Slot Q-Former init
Browse files- config.json +61 -0
- pytorch_model.bin +3 -0
config.json
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"image_size": 256,
|
| 3 |
+
"slot_num": 32,
|
| 4 |
+
"codebook_embed_dim": 32,
|
| 5 |
+
"n_embed": 8192,
|
| 6 |
+
"num_quantizers": 4,
|
| 7 |
+
"blocks_layers": 4,
|
| 8 |
+
"blocks_image_layers": 4,
|
| 9 |
+
"use_blocks_image": true,
|
| 10 |
+
"bypass_codebook": false,
|
| 11 |
+
"use_causal": true,
|
| 12 |
+
"use_slot": true,
|
| 13 |
+
"slot_config": {
|
| 14 |
+
"T": 1,
|
| 15 |
+
"num_iterations": 3,
|
| 16 |
+
"use_half_slot": false
|
| 17 |
+
},
|
| 18 |
+
"checkpoint_path": {
|
| 19 |
+
"model_path": null
|
| 20 |
+
},
|
| 21 |
+
"stage1": {
|
| 22 |
+
"dino_model_name": "dinov2_vitl14",
|
| 23 |
+
"unfreeze_unet": true,
|
| 24 |
+
"unfreeze_resnet": false,
|
| 25 |
+
"image_size": 256,
|
| 26 |
+
"loss_weight": {
|
| 27 |
+
"loss_itc": 0.5,
|
| 28 |
+
"loss_lm": 0.5,
|
| 29 |
+
"loss_diffusion": 1,
|
| 30 |
+
"loss_mse": 0.5
|
| 31 |
+
},
|
| 32 |
+
"use_causal": true,
|
| 33 |
+
"use_slot": true,
|
| 34 |
+
"slot_config": {
|
| 35 |
+
"T": 1,
|
| 36 |
+
"num_iterations": 3,
|
| 37 |
+
"use_half_slot": false
|
| 38 |
+
}
|
| 39 |
+
},
|
| 40 |
+
"stage2": {
|
| 41 |
+
"loss_weight": {
|
| 42 |
+
"loss_codebook": 1,
|
| 43 |
+
"loss_recon": 1,
|
| 44 |
+
"loss_diffusion": 0.1,
|
| 45 |
+
"loss_mse": 0.1
|
| 46 |
+
},
|
| 47 |
+
"unfreeze_unet": false,
|
| 48 |
+
"unfreeze_linear": false,
|
| 49 |
+
"blocks_layers": 4,
|
| 50 |
+
"blocks_image_layers": 4,
|
| 51 |
+
"use_blocks_image": true,
|
| 52 |
+
"unclip": false,
|
| 53 |
+
"vq": {
|
| 54 |
+
"vq_type": "residual_vq",
|
| 55 |
+
"num_quantizers": 4,
|
| 56 |
+
"codebook_embed_dim": 32,
|
| 57 |
+
"n_embed": 8192
|
| 58 |
+
},
|
| 59 |
+
"bypass_codebook": false
|
| 60 |
+
}
|
| 61 |
+
}
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ac8921b44fe3a0a7979b90988d6b24ae143ffe1bbb0958ac21b8bc4792ecd86b
|
| 3 |
+
size 7599744974
|