Upload folder using huggingface_hub
Browse files- config.json +65 -0
- model.safetensors +3 -0
- preprocessor_config.json +18 -0
- tokenizer_config.json +5 -0
config.json
ADDED
|
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"PS3Model"
|
| 4 |
+
],
|
| 5 |
+
"model_type": "ps3",
|
| 6 |
+
"vision_config": {
|
| 7 |
+
"architectures": [
|
| 8 |
+
"PS3VisionModel"
|
| 9 |
+
],
|
| 10 |
+
"model_type": "ps3_vision_model",
|
| 11 |
+
"model_name": "vit_so400m_patch14_siglip_384",
|
| 12 |
+
"hidden_size": 1152,
|
| 13 |
+
"pool": "map",
|
| 14 |
+
"s3_scales": [
|
| 15 |
+
378,
|
| 16 |
+
756,
|
| 17 |
+
1512,
|
| 18 |
+
3780
|
| 19 |
+
],
|
| 20 |
+
"select_based_on_layer": [
|
| 21 |
+
0,
|
| 22 |
+
9,
|
| 23 |
+
18,
|
| 24 |
+
26
|
| 25 |
+
],
|
| 26 |
+
"min_select_num": 1,
|
| 27 |
+
"max_select_num": 2560,
|
| 28 |
+
"seperate_pos_emb": true,
|
| 29 |
+
"highres_selection_feature": true,
|
| 30 |
+
"radio": false,
|
| 31 |
+
"radio_adapter_mlp_version": null,
|
| 32 |
+
"radio_adapter_mlp_input_dim": null,
|
| 33 |
+
"radio_adapter_mlp_hidden_dim": null,
|
| 34 |
+
"radio_adapter_mlp_output_dim": null,
|
| 35 |
+
"radio_adapter_mlp_num_inner": null,
|
| 36 |
+
"img_size": null,
|
| 37 |
+
"drop": 0.0,
|
| 38 |
+
"class_token": null,
|
| 39 |
+
"final_norm": false
|
| 40 |
+
},
|
| 41 |
+
"text_config": {
|
| 42 |
+
"context_length": 64,
|
| 43 |
+
"vocab_size": 32000,
|
| 44 |
+
"hf_tokenizer_name": "timm/ViT-B-16-SigLIP",
|
| 45 |
+
"tokenizer_kwargs": {
|
| 46 |
+
"clean": "canonicalize"
|
| 47 |
+
},
|
| 48 |
+
"width": 1152,
|
| 49 |
+
"heads": 16,
|
| 50 |
+
"layers": 27,
|
| 51 |
+
"mlp_ratio": 3.7362,
|
| 52 |
+
"no_causal_mask": true,
|
| 53 |
+
"proj_bias": true,
|
| 54 |
+
"pool_type": "last",
|
| 55 |
+
"norm_kwargs": {
|
| 56 |
+
"eps": 1e-06
|
| 57 |
+
},
|
| 58 |
+
"architectures": [
|
| 59 |
+
"PS3TextModel"
|
| 60 |
+
],
|
| 61 |
+
"model_type": "ps3_text_model",
|
| 62 |
+
"output_dim": 1152,
|
| 63 |
+
"prompt_proj_dim": 1152
|
| 64 |
+
}
|
| 65 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b6394d69e47b2ece97130d604b11c19198ccff9d4f248d9a96c1866b2db643b4
|
| 3 |
+
size 4112866144
|
preprocessor_config.json
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"image_size": [
|
| 3 |
+
3780,
|
| 4 |
+
3780
|
| 5 |
+
],
|
| 6 |
+
"mean": [
|
| 7 |
+
0.5,
|
| 8 |
+
0.5,
|
| 9 |
+
0.5
|
| 10 |
+
],
|
| 11 |
+
"std": [
|
| 12 |
+
0.5,
|
| 13 |
+
0.5,
|
| 14 |
+
0.5
|
| 15 |
+
],
|
| 16 |
+
"interpolation": "bicubic",
|
| 17 |
+
"resize_mode": "squash"
|
| 18 |
+
}
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"tokenizer_name": "timm/ViT-B-16-SigLIP",
|
| 3 |
+
"context_length": 64,
|
| 4 |
+
"clean": "canonicalize"
|
| 5 |
+
}
|