| { | |
| "architectures": [ | |
| "BaberuOCRModel" | |
| ], | |
| "attention_dropout": 0.0, | |
| "attn_logit_softcap": null, | |
| "bos_token_id": 1, | |
| "dtype": "float32", | |
| "eos_token_id": 2, | |
| "final_logit_softcap": 30.0, | |
| "freeze_vision_encoder": false, | |
| "head_dim": 64, | |
| "hidden_act": "silu", | |
| "hidden_size": 512, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 1536, | |
| "max_position_embeddings": 2048, | |
| "model_type": "baberu_ocr", | |
| "num_attention_heads": 8, | |
| "num_hidden_layers": 6, | |
| "num_key_value_heads": 2, | |
| "pad_token_id": 0, | |
| "projector_act": "gelu", | |
| "rms_norm_eps": 1e-06, | |
| "rope_theta": 10000.0, | |
| "sandwich_norm": true, | |
| "tie_word_embeddings": true, | |
| "transformers_version": "5.5.3", | |
| "unk_token_id": 3, | |
| "use_cache": true, | |
| "vision_hidden_size": 768, | |
| "vision_image_size": 224, | |
| "vision_model_name": "facebook/dinov2-base", | |
| "vision_num_tokens": 256, | |
| "vision_patch_size": 14, | |
| "vision_rope_2d": false, | |
| "vocab_size": 14630, | |
| "z_loss_weight": 0.0001 | |
| } | |