|
{ |
|
"drop_rate":0.1, |
|
"freeze_RoBERTa":false, |
|
"freeze_ViT":false, |
|
"freeze_layer_count_roberta":false, |
|
"freeze_layer_count_vit":false, |
|
"head_hidden_scale":2, |
|
"hidden_act":"gelu", |
|
"hidden_size":768, |
|
"image_size":288, |
|
"input_text_embed_size":768, |
|
"input_image_embed_size":768, |
|
"is_encoder_decoder":false, |
|
"layer_norm_eps":1e-5, |
|
"link_tower_shared":false, |
|
"link_tower_type":"add", |
|
"max_text_len":50, |
|
"mlp_ratio":4, |
|
"num_attention_heads":12, |
|
"num_hidden_layers":6, |
|
"resolution_before":224, |
|
"stop_gradient":false, |
|
"task_head_layers":2, |
|
"tie_word_embeddings":false, |
|
"tokenizer":"roberta-base", |
|
"unfreeze_RoBERTa_attention":false, |
|
"unfreeze_RoBERTa_embeddings":false, |
|
"unfreeze_RoBERTa_encoder":false, |
|
"unfreeze_RoBERTa_layernorm":false, |
|
"unfreeze_ViT_attention":false, |
|
"unfreeze_ViT_layernorm":false, |
|
"vit_embed_dim":512, |
|
"vit_layers":12, |
|
"vit_layernorm_init_from_vit":false, |
|
"vit_layernorm_shared":true, |
|
"vit_patch_size":16, |
|
"vit_remove_last":false, |
|
"vit_transformer_width":512, |
|
"vit_width":768, |
|
"vocab_size":50265 |
|
} |
|
|