Explicit vision_config
Browse files- config.json +20 -1
config.json
CHANGED
|
@@ -45,5 +45,24 @@
|
|
| 45 |
"use_cache": true,
|
| 46 |
"use_mm_proj": true,
|
| 47 |
"use_sliding_window": false,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 48 |
"vocab_size": 151936
|
| 49 |
-
}
|
|
|
|
| 45 |
"use_cache": true,
|
| 46 |
"use_mm_proj": true,
|
| 47 |
"use_sliding_window": false,
|
| 48 |
+
"vision_config": {
|
| 49 |
+
"cls_ratio": 2.0,
|
| 50 |
+
"down_patch_size": 7,
|
| 51 |
+
"down_stride": 2,
|
| 52 |
+
"downsamples": [true, true, true, true, true],
|
| 53 |
+
"embed_dims": [96, 192, 384, 768, 1536],
|
| 54 |
+
"hidden_size": 1024,
|
| 55 |
+
"image_size": 1024,
|
| 56 |
+
"intermediate_size": 3072,
|
| 57 |
+
"layer_scale_init_value": 1e-5,
|
| 58 |
+
"layers": [2, 12, 24, 4, 2],
|
| 59 |
+
"mlp_ratios": [4, 4, 4, 4, 4],
|
| 60 |
+
"num_classes": 1000,
|
| 61 |
+
"patch_size": 64,
|
| 62 |
+
"pos_embs_shapes": [null, null, null, [7, 7], [7, 7]],
|
| 63 |
+
"projection_dim": 768,
|
| 64 |
+
"repmixer_kernel_size": 3,
|
| 65 |
+
"token_mixers": ["repmixer", "repmixer", "repmixer", "attention", "attention"]
|
| 66 |
+
},
|
| 67 |
"vocab_size": 151936
|
| 68 |
+
}
|