Update config.json
Browse files- config.json +49 -8
config.json
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
{
|
2 |
-
"_name_or_path": "
|
3 |
"architectures": [
|
4 |
-
"
|
5 |
],
|
6 |
"attention_dropout": 0.0,
|
7 |
"bos_token_id": 151645,
|
@@ -12,16 +12,55 @@
|
|
12 |
"image_aspect_ratio": "pad",
|
13 |
"initializer_range": 0.02,
|
14 |
"intermediate_size": 2816,
|
15 |
-
"language_model": "vilm/Quyen-SE-v0.1",
|
16 |
"max_position_embeddings": 32768,
|
17 |
"max_window_layers": 21,
|
18 |
"mm_hidden_size": 1152,
|
19 |
"mm_projector_lr": null,
|
20 |
"mm_projector_type": "mlp2x_gelu",
|
21 |
"mm_vision_tower": "google/siglip-so400m-patch14-384",
|
22 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
"num_attention_heads": 16,
|
24 |
"num_hidden_layers": 24,
|
|
|
25 |
"num_key_value_heads": 16,
|
26 |
"rms_norm_eps": 1e-06,
|
27 |
"rope_theta": 1000000.0,
|
@@ -29,10 +68,12 @@
|
|
29 |
"tie_word_embeddings": false,
|
30 |
"tokenizer_model_max_length": 4096,
|
31 |
"tokenizer_padding_side": "right",
|
32 |
-
"
|
|
|
33 |
"tune_mm_mlp_adapter": false,
|
34 |
-
"use_cache":
|
35 |
"use_mm_proj": true,
|
36 |
"use_sliding_window": false,
|
37 |
-
"vocab_size": 151936
|
38 |
-
|
|
|
|
1 |
{
|
2 |
+
"_name_or_path": "qnguyen3/nanoLLaVA-1.5",
|
3 |
"architectures": [
|
4 |
+
"BunnyQwenForCausalLM"
|
5 |
],
|
6 |
"attention_dropout": 0.0,
|
7 |
"bos_token_id": 151645,
|
|
|
12 |
"image_aspect_ratio": "pad",
|
13 |
"initializer_range": 0.02,
|
14 |
"intermediate_size": 2816,
|
|
|
15 |
"max_position_embeddings": 32768,
|
16 |
"max_window_layers": 21,
|
17 |
"mm_hidden_size": 1152,
|
18 |
"mm_projector_lr": null,
|
19 |
"mm_projector_type": "mlp2x_gelu",
|
20 |
"mm_vision_tower": "google/siglip-so400m-patch14-384",
|
21 |
+
"language_model": "vilm/Quyen-SE-v0.1",
|
22 |
+
"model_type": "llava",
|
23 |
+
"text_config": {
|
24 |
+
"_name_or_path": "vilm/Quyen-SE-v0.1",
|
25 |
+
"architectures": [
|
26 |
+
"Qwen2ForCausalLM"
|
27 |
+
],
|
28 |
+
"attention_dropout": 0.0,
|
29 |
+
"eos_token_id": 151645,
|
30 |
+
"hidden_act": "silu",
|
31 |
+
"hidden_size": 1024,
|
32 |
+
"initializer_range": 0.02,
|
33 |
+
"intermediate_size": 2816,
|
34 |
+
"max_position_embeddings": 32768,
|
35 |
+
"max_window_layers": 21,
|
36 |
+
"model_type": "qwen2",
|
37 |
+
"num_attention_heads": 16,
|
38 |
+
"num_hidden_layers": 24,
|
39 |
+
"num_image_tokens": 729,
|
40 |
+
"num_key_value_heads": 16,
|
41 |
+
"rms_norm_eps": 1e-06,
|
42 |
+
"rope_theta": 1000000.0,
|
43 |
+
"sliding_window": 4096,
|
44 |
+
"tie_word_embeddings": false,
|
45 |
+
"torch_dtype": "bfloat16",
|
46 |
+
"transformers_version": "4.38.1",
|
47 |
+
"use_cache": false,
|
48 |
+
"use_sliding_window": false,
|
49 |
+
"vocab_size": 151936
|
50 |
+
},
|
51 |
+
"vision_config": {
|
52 |
+
"hidden_size": 1152,
|
53 |
+
"image_size": 384,
|
54 |
+
"intermediate_size": 4304,
|
55 |
+
"model_type": "siglip_vision_model",
|
56 |
+
"num_attention_heads": 16,
|
57 |
+
"num_hidden_layers": 27,
|
58 |
+
"num_image_tokens": 729,
|
59 |
+
"patch_size": 14
|
60 |
+
},
|
61 |
"num_attention_heads": 16,
|
62 |
"num_hidden_layers": 24,
|
63 |
+
"num_image_tokens": 729,
|
64 |
"num_key_value_heads": 16,
|
65 |
"rms_norm_eps": 1e-06,
|
66 |
"rope_theta": 1000000.0,
|
|
|
68 |
"tie_word_embeddings": false,
|
69 |
"tokenizer_model_max_length": 4096,
|
70 |
"tokenizer_padding_side": "right",
|
71 |
+
"torch_dtype": "bfloat16",
|
72 |
+
"transformers_version": "4.39.2",
|
73 |
"tune_mm_mlp_adapter": false,
|
74 |
+
"use_cache": false,
|
75 |
"use_mm_proj": true,
|
76 |
"use_sliding_window": false,
|
77 |
+
"vocab_size": 151936,
|
78 |
+
"image_token_index": -200
|
79 |
+
}
|