Upload 6 files
Browse files- coco_grounded/config.json +54 -0
- coco_grounded/model.pt +3 -0
- grit/config.json +31 -0
- grit/model.pt +3 -0
- grit_ft_coco_grounded/config.json +54 -0
- grit_ft_coco_grounded/model.pt +3 -0
coco_grounded/config.json
ADDED
|
@@ -0,0 +1,54 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"overview": "Running on main branch",
|
| 3 |
+
"save_embed": true,
|
| 4 |
+
"embed_dir": "cache/embeds_old",
|
| 5 |
+
"result_dir": "results",
|
| 6 |
+
"datasets": {
|
| 7 |
+
"COCOCaptionGrounded": {
|
| 8 |
+
"root": "datasets/coco-2017",
|
| 9 |
+
"split": "train",
|
| 10 |
+
"grounded_dir": "datasets/COCOCaptionGrounded",
|
| 11 |
+
"shuffle_bbox": false,
|
| 12 |
+
"crop_augment": true,
|
| 13 |
+
"caption_augment": true,
|
| 14 |
+
"ignore_caption_id_file": "assets/nsr_val_coco_train_overlap.json"
|
| 15 |
+
},
|
| 16 |
+
"COCOCaptionGroundedSpatial": {
|
| 17 |
+
"coco_root": "datasets/coco-2017",
|
| 18 |
+
"nsr_root": "datasets/COCOCaptionGroundedSpatial",
|
| 19 |
+
"split": "train",
|
| 20 |
+
"shuffle_bbox": false,
|
| 21 |
+
"crop_augment": true,
|
| 22 |
+
"data_augmentation": true,
|
| 23 |
+
"use_gt_bboxs": true
|
| 24 |
+
},
|
| 25 |
+
"NSR1KSpatial": {
|
| 26 |
+
"scale_factor": 10,
|
| 27 |
+
"coco_root": "datasets/coco-2017",
|
| 28 |
+
"nsr_root": "datasets/NSR-1K",
|
| 29 |
+
"split": "train",
|
| 30 |
+
"shuffle_bbox": false,
|
| 31 |
+
"crop_augment": true,
|
| 32 |
+
"data_augmentation": true
|
| 33 |
+
}
|
| 34 |
+
},
|
| 35 |
+
"concept_embedder_batch_size": 256,
|
| 36 |
+
"caption_embedder_batch_size": 256,
|
| 37 |
+
"model": "DiT-S",
|
| 38 |
+
"in_channel": 4,
|
| 39 |
+
"concept_in_channel": 768,
|
| 40 |
+
"y_in_channel": 768,
|
| 41 |
+
"max_in_len": 60,
|
| 42 |
+
"max_y_len": 120,
|
| 43 |
+
"scale": 2.0,
|
| 44 |
+
"noise_schedule": "linear",
|
| 45 |
+
"layout_type": "xyxy",
|
| 46 |
+
"diffusion_steps": 100,
|
| 47 |
+
"epochs": 400,
|
| 48 |
+
"global_batch_size": 256,
|
| 49 |
+
"global_seed": 0,
|
| 50 |
+
"num_workers": 4,
|
| 51 |
+
"log_every": 100,
|
| 52 |
+
"ckpt_every": 25000,
|
| 53 |
+
"t5_size": "base"
|
| 54 |
+
}
|
coco_grounded/model.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c4b4f7bf85f6ff2835953ec7aa460df70899720062b67538d2ce0000f8da2f81
|
| 3 |
+
size 73431836
|
grit/config.json
ADDED
|
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"overview": "Running on main branch",
|
| 3 |
+
"save_embed": true,
|
| 4 |
+
"embed_dir": "cache/pretrain_embeds_grit",
|
| 5 |
+
"result_dir": "results",
|
| 6 |
+
"datasets": {
|
| 7 |
+
"GRIT": {
|
| 8 |
+
"path": "datasets/GRIT/grit-20m",
|
| 9 |
+
"crop_augment": true
|
| 10 |
+
}
|
| 11 |
+
},
|
| 12 |
+
"concept_embedder_batch_size": 256,
|
| 13 |
+
"caption_embedder_batch_size": 256,
|
| 14 |
+
"model": "DiT-S",
|
| 15 |
+
"in_channel": 4,
|
| 16 |
+
"concept_in_channel": 768,
|
| 17 |
+
"y_in_channel": 768,
|
| 18 |
+
"max_in_len": 60,
|
| 19 |
+
"max_y_len": 120,
|
| 20 |
+
"scale": 2.0,
|
| 21 |
+
"noise_schedule": "linear",
|
| 22 |
+
"layout_type": "xyxy",
|
| 23 |
+
"diffusion_steps": 100,
|
| 24 |
+
"epochs": 400,
|
| 25 |
+
"global_batch_size": 256,
|
| 26 |
+
"global_seed": 0,
|
| 27 |
+
"num_workers": 12,
|
| 28 |
+
"log_every": 100,
|
| 29 |
+
"ckpt_every": 25000,
|
| 30 |
+
"t5_size": "base"
|
| 31 |
+
}
|
grit/model.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ed9b34b9c8f19327169e6f6188a5ff215c4f42acef0a0bb7bab6b616da055988
|
| 3 |
+
size 73420623
|
grit_ft_coco_grounded/config.json
ADDED
|
@@ -0,0 +1,54 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"overview": "Running on main branch",
|
| 3 |
+
"save_embed": true,
|
| 4 |
+
"embed_dir": "cache/embeds_old",
|
| 5 |
+
"result_dir": "results",
|
| 6 |
+
"datasets": {
|
| 7 |
+
"COCOCaptionGrounded": {
|
| 8 |
+
"root": "datasets/coco-2017",
|
| 9 |
+
"split": "train",
|
| 10 |
+
"grounded_dir": "datasets/COCOCaptionGrounded",
|
| 11 |
+
"shuffle_bbox": false,
|
| 12 |
+
"crop_augment": true,
|
| 13 |
+
"caption_augment": true,
|
| 14 |
+
"ignore_caption_id_file": "assets/nsr_val_coco_train_overlap.json"
|
| 15 |
+
},
|
| 16 |
+
"COCOCaptionGroundedSpatial": {
|
| 17 |
+
"coco_root": "datasets/coco-2017",
|
| 18 |
+
"nsr_root": "datasets/COCOCaptionGroundedSpatial",
|
| 19 |
+
"split": "train",
|
| 20 |
+
"shuffle_bbox": false,
|
| 21 |
+
"crop_augment": true,
|
| 22 |
+
"data_augmentation": true,
|
| 23 |
+
"use_gt_bboxs": true
|
| 24 |
+
},
|
| 25 |
+
"NSR1KSpatial": {
|
| 26 |
+
"scale_factor": 10,
|
| 27 |
+
"coco_root": "datasets/coco-2017",
|
| 28 |
+
"nsr_root": "datasets/NSR-1K",
|
| 29 |
+
"split": "train",
|
| 30 |
+
"shuffle_bbox": false,
|
| 31 |
+
"crop_augment": true,
|
| 32 |
+
"data_augmentation": true
|
| 33 |
+
}
|
| 34 |
+
},
|
| 35 |
+
"concept_embedder_batch_size": 256,
|
| 36 |
+
"caption_embedder_batch_size": 256,
|
| 37 |
+
"model": "DiT-S",
|
| 38 |
+
"in_channel": 4,
|
| 39 |
+
"concept_in_channel": 768,
|
| 40 |
+
"y_in_channel": 768,
|
| 41 |
+
"max_in_len": 60,
|
| 42 |
+
"max_y_len": 120,
|
| 43 |
+
"scale": 2.0,
|
| 44 |
+
"noise_schedule": "linear",
|
| 45 |
+
"layout_type": "xyxy",
|
| 46 |
+
"diffusion_steps": 100,
|
| 47 |
+
"epochs": 400,
|
| 48 |
+
"global_batch_size": 256,
|
| 49 |
+
"global_seed": 0,
|
| 50 |
+
"num_workers": 4,
|
| 51 |
+
"log_every": 100,
|
| 52 |
+
"ckpt_every": 25000,
|
| 53 |
+
"t5_size": "base"
|
| 54 |
+
}
|
grit_ft_coco_grounded/model.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:af27bc9ed0613d5f49d2e50e38e816621664d740bf24de04fcbc7c0941bc461e
|
| 3 |
+
size 73420623
|