| { |
| "architectures": [ |
| "FlavaForPreTraining" |
| ], |
| "ce_ignore_index": -100, |
| "dtype": "float32", |
| "global_backprop_contrastive": true, |
| "global_contrastive_weight": 1.0, |
| "hidden_size": 32, |
| "image_codebook_config": { |
| "freeze": true, |
| "hidden_size": 32, |
| "initializer_range": 0.02, |
| "input_channels": 3, |
| "model_type": "", |
| "num_blocks_per_group": 2, |
| "num_groups": 2, |
| "vocab_size": 99 |
| }, |
| "image_config": { |
| "attention_probs_dropout_prob": 0.0, |
| "hidden_act": "gelu", |
| "hidden_dropout_prob": 0.0, |
| "hidden_size": 32, |
| "image_size": 30, |
| "initializer_range": 0.02, |
| "intermediate_size": 37, |
| "layer_norm_eps": 1e-12, |
| "mask_token": true, |
| "model_type": "flava_image_model", |
| "num_attention_heads": 4, |
| "num_channels": 3, |
| "num_hidden_layers": 2, |
| "patch_size": 2, |
| "qkv_bias": true, |
| "vocab_size": 99 |
| }, |
| "init_codebook": true, |
| "initializer_factor": 1.0, |
| "initializer_range": 0.02, |
| "itm_weight": 1.0, |
| "layer_norm_eps": 1e-12, |
| "logit_scale_init_value": 2.6592, |
| "mim_weight": 1.0, |
| "mlm_weight": 1.0, |
| "mmm_image_weight": 1.0, |
| "mmm_text_weight": 1.0, |
| "model_type": "flava", |
| "multimodal_config": { |
| "attention_probs_dropout_prob": 0.0, |
| "ce_ignore_index": -100, |
| "hidden_act": "gelu", |
| "hidden_dropout_prob": 0.0, |
| "hidden_size": 32, |
| "initializer_range": 0.02, |
| "intermediate_size": 37, |
| "layer_norm_eps": 1e-12, |
| "model_type": "flava_multimodal_model", |
| "num_attention_heads": 4, |
| "num_hidden_layers": 2, |
| "qkv_bias": true, |
| "use_cls_token": true |
| }, |
| "projection_dim": 32, |
| "return_loss": true, |
| "skip_unmasked_multimodal_encoder": true, |
| "text_config": { |
| "attention_probs_dropout_prob": 0.0, |
| "hidden_act": "gelu", |
| "hidden_dropout_prob": 0.0, |
| "hidden_size": 32, |
| "initializer_range": 0.02, |
| "intermediate_size": 37, |
| "layer_norm_eps": 1e-12, |
| "max_position_embeddings": 512, |
| "model_type": "flava_text_model", |
| "num_attention_heads": 4, |
| "num_hidden_layers": 2, |
| "pad_token_id": 0, |
| "qkv_bias": true, |
| "type_vocab_size": 2, |
| "vocab_size": 30522 |
| }, |
| "tie_word_embeddings": true, |
| "transformers_version": "5.16.0.dev0" |
| } |
|
|
