Upload folder using huggingface_hub

checkpoint_best.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb79d93443190b143a46300721cba4d74d695c7f4d2f0409487b57e9c604f1fb
+size 2650029567

evaluate.txt

@@ -0,0 +1,6 @@
+{"txt_r1": 77.26, "txt_r5": 93.48, "txt_r10": 96.86, "txt_r_mean": 89.2, "img_r1": 60.035985605757695, "img_r5": 83.80247900839665, "img_r10": 90.76369452219113, "img_r_mean": 78.20071971211514, "r_mean": 83.70035985605757, "agg_metrics": 89.2}
+{"txt_r1": 77.8, "txt_r5": 93.72, "txt_r10": 97.12, "txt_r_mean": 89.54666666666667, "img_r1": 60.519792083166735, "img_r5": 84.406237504998, "img_r10": 90.89964014394242, "img_r_mean": 78.60855657736904, "r_mean": 84.07761162201786, "agg_metrics": 89.54666666666667}
+{"txt_r1": 77.62, "txt_r5": 94.3, "txt_r10": 97.24, "txt_r_mean": 89.72000000000001, "img_r1": 60.59576169532187, "img_r5": 84.43822471011596, "img_r10": 91.05557776889245, "img_r_mean": 78.69652139144343, "r_mean": 84.20826069572172, "agg_metrics": 89.72000000000001}
+{"txt_r1": 78.56, "txt_r5": 94.1, "txt_r10": 97.24, "txt_r_mean": 89.96666666666665, "img_r1": 61.12754898040784, "img_r5": 84.5701719312275, "img_r10": 91.1595361855258, "img_r_mean": 78.95241903238706, "r_mean": 84.45954284952685, "agg_metrics": 89.96666666666665}
+{"txt_r1": 78.36, "txt_r5": 94.36, "txt_r10": 97.36, "txt_r_mean": 90.02666666666666, "img_r1": 61.10355857656937, "img_r5": 84.41423430627749, "img_r10": 91.06357457017194, "img_r_mean": 78.86045581767293, "r_mean": 84.44356124216979, "agg_metrics": 90.02666666666666}
+{"txt_r1": 77.46, "txt_r5": 94.16, "txt_r10": 97.52, "txt_r_mean": 89.71333333333332, "img_r1": 60.57177129148341, "img_r5": 84.12235105957618, "img_r10": 90.51179528188725, "img_r_mean": 78.4019725443156, "r_mean": 84.05765293882446, "agg_metrics": 89.71333333333332}

log.txt

@@ -0,0 +1,133 @@
+{
+    "run": {
+        "task": "retrieval",
+        "lr_sched": "linear_warmup_cosine_lr",
+        "init_lr": 1e-05,
+        "min_lr": 1e-06,
+        "weight_decay": 0.05,
+        "train_splits": [
+            "train"
+        ],
+        "valid_splits": [
+            "val"
+        ],
+        "test_splits": [
+            "test"
+        ],
+        "k_test": 256,
+        "output_dir": "./output/original/albef/retrieval_coco",
+        "max_epoch": 5,
+        "log_freq": 2000,
+        "num_workers": 8,
+        "batch_size_train": 32,
+        "batch_size_eval": 64,
+        "device": "cuda",
+        "world_size": 2,
+        "dist_url": "env://",
+        "distributed": true,
+        "use_dist_eval_sampler": false,
+        "seed": 42,
+        "amp": false,
+        "resume_ckpt_path": null,
+        "evaluate": false,
+        "rank": 0,
+        "gpu": 0,
+        "dist_backend": "nccl"
+    },
+    "model": {
+        "arch": "albef_retrieval",
+        "load_finetuned": false,
+        "pretrained": "https://storage.googleapis.com/sfr-pcl-data-research/ALBEF/ALBEF.pth",
+        "finetuned": "https://storage.googleapis.com/sfr-vision-language-research/LAVIS/models/ALBEF/albef_coco_retrieval_lavis.pt",
+        "queue_size": 65536,
+        "vit_type": "base",
+        "image_size": 384,
+        "vit_ckpt_layer": 0,
+        "vit_drop_path_rate": 0,
+        "vit_layer_norm_epsilon": 1e-06,
+        "vit_grad_ckpt": false,
+        "med_config_path": "configs/models/med_config_albef.json",
+        "embed_dim": 256,
+        "momentum": 0.995,
+        "alpha": 0.4,
+        "temp": 0.07,
+        "use_distill": true,
+        "max_txt_len": 30,
+        "model_type": "coco"
+    },
+    "preprocess": {
+        "vis_processor": {
+            "train": {
+                "name": "blip_image_train",
+                "image_size": 384
+            },
+            "eval": {
+                "name": "blip_image_eval",
+                "image_size": 384
+            }
+        },
+        "text_processor": {
+            "train": {
+                "name": "blip_caption"
+            },
+            "eval": {
+                "name": "blip_caption"
+            }
+        }
+    },
+    "datasets": {
+        "coco_retrieval": {
+            "data_type": "images",
+            "build_info": {
+                "annotations": {
+                    "train": {
+                        "url": "https://storage.googleapis.com/sfr-vision-language-research/datasets/coco_karpathy_train.json",
+                        "md5": "aa31ac474cf6250ebb81d18348a07ed8",
+                        "storage": "coco/annotations/coco_karpathy_train.json"
+                    },
+                    "val": {
+                        "url": "https://storage.googleapis.com/sfr-vision-language-research/datasets/coco_karpathy_val.json",
+                        "md5": "b273847456ef5580e33713b1f7de52a0",
+                        "storage": "coco/annotations/coco_karpathy_val.json"
+                    },
+                    "test": {
+                        "url": "https://storage.googleapis.com/sfr-vision-language-research/datasets/coco_karpathy_test.json",
+                        "md5": "3ff34b0ef2db02d01c37399f6a2a6cd1",
+                        "storage": "coco/annotations/coco_karpathy_test.json"
+                    }
+                },
+                "images": {
+                    "storage": "coco/images/"
+                }
+            },
+            "vis_processor": {
+                "train": {
+                    "name": "blip_image_train",
+                    "image_size": 384
+                },
+                "eval": {
+                    "name": "blip_image_eval",
+                    "image_size": 384
+                }
+            },
+            "text_processor": {
+                "train": {
+                    "name": "blip_caption"
+                },
+                "eval": {
+                    "name": "blip_caption"
+                }
+            }
+        }
+    }
+}
+{"train_lr": "0.000", "train_loss": "2.636"}
+{"val_txt_r1": 77.26, "val_txt_r5": 93.48, "val_txt_r10": 96.86, "val_txt_r_mean": 89.2, "val_img_r1": 60.035985605757695, "val_img_r5": 83.80247900839665, "val_img_r10": 90.76369452219113, "val_img_r_mean": 78.20071971211514, "val_r_mean": 83.70035985605757, "val_agg_metrics": 89.2, "val_best_epoch": 0}
+{"train_lr": "0.000", "train_loss": "2.808"}
+{"val_txt_r1": 77.8, "val_txt_r5": 93.72, "val_txt_r10": 97.12, "val_txt_r_mean": 89.54666666666667, "val_img_r1": 60.519792083166735, "val_img_r5": 84.406237504998, "val_img_r10": 90.89964014394242, "val_img_r_mean": 78.60855657736904, "val_r_mean": 84.07761162201786, "val_agg_metrics": 89.54666666666667, "val_best_epoch": 1}
+{"train_lr": "0.000", "train_loss": "2.842"}
+{"val_txt_r1": 77.62, "val_txt_r5": 94.3, "val_txt_r10": 97.24, "val_txt_r_mean": 89.72000000000001, "val_img_r1": 60.59576169532187, "val_img_r5": 84.43822471011596, "val_img_r10": 91.05557776889245, "val_img_r_mean": 78.69652139144343, "val_r_mean": 84.20826069572172, "val_agg_metrics": 89.72000000000001, "val_best_epoch": 2}
+{"train_lr": "0.000", "train_loss": "2.833"}
+{"val_txt_r1": 78.56, "val_txt_r5": 94.1, "val_txt_r10": 97.24, "val_txt_r_mean": 89.96666666666665, "val_img_r1": 61.12754898040784, "val_img_r5": 84.5701719312275, "val_img_r10": 91.1595361855258, "val_img_r_mean": 78.95241903238706, "val_r_mean": 84.45954284952685, "val_agg_metrics": 89.96666666666665, "val_best_epoch": 3}
+{"train_lr": "0.000", "train_loss": "2.738"}
+{"val_txt_r1": 78.36, "val_txt_r5": 94.36, "val_txt_r10": 97.36, "val_txt_r_mean": 90.02666666666666, "val_img_r1": 61.10355857656937, "val_img_r5": 84.41423430627749, "val_img_r10": 91.06357457017194, "val_img_r_mean": 78.86045581767293, "val_r_mean": 84.44356124216979, "val_agg_metrics": 90.02666666666666, "val_best_epoch": 4}