self_attention_layer/output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_16/_self_attention_layer/query_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_16/_self_attention_layer/value_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_16/_self_attention_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_17/_feedforward_intermediate_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_17/_feedforward_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_17/_feedforward_output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_17/_self_attention_layer/key_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_17/_self_attention_layer/output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_17/_self_attention_layer/query_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_17/_self_attention_layer/value_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_17/_self_attention_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_18/_feedforward_intermediate_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_18/_feedforward_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_18/_feedforward_output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_18/_self_attention_layer/key_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_18/_self_attention_layer/output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_18/_self_attention_layer/query_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_18/_self_attention_layer/value_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_18/_self_attention_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_19/_feedforward_intermediate_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_19/_feedforward_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_19/_feedforward_output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_19/_self_attention_layer/key_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_19/_self_attention_layer/output_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_19/_self_attention_layer/query_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_19/_self_attention_layer/value_dense/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_19/_self_attention_layer_norm/vars": [
"model_00001.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_20/_feedforward_intermediate_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_20/_feedforward_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_20/_feedforward_output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_20/_self_attention_layer/key_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_20/_self_attention_layer/output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_20/_self_attention_layer/query_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_20/_self_attention_layer/value_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_20/_self_attention_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_21/_feedforward_intermediate_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_21/_feedforward_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_21/_feedforward_output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_21/_self_attention_layer/key_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_21/_self_attention_layer/output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_21/_self_attention_layer/query_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_21/_self_attention_layer/value_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_21/_self_attention_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_22/_feedforward_intermediate_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_22/_feedforward_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_22/_feedforward_output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_22/_self_attention_layer/key_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_22/_self_attention_layer/output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_22/_self_attention_layer/query_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_22/_self_attention_layer/value_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_22/_self_attention_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_23/_feedforward_intermediate_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_23/_feedforward_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_23/_feedforward_output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_23/_self_attention_layer/key_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_23/_self_attention_layer/output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_23/_self_attention_layer/query_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_23/_self_attention_layer/value_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_23/_self_attention_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_24/_feedforward_intermediate_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_24/_feedforward_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_24/_feedforward_output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_24/_self_attention_layer/key_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_24/_self_attention_layer/output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_24/_self_attention_layer/query_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_24/_self_attention_layer/value_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_24/_self_attention_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_25/_feedforward_intermediate_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_25/_feedforward_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_25/_feedforward_output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_25/_self_attention_layer/key_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_25/_self_attention_layer/output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_25/_self_attention_layer/query_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_25/_self_attention_layer/value_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_25/_self_attention_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_26/_feedforward_intermediate_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_26/_feedforward_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_26/_feedforward_output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_26/_self_attention_layer/key_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_26/_self_attention_layer/output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_26/_self_attention_layer/query_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_26/_self_attention_layer/value_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_26/_self_attention_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_27/_feedforward_intermediate_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_27/_feedforward_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_27/_feedforward_output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_27/_self_attention_layer/key_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_27/_self_attention_layer/output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_27/_self_attention_layer/query_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_27/_self_attention_layer/value_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_27/_self_attention_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_28/_feedforward_intermediate_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_28/_feedforward_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_28/_feedforward_output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_28/_self_attention_layer/key_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_28/_self_attention_layer/output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_28/_self_attention_layer/query_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_28/_self_attention_layer/value_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_28/_self_attention_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_29/_feedforward_intermediate_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_29/_feedforward_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_29/_feedforward_output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_29/_self_attention_layer/key_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_29/_self_attention_layer/output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_29/_self_attention_layer/query_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_29/_self_attention_layer/value_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_29/_self_attention_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_30/_feedforward_intermediate_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_30/_feedforward_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_30/_feedforward_output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_30/_self_attention_layer/key_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_30/_self_attention_layer/output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_30/_self_attention_layer/query_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_30/_self_attention_layer/value_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_30/_self_attention_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_31/_feedforward_intermediate_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_31/_feedforward_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_31/_feedforward_output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_31/_self_attention_layer/key_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_31/_self_attention_layer/output_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_31/_self_attention_layer/query_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_31/_self_attention_layer/value_dense/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/transformer_decoder_31/_self_attention_layer_norm/vars": [
"model_00002.weights.h5"
],
"/layers/blip2_custom_opt/layers/layer_normalization/vars": [
"model_00002.weights.h5"
]
}
}