Upload config.json with huggingface_hub

config.json

@@ -0,0 +1,48 @@
+{
+  "flickr_image_root": "/gfs/shared/public/datasets/flickr",
+  "flickr_ann_root": "/gfs/shared/public/datasets/flickr",
+  "coco_image_root": "/gfs/shared/public/datasets/albef/coco/images",
+  "coco_ann_root": "/gfs/shared/public/datasets/albef/json_finetune",
+  "pretrain_files": [
+    "/gfs/shared/public/datasets/albef/ccs_synthetic_filtered_large_local_cp.json",
+    "/gfs/shared/public/datasets/albef/json_pretrain/json_pretrain/vg.json",
+    "/gfs/shared/public/datasets/albef/json_pretrain/json_pretrain/coco.json"
+  ],
+  "finetune_files": [
+    "/gfs/shared/public/datasets/albef/json_pretrain/json_pretrain/coco.json"
+  ],
+  "zs_dataset": "/gfs/shared/public/datasets/albef/vg_large_objects.json",
+  "batch_size": 32,
+  "data": "",
+  "teacher_siglip_path": "google/siglip2-large-patch16-384",
+  "siglip_path": "google/siglip2-large-patch16-384",
+  "language_model_path": "microsoft/MiniLM-L12-H384-uncased",
+  "multimodal_projection_hidden_dim": 8192,
+  "num_negatives_per_sample": 3,
+  "num_hard_negatives": 0,
+  "num_compressed_tokens": 64,
+  "weight_decay": 0.05,
+  "pretraining_lr": 0.0003,
+  "finetune_lr": 0.0003,
+  "min_lr": 1e-06,
+  "warmup_lr": 1e-06,
+  "lr_decay_rate": 0.9,
+  "max_epoch": 45,
+  "warmup_steps": 100,
+  "k": 10,
+  "test_set": "flickr",
+  "imagenet_classnames_path": "/gfs/shared/public/datasets/imagenet/classnames.txt",
+  "templates_path": "zeroshot_classification_templates.txt",
+  "experiment": {
+    "backend": "comet",
+    "project": "efficient-fusion",
+    "entity": "mitchellkt",
+    "run_name": "multimodal-retrieval-training"
+  },
+  "dry_mode": false,
+  "finetune": true,
+  "init_lr": 0.0003,
+  "train_file": [
+    "/gfs/shared/public/datasets/albef/json_pretrain/json_pretrain/coco.json"
+  ]
+}