{ "baseline_beta": 1.0, "baseline_type": "ema", "bos_token_id": 1, "eos_token_id": 2, "eval_legacy": true, "image_feature_scale_num": 1, "max_length": 4096, "mode": 3, "num_layers": 33, "out_dim": 256, "pad_token_id": 0, "pad_train_clip_images": true, "rej_token_idx": 32001, "resize_vision_tower": false, "resize_vision_tower_size": 224, "seg_token_idx": 32000, "seg_token_num": 1, "separate_mm_projector": true, "strategy": "policy_walker", "train_mask_decoder": true, "transformers_version": "4.31.0", "vision_tower": "../dataset_sesame/clip-vit-large-patch14-336", "vision_tower_for_mask": false }