Update open_clip_config.json
Browse files- open_clip_config.json +29 -1
open_clip_config.json
CHANGED
|
@@ -1,5 +1,33 @@
|
|
| 1 |
{
|
| 2 |
-
"model_cfg":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
"preprocess_cfg": {
|
| 4 |
"mean": [
|
| 5 |
0.5,
|
|
|
|
| 1 |
{
|
| 2 |
+
"model_cfg": {
|
| 3 |
+
"embed_dim": 1024,
|
| 4 |
+
"init_logit_bias": -10,
|
| 5 |
+
"custom_text": true,
|
| 6 |
+
"vision_cfg": {
|
| 7 |
+
"image_size": 256,
|
| 8 |
+
"timm_model_name": "vit_large_patch16_siglip_256",
|
| 9 |
+
"timm_model_pretrained": false,
|
| 10 |
+
"timm_pool": "map",
|
| 11 |
+
"timm_proj": "none"
|
| 12 |
+
},
|
| 13 |
+
"text_cfg": {
|
| 14 |
+
"context_length": 64,
|
| 15 |
+
"vocab_size": 32000,
|
| 16 |
+
"tokenizer_kwargs": {
|
| 17 |
+
"clean": "canonicalize"
|
| 18 |
+
},
|
| 19 |
+
"width": 1024,
|
| 20 |
+
"heads": 16,
|
| 21 |
+
"layers": 24,
|
| 22 |
+
"no_causal_mask": true,
|
| 23 |
+
"proj_bias": true,
|
| 24 |
+
"pool_type": "last",
|
| 25 |
+
"norm_kwargs": {
|
| 26 |
+
"eps": 1e-06
|
| 27 |
+
},
|
| 28 |
+
"hf_tokenizer_name": "timm/ViT-L-16-SigLIP-256"
|
| 29 |
+
}
|
| 30 |
+
},
|
| 31 |
"preprocess_cfg": {
|
| 32 |
"mean": [
|
| 33 |
0.5,
|