Upload 2 files
Browse files- stepvit_config.json +17 -0
- stepvit_info.json +29 -0
stepvit_config.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"width": 1536,
|
| 3 |
+
"layers": 47,
|
| 4 |
+
"heads": 16,
|
| 5 |
+
"patch_size": 14,
|
| 6 |
+
"image_size": 728,
|
| 7 |
+
"mlp_ratio": 5.833333333333333,
|
| 8 |
+
"hidden_act": "quick_gelu",
|
| 9 |
+
"layer_norm_eps": 1e-05,
|
| 10 |
+
"ls_init_value": 0.1,
|
| 11 |
+
"use_cls_token": false,
|
| 12 |
+
"use_ln_pre": true,
|
| 13 |
+
"use_ln_post": false,
|
| 14 |
+
"use_abs_posemb": true,
|
| 15 |
+
"use_rope2d": true,
|
| 16 |
+
"rope_theta": 10000
|
| 17 |
+
}
|
stepvit_info.json
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "stepvit",
|
| 3 |
+
"source": "stepfun-ai/Step3-VL-10B",
|
| 4 |
+
"total_params": 1955848960,
|
| 5 |
+
"model_size_mb": 7823.6,
|
| 6 |
+
"architecture": {
|
| 7 |
+
"width": 1536,
|
| 8 |
+
"layers": 47,
|
| 9 |
+
"heads": 16,
|
| 10 |
+
"patch_size": 14,
|
| 11 |
+
"image_size": 728,
|
| 12 |
+
"mlp_ratio": 5.8333,
|
| 13 |
+
"hidden_act": "quick_gelu",
|
| 14 |
+
"ls_init_value": 0.1,
|
| 15 |
+
"use_rope2d": true,
|
| 16 |
+
"use_abs_posemb": true,
|
| 17 |
+
"use_cls_token": false
|
| 18 |
+
},
|
| 19 |
+
"features": [
|
| 20 |
+
"2D RoPE (rotary position embedding)",
|
| 21 |
+
"Absolute position embedding (bilinear interpolation)",
|
| 22 |
+
"LayerScale (init=0.1)",
|
| 23 |
+
"quick_gelu activation",
|
| 24 |
+
"No CLS token",
|
| 25 |
+
"2x spatial downsampler (Conv2d stride-2 x2)",
|
| 26 |
+
"Fused QKV projection",
|
| 27 |
+
"SDPA attention"
|
| 28 |
+
]
|
| 29 |
+
}
|