Add converted Keras weights + kf_config + card
Browse files- kf_config.json +45 -0
kf_config.json
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"library_name": "kerasformers",
|
| 3 |
+
"kerasformers_version": "1.2.4",
|
| 4 |
+
"model_module": "kerasformers.models.internvl",
|
| 5 |
+
"model_class": "InternVLConditionalGenerate",
|
| 6 |
+
"variant": "internvl3-1b",
|
| 7 |
+
"weights": "model.weights.json",
|
| 8 |
+
"schema_version": 2,
|
| 9 |
+
"weight_dtype": "bfloat16",
|
| 10 |
+
"model_type": "internvl",
|
| 11 |
+
"text_config": {
|
| 12 |
+
"vocab_size": 151674,
|
| 13 |
+
"embed_dim": 896,
|
| 14 |
+
"mlp_dim": 4864,
|
| 15 |
+
"num_layers": 24,
|
| 16 |
+
"num_heads": 14,
|
| 17 |
+
"num_kv_heads": 2,
|
| 18 |
+
"head_dim": null,
|
| 19 |
+
"text_backbone": "qwen2",
|
| 20 |
+
"norm_eps": 1e-06,
|
| 21 |
+
"rope_theta": 1000000.0,
|
| 22 |
+
"tie_embeddings": false,
|
| 23 |
+
"num_experts": 0,
|
| 24 |
+
"num_experts_per_tok": 0,
|
| 25 |
+
"moe_mlp_dim": 0,
|
| 26 |
+
"norm_topk_prob": true,
|
| 27 |
+
"decoder_sparse_step": 1,
|
| 28 |
+
"mlp_only_layers": []
|
| 29 |
+
},
|
| 30 |
+
"vision_config": {
|
| 31 |
+
"embed_dim": 1024,
|
| 32 |
+
"mlp_dim": 4096,
|
| 33 |
+
"num_layers": 24,
|
| 34 |
+
"num_heads": 16,
|
| 35 |
+
"image_size": 448,
|
| 36 |
+
"patch_size": 14,
|
| 37 |
+
"attention_bias": true,
|
| 38 |
+
"qk_norm": false,
|
| 39 |
+
"norm_type": "layer_norm",
|
| 40 |
+
"norm_eps": 1e-06,
|
| 41 |
+
"layer_scale_init": 0.1
|
| 42 |
+
},
|
| 43 |
+
"downsample_ratio": 0.5,
|
| 44 |
+
"image_token_id": 151667
|
| 45 |
+
}
|