Add config from convert_d_fine_original_pytorch_checkpoint_to_hf.py
Browse files- config.json +57 -7
config.json
CHANGED
|
@@ -1,17 +1,31 @@
|
|
| 1 |
{
|
|
|
|
| 2 |
"activation_dropout": 0.0,
|
| 3 |
"activation_function": "silu",
|
| 4 |
"anchor_image_size": null,
|
| 5 |
-
"architectures": [
|
| 6 |
-
"DFineForObjectDetection"
|
| 7 |
-
],
|
| 8 |
"attention_dropout": 0.0,
|
| 9 |
"auxiliary_loss": true,
|
| 10 |
"backbone": null,
|
| 11 |
"backbone_config": {
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
"embedding_size": 32,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
"layer_type": "basic",
|
| 14 |
-
"model_type": "
|
|
|
|
| 15 |
"out_features": [
|
| 16 |
"stage2",
|
| 17 |
"stage3",
|
|
@@ -22,29 +36,67 @@
|
|
| 22 |
3,
|
| 23 |
4
|
| 24 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
"stage_in_channels": [
|
| 26 |
64,
|
| 27 |
128,
|
| 28 |
512,
|
| 29 |
1024
|
| 30 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
"stage_mid_channels": [
|
| 32 |
64,
|
| 33 |
128,
|
| 34 |
256,
|
| 35 |
512
|
| 36 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
"stage_num_blocks": [
|
| 38 |
1,
|
| 39 |
2,
|
| 40 |
5,
|
| 41 |
2
|
| 42 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
"stem_channels": [
|
| 44 |
3,
|
| 45 |
32,
|
| 46 |
64
|
| 47 |
-
]
|
|
|
|
| 48 |
},
|
| 49 |
"backbone_kwargs": null,
|
| 50 |
"batch_norm_eps": 1e-05,
|
|
@@ -67,7 +119,6 @@
|
|
| 67 |
],
|
| 68 |
"decoder_offset_scale": 0.5,
|
| 69 |
"depth_mult": 1.0,
|
| 70 |
-
"disable_custom_kernels": true,
|
| 71 |
"dropout": 0.0,
|
| 72 |
"encode_proj_layers": [
|
| 73 |
2
|
|
@@ -281,7 +332,6 @@
|
|
| 281 |
"positional_encoding_temperature": 10000,
|
| 282 |
"reg_scale": 4.0,
|
| 283 |
"top_prob_values": 4,
|
| 284 |
-
"torch_dtype": "float32",
|
| 285 |
"transformers_version": "4.50.0.dev0",
|
| 286 |
"use_focal_loss": true,
|
| 287 |
"use_pretrained_backbone": false,
|
|
|
|
| 1 |
{
|
| 2 |
+
"_attn_implementation_autoset": true,
|
| 3 |
"activation_dropout": 0.0,
|
| 4 |
"activation_function": "silu",
|
| 5 |
"anchor_image_size": null,
|
|
|
|
|
|
|
|
|
|
| 6 |
"attention_dropout": 0.0,
|
| 7 |
"auxiliary_loss": true,
|
| 8 |
"backbone": null,
|
| 9 |
"backbone_config": {
|
| 10 |
+
"depths": [
|
| 11 |
+
3,
|
| 12 |
+
4,
|
| 13 |
+
6,
|
| 14 |
+
3
|
| 15 |
+
],
|
| 16 |
+
"downsample_in_bottleneck": false,
|
| 17 |
+
"downsample_in_first_stage": false,
|
| 18 |
"embedding_size": 32,
|
| 19 |
+
"hidden_act": "relu",
|
| 20 |
+
"hidden_sizes": [
|
| 21 |
+
256,
|
| 22 |
+
512,
|
| 23 |
+
1024,
|
| 24 |
+
2048
|
| 25 |
+
],
|
| 26 |
"layer_type": "basic",
|
| 27 |
+
"model_type": "hgnet_v2",
|
| 28 |
+
"num_channels": 3,
|
| 29 |
"out_features": [
|
| 30 |
"stage2",
|
| 31 |
"stage3",
|
|
|
|
| 36 |
3,
|
| 37 |
4
|
| 38 |
],
|
| 39 |
+
"stage_downsample": [
|
| 40 |
+
false,
|
| 41 |
+
true,
|
| 42 |
+
true,
|
| 43 |
+
true
|
| 44 |
+
],
|
| 45 |
"stage_in_channels": [
|
| 46 |
64,
|
| 47 |
128,
|
| 48 |
512,
|
| 49 |
1024
|
| 50 |
],
|
| 51 |
+
"stage_kernel_size": [
|
| 52 |
+
3,
|
| 53 |
+
3,
|
| 54 |
+
5,
|
| 55 |
+
5
|
| 56 |
+
],
|
| 57 |
+
"stage_light_block": [
|
| 58 |
+
false,
|
| 59 |
+
false,
|
| 60 |
+
true,
|
| 61 |
+
true
|
| 62 |
+
],
|
| 63 |
"stage_mid_channels": [
|
| 64 |
64,
|
| 65 |
128,
|
| 66 |
256,
|
| 67 |
512
|
| 68 |
],
|
| 69 |
+
"stage_names": [
|
| 70 |
+
"stem",
|
| 71 |
+
"stage1",
|
| 72 |
+
"stage2",
|
| 73 |
+
"stage3",
|
| 74 |
+
"stage4"
|
| 75 |
+
],
|
| 76 |
"stage_num_blocks": [
|
| 77 |
1,
|
| 78 |
2,
|
| 79 |
5,
|
| 80 |
2
|
| 81 |
],
|
| 82 |
+
"stage_numb_of_layers": [
|
| 83 |
+
6,
|
| 84 |
+
6,
|
| 85 |
+
6,
|
| 86 |
+
6
|
| 87 |
+
],
|
| 88 |
+
"stage_out_channels": [
|
| 89 |
+
128,
|
| 90 |
+
512,
|
| 91 |
+
1024,
|
| 92 |
+
2048
|
| 93 |
+
],
|
| 94 |
"stem_channels": [
|
| 95 |
3,
|
| 96 |
32,
|
| 97 |
64
|
| 98 |
+
],
|
| 99 |
+
"use_learnable_affine_block": false
|
| 100 |
},
|
| 101 |
"backbone_kwargs": null,
|
| 102 |
"batch_norm_eps": 1e-05,
|
|
|
|
| 119 |
],
|
| 120 |
"decoder_offset_scale": 0.5,
|
| 121 |
"depth_mult": 1.0,
|
|
|
|
| 122 |
"dropout": 0.0,
|
| 123 |
"encode_proj_layers": [
|
| 124 |
2
|
|
|
|
| 332 |
"positional_encoding_temperature": 10000,
|
| 333 |
"reg_scale": 4.0,
|
| 334 |
"top_prob_values": 4,
|
|
|
|
| 335 |
"transformers_version": "4.50.0.dev0",
|
| 336 |
"use_focal_loss": true,
|
| 337 |
"use_pretrained_backbone": false,
|