| { | |
| "_name_or_path": "./PreFLMR_ViT-G", | |
| "architectures": [ | |
| "FLMRModelForRetrieval" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_flmr.FLMRConfig", | |
| "AutoModel": "modeling_flmr.FLMRModelForRetrieval" | |
| }, | |
| "context_concat_output_from_text_encoder": true, | |
| "context_concat_output_from_vision_encoder": false, | |
| "dim": 128, | |
| "initializer_range": 0.02, | |
| "load_cpu_extension": false, | |
| "mapping_network_prefix_length": 32, | |
| "mask_instruction_token": ":", | |
| "mask_punctuation": true, | |
| "model_type": "flmr", | |
| "query_concat_output_from_text_encoder": true, | |
| "query_concat_output_from_vision_encoder": true, | |
| "separate_query_and_context_text_encoder": true, | |
| "separate_query_and_context_vision_encoder": false, | |
| "text_config": { | |
| "architectures": [ | |
| "BertForMaskedLM" | |
| ], | |
| "gradient_checkpointing": false, | |
| "model_type": "flmr_text_model", | |
| "use_cache": true | |
| }, | |
| "torch_dtype": "float32", | |
| "transformer_mapping_config_base": "bert-base-uncased", | |
| "transformer_mapping_cross_attention_length": 32, | |
| "transformer_mapping_num_hidden_layers": 1, | |
| "transformers_version": "4.37.2", | |
| "use_transformer_mapping_network": true, | |
| "use_vision_encoder": true, | |
| "vision_config": { | |
| "dropout": 0.0, | |
| "hidden_act": "gelu", | |
| "hidden_size": 1664, | |
| "intermediate_size": 8192, | |
| "model_type": "flmr_vision_model", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 48, | |
| "patch_size": 14, | |
| "projection_dim": 1280 | |
| }, | |
| "vision_model_version": "laion/CLIP-ViT-bigG-14-laion2B-39B-b160k" | |
| } | |