{
    "run": {
        "task": "image_text_pretrain",
        "lr_sched": "linear_warmup_cosine_lr",
        "init_lr": 0.0001,
        "min_lr": 8e-05,
        "warmup_lr": 1e-06,
        "weight_decay": 0.05,
        "max_epoch": 150,
        "batch_size_train": 120,
        "batch_size_eval": 3,
        "num_workers": 4,
        "warmup_steps": 500,
        "iters_per_epoch": 500,
        "seed": 42,
        "output_dir": "output/xraygpt_mimic_pretrain",
        "amp": true,
        "resume_ckpt_path": "24_10_22_08_14_39_epoch99.pth",
        "evaluate": false,
        "train_splits": [
            "train"
        ],
        "device": "cuda",
        "world_size": 4,
        "dist_url": "env://",
        "distributed": true,
        "rank": 0,
        "gpu": 0,
        "dist_backend": "nccl"
    },
    "model": {
        "arch": "mini_gpt4",
        "image_size": 224,
        "drop_path_rate": 0,
        "use_grad_checkpoint": false,
        "vit_precision": "fp16",
        "freeze_vit": true,
        "freeze_qformer": true,
        "num_query_token": 32,
        "llama_model": "Joycean0301/Llama-3.2-3B-Instruct-Medical-Conversational",
        "prompt": "",
        "model_type": "pretrain_vicuna"
    },
    "preprocess": {
        "vis_processor": {
            "train": {
                "name": "blip2_image_train",
                "image_size": 224
            },
            "eval": {
                "name": "blip2_image_eval",
                "image_size": 224
            }
        },
        "text_processor": {
            "train": {
                "name": "blip_caption"
            },
            "eval": {
                "name": "blip_caption"
            }
        }
    },
    "datasets": {
        "mimic": {
            "data_type": "images",
            "build_info": {
                "storage": "dataset/mimic"
            },
            "vis_processor": {
                "train": {
                    "name": "blip2_image_train",
                    "image_size": 224
                }
            },
            "text_processor": {
                "train": {
                    "name": "blip_caption"
                }
            }
        }
    }
}
{
    "run": {
        "task": "image_text_pretrain",
        "lr_sched": "linear_warmup_cosine_lr",
        "init_lr": 0.0001,
        "min_lr": 8e-05,
        "warmup_lr": 1e-06,
        "weight_decay": 0.05,
        "max_epoch": 150,
        "batch_size_train": 120,
        "batch_size_eval": 3,
        "num_workers": 4,
        "warmup_steps": 500,
        "iters_per_epoch": 500,
        "seed": 42,
        "output_dir": "output/xraygpt_mimic_pretrain",
        "amp": true,
        "resume_ckpt_path": "24_10_22_08_14_39_epoch99.pth",
        "evaluate": false,
        "train_splits": [
            "train"
        ],
        "device": "cuda",
        "world_size": 4,
        "dist_url": "env://",
        "distributed": true,
        "rank": 0,
        "gpu": 0,
        "dist_backend": "nccl"
    },
    "model": {
        "arch": "mini_gpt4",
        "image_size": 224,
        "drop_path_rate": 0,
        "use_grad_checkpoint": false,
        "vit_precision": "fp16",
        "freeze_vit": true,
        "freeze_qformer": true,
        "num_query_token": 32,
        "llama_model": "Joycean0301/Llama-3.2-3B-Instruct-Medical-Conversational",
        "prompt": "",
        "model_type": "pretrain_vicuna"
    },
    "preprocess": {
        "vis_processor": {
            "train": {
                "name": "blip2_image_train",
                "image_size": 224
            },
            "eval": {
                "name": "blip2_image_eval",
                "image_size": 224
            }
        },
        "text_processor": {
            "train": {
                "name": "blip_caption"
            },
            "eval": {
                "name": "blip_caption"
            }
        }
    },
    "datasets": {
        "mimic": {
            "data_type": "images",
            "build_info": {
                "storage": "dataset/mimic"
            },
            "vis_processor": {
                "train": {
                    "name": "blip2_image_train",
                    "image_size": 224
                }
            },
            "text_processor": {
                "train": {
                    "name": "blip_caption"
                }
            }
        }
    }
}
{"train_lr": "0.000085", "train_loss": "1.176420"}
{"train_lr": "0.000085", "train_loss": "1.175911"}
{"train_lr": "0.000085", "train_loss": "1.177007"}
{"train_lr": "0.000084", "train_loss": "1.174899"}
{"train_lr": "0.000084", "train_loss": "1.176167"}
{"train_lr": "0.000084", "train_loss": "1.175371"}
{"train_lr": "0.000084", "train_loss": "1.175089"}
{"train_lr": "0.000084", "train_loss": "1.174685"}
{"train_lr": "0.000084", "train_loss": "1.173965"}
{"train_lr": "0.000083", "train_loss": "1.173050"}
{"train_lr": "0.000083", "train_loss": "1.174454"}
{"train_lr": "0.000083", "train_loss": "1.174522"}
{"train_lr": "0.000083", "train_loss": "1.171317"}
{"train_lr": "0.000083", "train_loss": "1.173012"}
{"train_lr": "0.000083", "train_loss": "1.173456"}
{"train_lr": "0.000082", "train_loss": "1.172261"}
{"train_lr": "0.000082", "train_loss": "1.171457"}
{"train_lr": "0.000082", "train_loss": "1.172554"}
{"train_lr": "0.000082", "train_loss": "1.169906"}
{"train_lr": "0.000082", "train_loss": "1.171817"}
{"train_lr": "0.000082", "train_loss": "1.170531"}
{"train_lr": "0.000082", "train_loss": "1.170232"}
{"train_lr": "0.000082", "train_loss": "1.171015"}
{"train_lr": "0.000082", "train_loss": "1.169610"}
{"train_lr": "0.000081", "train_loss": "1.170178"}
{"train_lr": "0.000081", "train_loss": "1.169042"}
{"train_lr": "0.000081", "train_loss": "1.170149"}
{"train_lr": "0.000081", "train_loss": "1.168049"}
{"train_lr": "0.000081", "train_loss": "1.169198"}
{"train_lr": "0.000081", "train_loss": "1.169054"}
{"train_lr": "0.000081", "train_loss": "1.168359"}
{"train_lr": "0.000081", "train_loss": "1.168657"}
{"train_lr": "0.000081", "train_loss": "1.168694"}
{"train_lr": "0.000081", "train_loss": "1.168019"}
{"train_lr": "0.000081", "train_loss": "1.167548"}
{"train_lr": "0.000080", "train_loss": "1.165927"}
{"train_lr": "0.000080", "train_loss": "1.168800"}v
{"train_lr": "0.000084", "train_loss": "1.166641"}
{"train_lr": "0.000084", "train_loss": "1.165976"}
{"train_lr": "0.000084", "train_loss": "1.167195"}
{"train_lr": "0.000084", "train_loss": "1.165221"}
{"train_lr": "0.000084", "train_loss": "1.166641"}
{"train_lr": "0.000084", "train_loss": "1.165935"}
{"train_lr": "0.000084", "train_loss": "1.165807"}
{"train_lr": "0.000084", "train_loss": "1.165503"}
{"train_lr": "0.000083", "train_loss": "1.164865"}
{"train_lr": "0.000083", "train_loss": "1.164062"}
{"train_lr": "0.000083", "train_loss": "1.165597"}
{"train_lr": "0.000083", "train_loss": "1.165696"}
{"train_lr": "0.000083", "train_loss": "1.162653"}
{"train_lr": "0.000083", "train_loss": "1.164436"}
{"train_lr": "0.000083", "train_loss": "1.164961"}
{"train_lr": "0.000083", "train_loss": "1.163872"}
{"train_lr": "0.000083", "train_loss": "1.163161"}
{"train_lr": "0.000082", "train_loss": "1.164320"}
{"train_lr": "0.000082", "train_loss": "1.161806"}
{"train_lr": "0.000082", "train_loss": "1.163766"}
{"train_lr": "0.000082", "train_loss": "1.162590"}
{"train_lr": "0.000082", "train_loss": "1.162393"}
{"train_lr": "0.000082", "train_loss": "1.163239"}
{"train_lr": "0.000082", "train_loss": "1.161929"}