prince-canuma commited on
Commit
c49a90a
·
verified ·
1 Parent(s): 12a2351

Upload folder using huggingface_hub

Browse files
README.md ADDED
@@ -0,0 +1,53 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ - zh
5
+ license: mit
6
+ pipeline_tag: automatic-speech-recognition
7
+ tags:
8
+ - ASR
9
+ - Transcriptoin
10
+ - Diarization
11
+ - Speech-to-Text
12
+ - mlx
13
+ - speech-to-text
14
+ - speech
15
+ - transcription
16
+ - asr
17
+ - stt
18
+ - mlx-audio
19
+ library_name: mlx-audio
20
+ ---
21
+ # mlx-community/VibeVoice-ASR-6bit
22
+
23
+ This model was converted to MLX format from [`microsoft/VibeVoice-ASR`](https://huggingface.co/microsoft/VibeVoice-ASR) using mlx-audio version **0.3.0rc1**.
24
+
25
+ Refer to the [original model card](https://huggingface.co/microsoft/VibeVoice-ASR) for more details on the model.
26
+
27
+ ## Use with mlx-audio
28
+
29
+ ```bash
30
+ pip install -U mlx-audio
31
+ ```
32
+
33
+ ### CLI Example:
34
+ ```bash
35
+ python -m mlx_audio.stt.generate --model mlx-community/VibeVoice-ASR-6bit --audio "audio.wav"
36
+ ```
37
+
38
+ ### Python Example:
39
+ ```python
40
+ from mlx_audio.stt.utils import load_model
41
+ from mlx_audio.stt.generate import generate_transcription
42
+
43
+ model = load_model("mlx-community/VibeVoice-ASR-6bit")
44
+ transcription = generate_transcription(
45
+ model=model,
46
+ audio_path="path_to_audio.wav",
47
+ output_path="path_to_output.txt",
48
+ format="txt",
49
+ verbose=True,
50
+ )
51
+ print(transcription.text)
52
+
53
+ ```
config.json ADDED
@@ -0,0 +1,158 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": [
12
+ 8,
13
+ 5,
14
+ 5,
15
+ 4,
16
+ 2,
17
+ 2
18
+ ],
19
+ "disable_last_norm": true,
20
+ "dtype": "bfloat16",
21
+ "encoder_depths": "3-3-3-3-3-3-8",
22
+ "encoder_n_filters": 32,
23
+ "encoder_ratios": [
24
+ 8,
25
+ 5,
26
+ 5,
27
+ 4,
28
+ 2,
29
+ 2
30
+ ],
31
+ "fix_std": 0.5,
32
+ "layer_scale_init_value": 1e-06,
33
+ "layernorm": "RMSNorm",
34
+ "layernorm_elementwise_affine": true,
35
+ "layernorm_eps": 1e-05,
36
+ "mixer_layer": "depthwise_conv",
37
+ "model_type": "vibevoice_acoustic_tokenizer",
38
+ "pad_mode": "constant",
39
+ "std_dist_type": "gaussian",
40
+ "vae_dim": 64,
41
+ "weight_init_value": 0.01
42
+ },
43
+ "acoustic_vae_dim": 64,
44
+ "architectures": [
45
+ "VibeVoiceForASRTraining"
46
+ ],
47
+ "decoder_config": {
48
+ "attention_dropout": 0.0,
49
+ "dtype": "bfloat16",
50
+ "hidden_act": "silu",
51
+ "hidden_size": 3584,
52
+ "initializer_range": 0.02,
53
+ "intermediate_size": 18944,
54
+ "layer_types": [
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
+ "full_attention",
67
+ "full_attention",
68
+ "full_attention",
69
+ "full_attention",
70
+ "full_attention",
71
+ "full_attention",
72
+ "full_attention",
73
+ "full_attention",
74
+ "full_attention",
75
+ "full_attention",
76
+ "full_attention",
77
+ "full_attention",
78
+ "full_attention",
79
+ "full_attention",
80
+ "full_attention",
81
+ "full_attention",
82
+ "full_attention"
83
+ ],
84
+ "max_position_embeddings": 131072,
85
+ "max_window_layers": 28,
86
+ "model_type": "qwen2",
87
+ "num_attention_heads": 28,
88
+ "num_hidden_layers": 28,
89
+ "num_key_value_heads": 4,
90
+ "rms_norm_eps": 1e-06,
91
+ "rope_scaling": null,
92
+ "rope_theta": 1000000.0,
93
+ "sliding_window": null,
94
+ "use_cache": true,
95
+ "use_mrope": false,
96
+ "use_sliding_window": false,
97
+ "vocab_size": 152064
98
+ },
99
+ "diffusion_head_config": {
100
+ "ddpm_batch_mul": 4,
101
+ "ddpm_beta_schedule": "cosine",
102
+ "ddpm_num_inference_steps": 20,
103
+ "ddpm_num_steps": 1000,
104
+ "diffusion_type": "ddpm",
105
+ "head_ffn_ratio": 3.0,
106
+ "head_layers": 4,
107
+ "hidden_size": 3584,
108
+ "latent_size": 64,
109
+ "model_type": "vibepod_diffusion_head",
110
+ "prediction_type": "v_prediction",
111
+ "rms_norm_eps": 1e-05,
112
+ "speech_vae_dim": 64
113
+ },
114
+ "dtype": "float32",
115
+ "model_type": "vibevoice_asr",
116
+ "quantization": {
117
+ "group_size": 64,
118
+ "bits": 6,
119
+ "mode": "affine"
120
+ },
121
+ "quantization_config": {
122
+ "group_size": 64,
123
+ "bits": 6,
124
+ "mode": "affine"
125
+ },
126
+ "semantic_tokenizer_config": {
127
+ "causal": true,
128
+ "channels": 1,
129
+ "conv_bias": true,
130
+ "conv_norm": "none",
131
+ "corpus_normalize": 0.0,
132
+ "disable_last_norm": true,
133
+ "dtype": "bfloat16",
134
+ "encoder_depths": "3-3-3-3-3-3-8",
135
+ "encoder_n_filters": 32,
136
+ "encoder_ratios": [
137
+ 8,
138
+ 5,
139
+ 5,
140
+ 4,
141
+ 2,
142
+ 2
143
+ ],
144
+ "fix_std": 0,
145
+ "layer_scale_init_value": 1e-06,
146
+ "layernorm": "RMSNorm",
147
+ "layernorm_elementwise_affine": true,
148
+ "layernorm_eps": 1e-05,
149
+ "mixer_layer": "depthwise_conv",
150
+ "model_type": "vibevoice_semantic_tokenizer",
151
+ "pad_mode": "constant",
152
+ "std_dist_type": "none",
153
+ "vae_dim": 128,
154
+ "weight_init_value": 0.01
155
+ },
156
+ "semantic_vae_dim": 128,
157
+ "transformers_version": "4.57.6"
158
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eb034800ed0b1d94469fa7445f0df429d9fc6e3cf6e511330b188b71cad4463a
3
+ size 5360114475
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5890215e0574892fd36be7357645685657761cc540cdae32d1a48096d168a0f
3
+ size 2257554265
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff