Hugging Face's logo

deepseek-community
/
Janus-Pro-7B

Any-to-Any
Transformers
Safetensors
janus
image-text-to-text
muiltimodal
text-to-image
unified-model
Model card Files
xet
Community
hugosilva664 commited on
Commit
6c27207
·
verified ·
1 Parent(s): d47afed

Upload config

Browse files
Files changed (1) hide show
  1. config.json +54 -10
config.json CHANGED
@@ -1,33 +1,77 @@
1
  {
2
- "architectures": [
3
- "JanusForConditionalGeneration"
4
- ],
5
  "image_token_index": 100594,
6
  "model_type": "janus",
7
  "text_config": {
 
 
8
  "bos_token_id": 100000,
9
  "eos_token_id": 100001,
 
 
10
  "hidden_size": 4096,
 
11
  "intermediate_size": 11008,
12
  "max_position_embeddings": 16384,
 
13
  "model_type": "llama",
14
  "num_attention_heads": 32,
15
  "num_hidden_layers": 30,
16
  "num_key_value_heads": 32,
17
  "pad_token_id": 100015,
 
 
 
 
18
  "torch_dtype": "bfloat16",
 
19
  "vocab_size": 102400
20
  },
21
- "torch_dtype": "bfloat16",
22
- "transformers_version": "4.50.0.dev0",
23
  "vision_config": {
24
- "aligner_projection_size": 4096,
25
- "model_type": "siglip_vision_model"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
26
  },
27
  "vq_config": {
28
- "aligner_projection_size": 4096,
29
- "image_token_embed_size": 4096,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
30
  "model_type": "janus_vqgan",
31
- "num_patches": 24
 
 
 
 
 
32
  }
33
  }
 
1
  {
 
 
 
2
  "image_token_index": 100594,
3
  "model_type": "janus",
4
  "text_config": {
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
  "bos_token_id": 100000,
8
  "eos_token_id": 100001,
9
+ "head_dim": 128,
10
+ "hidden_act": "silu",
11
  "hidden_size": 4096,
12
+ "initializer_range": 0.02,
13
  "intermediate_size": 11008,
14
  "max_position_embeddings": 16384,
15
+ "mlp_bias": false,
16
  "model_type": "llama",
17
  "num_attention_heads": 32,
18
  "num_hidden_layers": 30,
19
  "num_key_value_heads": 32,
20
  "pad_token_id": 100015,
21
+ "pretraining_tp": 1,
22
+ "rms_norm_eps": 1e-06,
23
+ "rope_scaling": null,
24
+ "rope_theta": 10000.0,
25
  "torch_dtype": "bfloat16",
26
+ "use_cache": true,
27
  "vocab_size": 102400
28
  },
29
+ "transformers_version": "4.52.0.dev0",
 
30
  "vision_config": {
31
+ "attention_bias": true,
32
+ "attention_dropout": 0.0,
33
+ "depth": 2,
34
+ "hidden_act": "gelu",
35
+ "hidden_dropout_rate": 0.0,
36
+ "hidden_size": 1024,
37
+ "image_size": 384,
38
+ "initializer_range": 0.02,
39
+ "layer_norm_eps": 1e-06,
40
+ "mlp_ratio": 4.0,
41
+ "model_type": "janus_vision_model",
42
+ "num_attention_heads": 16,
43
+ "num_channels": 3,
44
+ "num_hidden_layers": 24,
45
+ "num_image_tokens": 576,
46
+ "patch_size": 16,
47
+ "projection_dim": 4096,
48
+ "projection_dropout": 0.0,
49
+ "use_qk_norm": false
50
  },
51
  "vq_config": {
52
+ "base_channels": 128,
53
+ "channel_multiplier": [
54
+ 1,
55
+ 1,
56
+ 2,
57
+ 2,
58
+ 4
59
+ ],
60
+ "depth": 2,
61
+ "double_latent": false,
62
+ "dropout": 0.0,
63
+ "embed_dim": 8,
64
+ "hidden_act": "gelu",
65
+ "image_token_embed_dim": 4096,
66
+ "in_channels": 3,
67
+ "initializer_range": 0.02,
68
+ "latent_channels": 256,
69
  "model_type": "janus_vqgan",
70
+ "num_embeddings": 16384,
71
+ "num_hidden_layers": 2,
72
+ "num_patches": 24,
73
+ "num_res_blocks": 2,
74
+ "out_channels": 3,
75
+ "projection_dim": 4096
76
  }
77
  }