Update tiny models for CHMv2ForDepthEstimation
Browse files- config.json +70 -0
- model.safetensors +3 -0
- preprocessor_config.json +27 -0
config.json
ADDED
|
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"CHMv2ForDepthEstimation"
|
| 4 |
+
],
|
| 5 |
+
"backbone_config": {
|
| 6 |
+
"apply_layernorm": true,
|
| 7 |
+
"attention_dropout": 0.0,
|
| 8 |
+
"drop_path_rate": 0.0,
|
| 9 |
+
"hidden_act": "gelu",
|
| 10 |
+
"hidden_size": 32,
|
| 11 |
+
"image_size": 32,
|
| 12 |
+
"initializer_range": 0.02,
|
| 13 |
+
"intermediate_size": 64,
|
| 14 |
+
"key_bias": true,
|
| 15 |
+
"layer_norm_eps": 1e-06,
|
| 16 |
+
"layerscale_value": 1.0,
|
| 17 |
+
"mlp_bias": true,
|
| 18 |
+
"model_type": "dinov3_vit",
|
| 19 |
+
"num_attention_heads": 4,
|
| 20 |
+
"num_channels": 3,
|
| 21 |
+
"num_hidden_layers": 2,
|
| 22 |
+
"num_register_tokens": 0,
|
| 23 |
+
"out_features": [
|
| 24 |
+
"stage1",
|
| 25 |
+
"stage2"
|
| 26 |
+
],
|
| 27 |
+
"out_indices": [
|
| 28 |
+
1,
|
| 29 |
+
2
|
| 30 |
+
],
|
| 31 |
+
"patch_size": 16,
|
| 32 |
+
"pos_embed_jitter": null,
|
| 33 |
+
"pos_embed_rescale": 2.0,
|
| 34 |
+
"pos_embed_shift": null,
|
| 35 |
+
"proj_bias": true,
|
| 36 |
+
"query_bias": true,
|
| 37 |
+
"reshape_hidden_states": true,
|
| 38 |
+
"return_class_token": true,
|
| 39 |
+
"rope_theta": 100.0,
|
| 40 |
+
"stage_names": [
|
| 41 |
+
"stem",
|
| 42 |
+
"stage1",
|
| 43 |
+
"stage2"
|
| 44 |
+
],
|
| 45 |
+
"use_gated_mlp": false,
|
| 46 |
+
"value_bias": true
|
| 47 |
+
},
|
| 48 |
+
"bins_strategy": "chmv2_mixlog",
|
| 49 |
+
"dtype": "float32",
|
| 50 |
+
"fusion_hidden_size": 16,
|
| 51 |
+
"head_hidden_size": 16,
|
| 52 |
+
"initializer_range": 0.02,
|
| 53 |
+
"max_depth": 96.0,
|
| 54 |
+
"min_depth": 0.001,
|
| 55 |
+
"model_type": "chmv2",
|
| 56 |
+
"norm_strategy": "chmv2_mixlog",
|
| 57 |
+
"number_output_channels": 4,
|
| 58 |
+
"patch_size": 16,
|
| 59 |
+
"post_process_channels": [
|
| 60 |
+
16,
|
| 61 |
+
16
|
| 62 |
+
],
|
| 63 |
+
"readout_type": "project",
|
| 64 |
+
"reassemble_factors": [
|
| 65 |
+
4,
|
| 66 |
+
2
|
| 67 |
+
],
|
| 68 |
+
"reassemble_hidden_size": 32,
|
| 69 |
+
"transformers_version": "5.16.0.dev0"
|
| 70 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3fd9255d5141e0f0f25ab85c41f764d234323baabf262ea391d780a8614fedf4
|
| 3 |
+
size 303376
|
preprocessor_config.json
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"data_format": "channels_first",
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_pad": true,
|
| 5 |
+
"do_rescale": true,
|
| 6 |
+
"do_resize": false,
|
| 7 |
+
"ensure_multiple_of": 16,
|
| 8 |
+
"image_mean": [
|
| 9 |
+
0.42,
|
| 10 |
+
0.411,
|
| 11 |
+
0.296
|
| 12 |
+
],
|
| 13 |
+
"image_processor_type": "CHMv2ImageProcessor",
|
| 14 |
+
"image_std": [
|
| 15 |
+
0.213,
|
| 16 |
+
0.156,
|
| 17 |
+
0.143
|
| 18 |
+
],
|
| 19 |
+
"keep_aspect_ratio": true,
|
| 20 |
+
"resample": 3,
|
| 21 |
+
"rescale_factor": 0.00392156862745098,
|
| 22 |
+
"size": {
|
| 23 |
+
"height": 384,
|
| 24 |
+
"width": 384
|
| 25 |
+
},
|
| 26 |
+
"size_divisor": 16
|
| 27 |
+
}
|