donut-epoch1 / config.json
tt1225's picture
Upload 9 files
122dfd3 verified
{
"architectures": [
"ChartReLAInstructSmall"
],
"decoder": {
"add_cross_attention": true,
"add_final_layer_norm": true,
"d_model": 512,
"decoder_layers": 4,
"is_decoder": true,
"is_encoder_decoder": false,
"max_position_embeddings": 1536,
"model_type": "mbart",
"scale_embedding": true,
"vocab_size": 57531
},
"encoder": {
"adapter": {
"hidden_dim": 512,
"image_feat_dims": {
"res2": 128,
"res3": 256,
"res4": 512,
"res5": 1024
},
"mask_dim": 512,
"num_queries": 128
},
"model_type": "multimodal_rela_adapter",
"text_encoder": {
"architectures": [
"BertForMaskedLM"
],
"gradient_checkpointing": false,
"vocab_size": 57531
},
"vision_encoder": {
"depths": [
2,
2,
14,
2
],
"embed_dim": 128,
"hidden_size": 1024,
"image_size": [
960,
960
],
"num_heads": [
4,
8,
16,
32
],
"window_size": 10
}
},
"is_encoder_decoder": true,
"model_type": "Chart-rela-instruct",
"processor_name_or_path": "ahmed-masry/unichart-chartqa-960",
"torch_dtype": "float32",
"transformers_version": "4.38.2"
}