Hugging Face's logo

Zuellni
/
VibeVoice-Backup

Safetensors
Model card Files
xet
Community
Zuellni commited on
Commit
e25dbc8
·
verified ·
1 Parent(s): 04ce3f2

Upload 6 files

Browse files
Files changed (6) hide show
  1. large/config.json +127 -0
  2. large/model.safetensors +3 -0
  3. large/preprocessor_config.json +13 -0
  4. smol/config.json +126 -0
  5. smol/model.safetensors +3 -0
  6. smol/preprocessor_config.json +13 -0
large/config.json ADDED
@@ -0,0 +1,127 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_attn_implementation_autoset": false,
3
+ "acoustic_tokenizer_config": {
4
+ "causal": true,
5
+ "channels": 1,
6
+ "conv_bias": true,
7
+ "conv_norm": "none",
8
+ "corpus_normalize": 0.0,
9
+ "decoder_depths": null,
10
+ "decoder_n_filters": 32,
11
+ "decoder_ratios": [8, 5, 5, 4, 2, 2],
12
+ "disable_last_norm": true,
13
+ "encoder_depths": "3-3-3-3-3-3-8",
14
+ "encoder_n_filters": 32,
15
+ "encoder_ratios": [8, 5, 5, 4, 2, 2],
16
+ "fix_std": 0.5,
17
+ "layer_scale_init_value": 1e-6,
18
+ "layernorm": "RMSNorm",
19
+ "layernorm_elementwise_affine": true,
20
+ "layernorm_eps": 1e-5,
21
+ "mixer_layer": "depthwise_conv",
22
+ "model_type": "vibevoice_acoustic_tokenizer",
23
+ "pad_mode": "constant",
24
+ "std_dist_type": "gaussian",
25
+ "torch_dtype": "bfloat16",
26
+ "vae_dim": 64,
27
+ "weight_init_value": 0.01
28
+ },
29
+ "acoustic_vae_dim": 64,
30
+ "architectures": ["VibeVoiceForConditionalGenerationInference"],
31
+ "decoder_config": {
32
+ "attention_dropout": 0.0,
33
+ "hidden_act": "silu",
34
+ "hidden_size": 3584,
35
+ "initializer_range": 0.02,
36
+ "intermediate_size": 18944,
37
+ "layer_types": [
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention",
41
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention",
45
+ "full_attention",
46
+ "full_attention",
47
+ "full_attention",
48
+ "full_attention",
49
+ "full_attention",
50
+ "full_attention",
51
+ "full_attention",
52
+ "full_attention",
53
+ "full_attention",
54
+ "full_attention",
55
+ "full_attention",
56
+ "full_attention",
57
+ "full_attention",
58
+ "full_attention",
59
+ "full_attention",
60
+ "full_attention",
61
+ "full_attention",
62
+ "full_attention",
63
+ "full_attention",
64
+ "full_attention",
65
+ "full_attention"
66
+ ],
67
+ "max_position_embeddings": 32768,
68
+ "max_window_layers": 28,
69
+ "model_type": "qwen2",
70
+ "num_attention_heads": 28,
71
+ "num_hidden_layers": 28,
72
+ "num_key_value_heads": 4,
73
+ "rms_norm_eps": 1e-6,
74
+ "rope_scaling": null,
75
+ "rope_theta": 1000000.0,
76
+ "sliding_window": null,
77
+ "torch_dtype": "bfloat16",
78
+ "use_cache": true,
79
+ "use_mrope": false,
80
+ "use_sliding_window": false,
81
+ "vocab_size": 152064
82
+ },
83
+ "diffusion_head_config": {
84
+ "ddpm_batch_mul": 4,
85
+ "ddpm_beta_schedule": "cosine",
86
+ "ddpm_num_inference_steps": 20,
87
+ "ddpm_num_steps": 1000,
88
+ "diffusion_type": "ddpm",
89
+ "head_ffn_ratio": 3.0,
90
+ "head_layers": 4,
91
+ "hidden_size": 3584,
92
+ "latent_size": 64,
93
+ "model_type": "vibevoice_diffusion_head",
94
+ "prediction_type": "v_prediction",
95
+ "rms_norm_eps": 1e-5,
96
+ "speech_vae_dim": 64,
97
+ "torch_dtype": "bfloat16"
98
+ },
99
+ "model_type": "vibevoice",
100
+ "semantic_tokenizer_config": {
101
+ "causal": true,
102
+ "channels": 1,
103
+ "conv_bias": true,
104
+ "conv_norm": "none",
105
+ "corpus_normalize": 0.0,
106
+ "disable_last_norm": true,
107
+ "encoder_depths": "3-3-3-3-3-3-8",
108
+ "encoder_n_filters": 32,
109
+ "encoder_ratios": [8, 5, 5, 4, 2, 2],
110
+ "fix_std": 0,
111
+ "layer_scale_init_value": 1e-6,
112
+ "layernorm": "RMSNorm",
113
+ "layernorm_elementwise_affine": true,
114
+ "layernorm_eps": 1e-5,
115
+ "mixer_layer": "depthwise_conv",
116
+ "model_type": "vibevoice_semantic_tokenizer",
117
+ "pad_mode": "constant",
118
+ "std_dist_type": "none",
119
+ "torch_dtype": "bfloat16",
120
+ "vae_dim": 128,
121
+ "weight_init_value": 0.01
122
+ },
123
+ "semantic_vae_dim": 128,
124
+ "tie_word_embeddings": false,
125
+ "torch_dtype": "bfloat16",
126
+ "transformers_version": "4.55.4"
127
+ }
large/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:523ba74883bc862e2963b996ba8236f76c5d4aac0deb97d9649a4859dfa56af2
3
+ size 18686872118
large/preprocessor_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "audio_processor": {
3
+ "eps": 1e-6,
4
+ "feature_extractor_type": "VibeVoiceTokenizerProcessor",
5
+ "normalize_audio": true,
6
+ "sampling_rate": 24000,
7
+ "target_dB_FS": -25
8
+ },
9
+ "db_normalize": true,
10
+ "language_model_pretrained_name": "Qwen/Qwen2.5-7B",
11
+ "processor_class": "VibeVoiceProcessor",
12
+ "speech_tok_compress_ratio": 3200
13
+ }
smol/config.json ADDED
@@ -0,0 +1,126 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_attn_implementation_autoset": false,
3
+ "acoustic_tokenizer_config": {
4
+ "causal": true,
5
+ "channels": 1,
6
+ "conv_bias": true,
7
+ "conv_norm": "none",
8
+ "corpus_normalize": 0.0,
9
+ "decoder_depths": null,
10
+ "decoder_n_filters": 32,
11
+ "decoder_ratios": [8, 5, 5, 4, 2, 2],
12
+ "disable_last_norm": true,
13
+ "encoder_depths": "3-3-3-3-3-3-8",
14
+ "encoder_n_filters": 32,
15
+ "encoder_ratios": [8, 5, 5, 4, 2, 2],
16
+ "fix_std": 0.5,
17
+ "layer_scale_init_value": 1e-6,
18
+ "layernorm": "RMSNorm",
19
+ "layernorm_elementwise_affine": true,
20
+ "layernorm_eps": 1e-5,
21
+ "mixer_layer": "depthwise_conv",
22
+ "model_type": "vibevoice_acoustic_tokenizer",
23
+ "pad_mode": "constant",
24
+ "std_dist_type": "gaussian",
25
+ "torch_dtype": "bfloat16",
26
+ "vae_dim": 64,
27
+ "weight_init_value": 0.01
28
+ },
29
+ "acoustic_vae_dim": 64,
30
+ "architectures": ["VibeVoiceForConditionalGenerationInference"],
31
+ "decoder_config": {
32
+ "attention_dropout": 0.0,
33
+ "hidden_act": "silu",
34
+ "hidden_size": 1536,
35
+ "initializer_range": 0.02,
36
+ "intermediate_size": 8960,
37
+ "layer_types": [
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention",
41
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention",
45
+ "full_attention",
46
+ "full_attention",
47
+ "full_attention",
48
+ "full_attention",
49
+ "full_attention",
50
+ "full_attention",
51
+ "full_attention",
52
+ "full_attention",
53
+ "full_attention",
54
+ "full_attention",
55
+ "full_attention",
56
+ "full_attention",
57
+ "full_attention",
58
+ "full_attention",
59
+ "full_attention",
60
+ "full_attention",
61
+ "full_attention",
62
+ "full_attention",
63
+ "full_attention",
64
+ "full_attention",
65
+ "full_attention"
66
+ ],
67
+ "max_position_embeddings": 65536,
68
+ "max_window_layers": 28,
69
+ "model_type": "qwen2",
70
+ "num_attention_heads": 12,
71
+ "num_hidden_layers": 28,
72
+ "num_key_value_heads": 2,
73
+ "rms_norm_eps": 1e-6,
74
+ "rope_scaling": null,
75
+ "rope_theta": 1000000.0,
76
+ "sliding_window": null,
77
+ "tie_word_embeddings": true,
78
+ "torch_dtype": "bfloat16",
79
+ "use_cache": true,
80
+ "use_sliding_window": false,
81
+ "vocab_size": 151936
82
+ },
83
+ "diffusion_head_config": {
84
+ "ddpm_batch_mul": 4,
85
+ "ddpm_beta_schedule": "cosine",
86
+ "ddpm_num_inference_steps": 20,
87
+ "ddpm_num_steps": 1000,
88
+ "diffusion_type": "ddpm",
89
+ "head_ffn_ratio": 3.0,
90
+ "head_layers": 4,
91
+ "hidden_size": 1536,
92
+ "latent_size": 64,
93
+ "model_type": "vibevoice_diffusion_head",
94
+ "prediction_type": "v_prediction",
95
+ "rms_norm_eps": 1e-5,
96
+ "speech_vae_dim": 64,
97
+ "torch_dtype": "bfloat16"
98
+ },
99
+ "model_type": "vibevoice",
100
+ "semantic_tokenizer_config": {
101
+ "causal": true,
102
+ "channels": 1,
103
+ "conv_bias": true,
104
+ "conv_norm": "none",
105
+ "corpus_normalize": 0.0,
106
+ "disable_last_norm": true,
107
+ "encoder_depths": "3-3-3-3-3-3-8",
108
+ "encoder_n_filters": 32,
109
+ "encoder_ratios": [8, 5, 5, 4, 2, 2],
110
+ "fix_std": 0,
111
+ "layer_scale_init_value": 1e-6,
112
+ "layernorm": "RMSNorm",
113
+ "layernorm_elementwise_affine": true,
114
+ "layernorm_eps": 1e-5,
115
+ "mixer_layer": "depthwise_conv",
116
+ "model_type": "vibevoice_semantic_tokenizer",
117
+ "pad_mode": "constant",
118
+ "std_dist_type": "none",
119
+ "torch_dtype": "bfloat16",
120
+ "vae_dim": 128,
121
+ "weight_init_value": 0.01
122
+ },
123
+ "semantic_vae_dim": 128,
124
+ "torch_dtype": "bfloat16",
125
+ "transformers_version": "4.55.4"
126
+ }
smol/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:73cdcd09359e004fea721f77535d152b8061e61e541cdad7671100b2cf23897a
3
+ size 5408203142
smol/preprocessor_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "audio_processor": {
3
+ "eps": 1e-6,
4
+ "feature_extractor_type": "VibeVoiceTokenizerProcessor",
5
+ "normalize_audio": true,
6
+ "sampling_rate": 24000,
7
+ "target_dB_FS": -25
8
+ },
9
+ "db_normalize": true,
10
+ "language_model_pretrained_name": "Qwen/Qwen2.5-1.5B",
11
+ "processor_class": "VibeVoiceProcessor",
12
+ "speech_tok_compress_ratio": 3200
13
+ }