Upload config.json with huggingface_hub
Browse files- config.json +78 -0
config.json
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"aligner_config": {
|
| 3 |
+
"cls": "MlpProjector",
|
| 4 |
+
"model_type": "aligner",
|
| 5 |
+
"params": {
|
| 6 |
+
"depth": 2,
|
| 7 |
+
"input_dim": 1024,
|
| 8 |
+
"n_embed": 4096,
|
| 9 |
+
"projector_type": "mlp_gelu"
|
| 10 |
+
}
|
| 11 |
+
},
|
| 12 |
+
"architectures": [
|
| 13 |
+
"JanusLLamaModel"
|
| 14 |
+
],
|
| 15 |
+
"gen_aligner_config": {
|
| 16 |
+
"cls": "MlpProjector",
|
| 17 |
+
"model_type": "gen_aligner",
|
| 18 |
+
"params": {
|
| 19 |
+
"depth": 2,
|
| 20 |
+
"input_dim": 8,
|
| 21 |
+
"n_embed": 4096,
|
| 22 |
+
"projector_type": "mlp_gelu"
|
| 23 |
+
}
|
| 24 |
+
},
|
| 25 |
+
"gen_head_config": {
|
| 26 |
+
"cls": "vision_head",
|
| 27 |
+
"model_type": "gen_head",
|
| 28 |
+
"params": {
|
| 29 |
+
"image_token_embed": 4096,
|
| 30 |
+
"image_token_size": 16384,
|
| 31 |
+
"n_embed": 4096
|
| 32 |
+
}
|
| 33 |
+
},
|
| 34 |
+
"gen_vision_config": {
|
| 35 |
+
"cls": "VQ-16",
|
| 36 |
+
"model_type": "gen_vision",
|
| 37 |
+
"params": {
|
| 38 |
+
"image_token_size": 16384,
|
| 39 |
+
"n_embed": 8
|
| 40 |
+
}
|
| 41 |
+
},
|
| 42 |
+
"language_config": {
|
| 43 |
+
"_attn_implementation_autoset": true,
|
| 44 |
+
"attention_bias": false,
|
| 45 |
+
"attention_dropout": 0.0,
|
| 46 |
+
"head_dim": 128,
|
| 47 |
+
"hidden_act": "silu",
|
| 48 |
+
"hidden_size": 4096,
|
| 49 |
+
"initializer_range": 0.02,
|
| 50 |
+
"intermediate_size": 11008,
|
| 51 |
+
"max_position_embeddings": 1576,
|
| 52 |
+
"mlp_bias": false,
|
| 53 |
+
"model_type": "llama",
|
| 54 |
+
"num_attention_heads": 32,
|
| 55 |
+
"num_hidden_layers": 30,
|
| 56 |
+
"num_key_value_heads": 32,
|
| 57 |
+
"pretraining_tp": 1,
|
| 58 |
+
"rms_norm_eps": 1e-06,
|
| 59 |
+
"rope_scaling": null,
|
| 60 |
+
"rope_theta": 10000.0,
|
| 61 |
+
"torch_dtype": "bfloat16",
|
| 62 |
+
"use_cache": true,
|
| 63 |
+
"vocab_size": 102400
|
| 64 |
+
},
|
| 65 |
+
"model_type": "multi_modality",
|
| 66 |
+
"torch_dtype": "bfloat16",
|
| 67 |
+
"transformers_version": "4.50.0",
|
| 68 |
+
"vision_config": {
|
| 69 |
+
"cls": "CLIPVisionTower",
|
| 70 |
+
"model_type": "vision",
|
| 71 |
+
"params": {
|
| 72 |
+
"image_size": 384,
|
| 73 |
+
"model_name": "siglip_large_patch16_384",
|
| 74 |
+
"select_feature": "same",
|
| 75 |
+
"select_layer": -1
|
| 76 |
+
}
|
| 77 |
+
}
|
| 78 |
+
}
|