Upload config.json with huggingface_hub
Browse files- config.json +46 -0
config.json
ADDED
|
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"LlavaForConditionalGeneration"
|
| 4 |
+
],
|
| 5 |
+
"ignore_index": -100,
|
| 6 |
+
"image_token_index": 128257,
|
| 7 |
+
"model_type": "llava",
|
| 8 |
+
"pad_token_id": 128258,
|
| 9 |
+
"projector_hidden_act": "gelu",
|
| 10 |
+
"text_config": {
|
| 11 |
+
"_name_or_path": "./iter_39620_xtuner",
|
| 12 |
+
"architectures": [
|
| 13 |
+
"LlamaForCausalLM"
|
| 14 |
+
],
|
| 15 |
+
"bos_token_id": 128000,
|
| 16 |
+
"eos_token_id": 128001,
|
| 17 |
+
"intermediate_size": 14336,
|
| 18 |
+
"max_position_embeddings": 8192,
|
| 19 |
+
"model_type": "llama",
|
| 20 |
+
"num_key_value_heads": 8,
|
| 21 |
+
"rms_norm_eps": 1e-05,
|
| 22 |
+
"rope_theta": 500000.0,
|
| 23 |
+
"torch_dtype": "float16",
|
| 24 |
+
"vocab_size": 128320
|
| 25 |
+
},
|
| 26 |
+
"torch_dtype": "float16",
|
| 27 |
+
"transformers_version": "4.40.1",
|
| 28 |
+
"vision_config": {
|
| 29 |
+
"_name_or_path": "./iter_39620_visual_encoder",
|
| 30 |
+
"architectures": [
|
| 31 |
+
"CLIPVisionModel"
|
| 32 |
+
],
|
| 33 |
+
"dropout": 0.0,
|
| 34 |
+
"hidden_size": 1024,
|
| 35 |
+
"image_size": 336,
|
| 36 |
+
"intermediate_size": 4096,
|
| 37 |
+
"model_type": "clip_vision_model",
|
| 38 |
+
"num_attention_heads": 16,
|
| 39 |
+
"num_hidden_layers": 24,
|
| 40 |
+
"patch_size": 14,
|
| 41 |
+
"projection_dim": 768,
|
| 42 |
+
"torch_dtype": "float32"
|
| 43 |
+
},
|
| 44 |
+
"vision_feature_layer": -2,
|
| 45 |
+
"vision_feature_select_strategy": "default"
|
| 46 |
+
}
|