Hugging Face's logo

AITRADER
/
Voxtral-4B-TTS-2603-mxfp8

Text-to-Speech
Safetensors
MLX
mlx-audio
voxtral_tts
mistral-common
speech
speech generation
voice cloning
tts
8-bit precision
Model card Files
xet
 Community
AITRADER commited on
Commit
423744a
·
verified ·
1 Parent(s): 2054288

Upload folder using huggingface_hub

Browse files
Files changed (7) hide show
  1. .gitattributes +1 -0
  2. README.md +60 -0
  3. config.json +140 -0
  4. model.safetensors +3 -0
  5. model.safetensors.index.json +599 -0
  6. params.json +130 -0
  7. tekken.json +3 -0
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ tekken.json filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: mlx-audio
3
+ language:
4
+ - en
5
+ - fr
6
+ - es
7
+ - pt
8
+ - it
9
+ - nl
10
+ - de
11
+ - ar
12
+ - hi
13
+ license: cc-by-nc-4.0
14
+ inference: false
15
+ base_model:
16
+ - mistralai/Ministral-3-3B-Base-2512
17
+ extra_gated_description: If you want to learn more about how we process your personal
18
+ data, please read our <a href="https://mistral.ai/terms/">Privacy Policy</a>.
19
+ tags:
20
+ - mistral-common
21
+ - mlx
22
+ - text-to-speech
23
+ - speech
24
+ - speech generation
25
+ - voice cloning
26
+ - tts
27
+ - mlx-audio
28
+ pipeline_tag: text-to-speech
29
+ ---
30
+ # AITRADER/Voxtral-4B-TTS-2603-mxfp8
31
+
32
+ This model was converted to MLX format from [`mistralai/Voxtral-4B-TTS-2603`](https://huggingface.co/mistralai/Voxtral-4B-TTS-2603) using mlx-audio version **0.4.1**.
33
+
34
+ Refer to the [original model card](https://huggingface.co/mistralai/Voxtral-4B-TTS-2603) for more details on the model.
35
+
36
+ ## Use with mlx-audio
37
+
38
+ ```bash
39
+ pip install -U mlx-audio
40
+ ```
41
+
42
+ ### CLI Example:
43
+ ```bash
44
+ python -m mlx_audio.tts.generate --model AITRADER/Voxtral-4B-TTS-2603-mxfp8 --text "Hello, this is a test."
45
+ ```
46
+
47
+ ### Python Example:
48
+ ```python
49
+ from mlx_audio.tts.utils import load_model
50
+ from mlx_audio.tts.generate import generate_audio
51
+
52
+ model = load_model("AITRADER/Voxtral-4B-TTS-2603-mxfp8")
53
+ generate_audio(
54
+ model=model,
55
+ text="Hello, this is a test.",
56
+ ref_audio="path_to_audio.wav",
57
+ file_prefix="test_audio",
58
+ )
59
+
60
+ ```
config.json ADDED
@@ -0,0 +1,140 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "attention_type": "FLASH_ATTN_3",
3
+ "causal": true,
4
+ "checkpoint": false,
5
+ "context_parallel": 1,
6
+ "cust_bwd": false,
7
+ "dim": 3072,
8
+ "dp_replicate_size": 1,
9
+ "dropout": 0.0,
10
+ "fp8_matmul": false,
11
+ "freeze_nonembedding": false,
12
+ "fsdp2": true,
13
+ "fsdp_optimize_backward_concat_if_pp": true,
14
+ "fused_rms_norm": true,
15
+ "head_dim": 128,
16
+ "hidden_dim": 9216,
17
+ "init": "NO_INIT",
18
+ "is_sequence_parallel": false,
19
+ "learnable_sinks": false,
20
+ "max_concurrent_tokens": 65536,
21
+ "max_position_embeddings": 128000,
22
+ "max_seq_len": 65536,
23
+ "model_parallel": 1,
24
+ "model_pipelining": 1,
25
+ "model_type": "voxtral_tts",
26
+ "multimodal": {
27
+ "bos_token_id": 1,
28
+ "audio_model_args": {
29
+ "semantic_codebook_size": 8192,
30
+ "acoustic_codebook_size": 21,
31
+ "n_acoustic_codebook": 36,
32
+ "audio_encoding_args": {
33
+ "codebook_pattern": "parallel",
34
+ "interleave_audio_tokens_per_segment": 8192,
35
+ "interleave_text_tokens_per_segment": 8192,
36
+ "single_trailing_segment": false,
37
+ "num_codebooks": 37,
38
+ "sampling_rate": 24000,
39
+ "frame_rate": 12.5
40
+ },
41
+ "audio_token_id": 24,
42
+ "begin_audio_token_id": 25,
43
+ "input_embedding_concat_type": "sum",
44
+ "acoustic_transformer_args": {
45
+ "input_dim": 3072,
46
+ "dim": 3072,
47
+ "n_layers": 3,
48
+ "head_dim": 128,
49
+ "hidden_dim": 9216,
50
+ "n_heads": 32,
51
+ "n_kv_heads": 8,
52
+ "use_biases": false,
53
+ "rope_theta": 10000.0,
54
+ "sigma": 1e-05,
55
+ "sigma_max": 1.0
56
+ },
57
+ "p_uncond": 0.0,
58
+ "text_feature_bugged": false,
59
+ "condition_dropped_token_id": 42
60
+ },
61
+ "audio_tokenizer_args": {
62
+ "channels": 1,
63
+ "sampling_rate": 24000,
64
+ "pretransform_patch_size": 240,
65
+ "patch_proj_kernel_size": 7,
66
+ "semantic_codebook_size": 8192,
67
+ "semantic_dim": 256,
68
+ "acoustic_codebook_size": 21,
69
+ "acoustic_dim": 36,
70
+ "conv_weight_norm": true,
71
+ "causal": true,
72
+ "attn_sliding_window_size": 16,
73
+ "half_attn_window_upon_downsampling": true,
74
+ "dim": 1024,
75
+ "hidden_dim": 4096,
76
+ "head_dim": 128,
77
+ "n_heads": 8,
78
+ "n_kv_heads": 8,
79
+ "qk_norm_eps": 1e-06,
80
+ "qk_norm": true,
81
+ "use_biases": false,
82
+ "norm_eps": 0.01,
83
+ "layer_scale": true,
84
+ "layer_scale_init": 0.01,
85
+ "decoder_transformer_lengths_str": "2,2,2,2",
86
+ "decoder_convs_kernels_str": "3,4,4,4",
87
+ "decoder_convs_strides_str": "1,2,2,2",
88
+ "voice": {
89
+ "casual_female": 0,
90
+ "casual_male": 1,
91
+ "cheerful_female": 2,
92
+ "neutral_female": 3,
93
+ "neutral_male": 4,
94
+ "pt_male": 5,
95
+ "pt_female": 6,
96
+ "nl_male": 7,
97
+ "nl_female": 8,
98
+ "it_male": 9,
99
+ "it_female": 10,
100
+ "fr_male": 11,
101
+ "fr_female": 12,
102
+ "es_male": 13,
103
+ "es_female": 14,
104
+ "de_male": 15,
105
+ "de_female": 16,
106
+ "ar_male": 17,
107
+ "hi_male": 18,
108
+ "hi_female": 19
109
+ }
110
+ }
111
+ },
112
+ "n_heads": 32,
113
+ "n_kv_heads": 8,
114
+ "n_layers": 26,
115
+ "norm_eps": 1e-05,
116
+ "override_parameters_str": "",
117
+ "quantization": {
118
+ "group_size": 32,
119
+ "bits": 8,
120
+ "mode": "mxfp8"
121
+ },
122
+ "quantization_config": {
123
+ "group_size": 32,
124
+ "bits": 8,
125
+ "mode": "mxfp8"
126
+ },
127
+ "recompute_attn_every": 0,
128
+ "recompute_w1_every": 0,
129
+ "recompute_w3_every": 0,
130
+ "rms_norm": "PRE",
131
+ "rope_theta": 1000000.0,
132
+ "shard_on_vocab_dim": false,
133
+ "tied_embeddings": true,
134
+ "torch_compile_swiglu_noncust_bwd": false,
135
+ "use_biases": false,
136
+ "use_cache": false,
137
+ "virtual_model_pipelining": 1,
138
+ "vocab_size": 131072,
139
+ "zero2": true
140
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9cb0074d57f355864efd6bfa991b6660b7158e4767c9f910566d399d3decb376
3
+ size 4716605637
model.safetensors.index.json ADDED
@@ -0,0 +1,599 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 4716536288,
4
+ "total_parameters": 4002337008
5
+ },
6
+ "weight_map": {
7
+ "acoustic_transformer.acoustic_codebook_output.weight": "model.safetensors",
8
+ "acoustic_transformer.input_projection.weight": "model.safetensors",
9
+ "acoustic_transformer.layers.0.attention.wk.scales": "model.safetensors",
10
+ "acoustic_transformer.layers.0.attention.wk.weight": "model.safetensors",
11
+ "acoustic_transformer.layers.0.attention.wo.scales": "model.safetensors",
12
+ "acoustic_transformer.layers.0.attention.wo.weight": "model.safetensors",
13
+ "acoustic_transformer.layers.0.attention.wq.scales": "model.safetensors",
14
+ "acoustic_transformer.layers.0.attention.wq.weight": "model.safetensors",
15
+ "acoustic_transformer.layers.0.attention.wv.scales": "model.safetensors",
16
+ "acoustic_transformer.layers.0.attention.wv.weight": "model.safetensors",
17
+ "acoustic_transformer.layers.0.attention_norm.weight": "model.safetensors",
18
+ "acoustic_transformer.layers.0.feed_forward_w1.scales": "model.safetensors",
19
+ "acoustic_transformer.layers.0.feed_forward_w1.weight": "model.safetensors",
20
+ "acoustic_transformer.layers.0.feed_forward_w2.scales": "model.safetensors",
21
+ "acoustic_transformer.layers.0.feed_forward_w2.weight": "model.safetensors",
22
+ "acoustic_transformer.layers.0.feed_forward_w3.scales": "model.safetensors",
23
+ "acoustic_transformer.layers.0.feed_forward_w3.weight": "model.safetensors",
24
+ "acoustic_transformer.layers.0.ffn_norm.weight": "model.safetensors",
25
+ "acoustic_transformer.layers.1.attention.wk.scales": "model.safetensors",
26
+ "acoustic_transformer.layers.1.attention.wk.weight": "model.safetensors",
27
+ "acoustic_transformer.layers.1.attention.wo.scales": "model.safetensors",
28
+ "acoustic_transformer.layers.1.attention.wo.weight": "model.safetensors",
29
+ "acoustic_transformer.layers.1.attention.wq.scales": "model.safetensors",
30
+ "acoustic_transformer.layers.1.attention.wq.weight": "model.safetensors",
31
+ "acoustic_transformer.layers.1.attention.wv.scales": "model.safetensors",
32
+ "acoustic_transformer.layers.1.attention.wv.weight": "model.safetensors",
33
+ "acoustic_transformer.layers.1.attention_norm.weight": "model.safetensors",
34
+ "acoustic_transformer.layers.1.feed_forward_w1.scales": "model.safetensors",
35
+ "acoustic_transformer.layers.1.feed_forward_w1.weight": "model.safetensors",
36
+ "acoustic_transformer.layers.1.feed_forward_w2.scales": "model.safetensors",
37
+ "acoustic_transformer.layers.1.feed_forward_w2.weight": "model.safetensors",
38
+ "acoustic_transformer.layers.1.feed_forward_w3.scales": "model.safetensors",
39
+ "acoustic_transformer.layers.1.feed_forward_w3.weight": "model.safetensors",
40
+ "acoustic_transformer.layers.1.ffn_norm.weight": "model.safetensors",
41
+ "acoustic_transformer.layers.2.attention.wk.scales": "model.safetensors",
42
+ "acoustic_transformer.layers.2.attention.wk.weight": "model.safetensors",
43
+ "acoustic_transformer.layers.2.attention.wo.scales": "model.safetensors",
44
+ "acoustic_transformer.layers.2.attention.wo.weight": "model.safetensors",
45
+ "acoustic_transformer.layers.2.attention.wq.scales": "model.safetensors",
46
+ "acoustic_transformer.layers.2.attention.wq.weight": "model.safetensors",
47
+ "acoustic_transformer.layers.2.attention.wv.scales": "model.safetensors",
48
+ "acoustic_transformer.layers.2.attention.wv.weight": "model.safetensors",
49
+ "acoustic_transformer.layers.2.attention_norm.weight": "model.safetensors",
50
+ "acoustic_transformer.layers.2.feed_forward_w1.scales": "model.safetensors",
51
+ "acoustic_transformer.layers.2.feed_forward_w1.weight": "model.safetensors",
52
+ "acoustic_transformer.layers.2.feed_forward_w2.scales": "model.safetensors",
53
+ "acoustic_transformer.layers.2.feed_forward_w2.weight": "model.safetensors",
54
+ "acoustic_transformer.layers.2.feed_forward_w3.scales": "model.safetensors",
55
+ "acoustic_transformer.layers.2.feed_forward_w3.weight": "model.safetensors",
56
+ "acoustic_transformer.layers.2.ffn_norm.weight": "model.safetensors",
57
+ "acoustic_transformer.llm_projection.scales": "model.safetensors",
58
+ "acoustic_transformer.llm_projection.weight": "model.safetensors",
59
+ "acoustic_transformer.norm.weight": "model.safetensors",
60
+ "acoustic_transformer.semantic_codebook_output.weight": "model.safetensors",
61
+ "acoustic_transformer.time_projection.scales": "model.safetensors",
62
+ "acoustic_transformer.time_projection.weight": "model.safetensors",
63
+ "audio_tokenizer_decoder.decoder_blocks.0.conv.original0": "model.safetensors",
64
+ "audio_tokenizer_decoder.decoder_blocks.0.conv.original1": "model.safetensors",
65
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.0.attention.k_norm.weight": "model.safetensors",
66
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.0.attention.q_norm.weight": "model.safetensors",
67
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.0.attention.wk.weight": "model.safetensors",
68
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.0.attention.wo.weight": "model.safetensors",
69
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.0.attention.wq.weight": "model.safetensors",
70
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.0.attention.wv.weight": "model.safetensors",
71
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.0.attention_norm.weight": "model.safetensors",
72
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.0.attention_scale": "model.safetensors",
73
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.0.feed_forward_w1.weight": "model.safetensors",
74
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.0.feed_forward_w2.weight": "model.safetensors",
75
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.0.feed_forward_w3.weight": "model.safetensors",
76
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.0.ffn_norm.weight": "model.safetensors",
77
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.0.ffn_scale": "model.safetensors",
78
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.1.attention.k_norm.weight": "model.safetensors",
79
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.1.attention.q_norm.weight": "model.safetensors",
80
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.1.attention.wk.weight": "model.safetensors",
81
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.1.attention.wo.weight": "model.safetensors",
82
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.1.attention.wq.weight": "model.safetensors",
83
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.1.attention.wv.weight": "model.safetensors",
84
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.1.attention_norm.weight": "model.safetensors",
85
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.1.attention_scale": "model.safetensors",
86
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.1.feed_forward_w1.weight": "model.safetensors",
87
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.1.feed_forward_w2.weight": "model.safetensors",
88
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.1.feed_forward_w3.weight": "model.safetensors",
89
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.1.ffn_norm.weight": "model.safetensors",
90
+ "audio_tokenizer_decoder.decoder_blocks.1.layers.1.ffn_scale": "model.safetensors",
91
+ "audio_tokenizer_decoder.decoder_blocks.2.conv.original0": "model.safetensors",
92
+ "audio_tokenizer_decoder.decoder_blocks.2.conv.original1": "model.safetensors",
93
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.0.attention.k_norm.weight": "model.safetensors",
94
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.0.attention.q_norm.weight": "model.safetensors",
95
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.0.attention.wk.weight": "model.safetensors",
96
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.0.attention.wo.weight": "model.safetensors",
97
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.0.attention.wq.weight": "model.safetensors",
98
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.0.attention.wv.weight": "model.safetensors",
99
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.0.attention_norm.weight": "model.safetensors",
100
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.0.attention_scale": "model.safetensors",
101
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.0.feed_forward_w1.weight": "model.safetensors",
102
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.0.feed_forward_w2.weight": "model.safetensors",
103
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.0.feed_forward_w3.weight": "model.safetensors",
104
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.0.ffn_norm.weight": "model.safetensors",
105
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.0.ffn_scale": "model.safetensors",
106
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.1.attention.k_norm.weight": "model.safetensors",
107
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.1.attention.q_norm.weight": "model.safetensors",
108
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.1.attention.wk.weight": "model.safetensors",
109
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.1.attention.wo.weight": "model.safetensors",
110
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.1.attention.wq.weight": "model.safetensors",
111
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.1.attention.wv.weight": "model.safetensors",
112
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.1.attention_norm.weight": "model.safetensors",
113
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.1.attention_scale": "model.safetensors",
114
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.1.feed_forward_w1.weight": "model.safetensors",
115
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.1.feed_forward_w2.weight": "model.safetensors",
116
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.1.feed_forward_w3.weight": "model.safetensors",
117
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.1.ffn_norm.weight": "model.safetensors",
118
+ "audio_tokenizer_decoder.decoder_blocks.3.layers.1.ffn_scale": "model.safetensors",
119
+ "audio_tokenizer_decoder.decoder_blocks.4.conv.original0": "model.safetensors",
120
+ "audio_tokenizer_decoder.decoder_blocks.4.conv.original1": "model.safetensors",
121
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.0.attention.k_norm.weight": "model.safetensors",
122
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.0.attention.q_norm.weight": "model.safetensors",
123
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.0.attention.wk.weight": "model.safetensors",
124
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.0.attention.wo.weight": "model.safetensors",
125
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.0.attention.wq.weight": "model.safetensors",
126
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.0.attention.wv.weight": "model.safetensors",
127
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.0.attention_norm.weight": "model.safetensors",
128
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.0.attention_scale": "model.safetensors",
129
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.0.feed_forward_w1.weight": "model.safetensors",
130
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.0.feed_forward_w2.weight": "model.safetensors",
131
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.0.feed_forward_w3.weight": "model.safetensors",
132
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.0.ffn_norm.weight": "model.safetensors",
133
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.0.ffn_scale": "model.safetensors",
134
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.1.attention.k_norm.weight": "model.safetensors",
135
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.1.attention.q_norm.weight": "model.safetensors",
136
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.1.attention.wk.weight": "model.safetensors",
137
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.1.attention.wo.weight": "model.safetensors",
138
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.1.attention.wq.weight": "model.safetensors",
139
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.1.attention.wv.weight": "model.safetensors",
140
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.1.attention_norm.weight": "model.safetensors",
141
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.1.attention_scale": "model.safetensors",
142
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.1.feed_forward_w1.weight": "model.safetensors",
143
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.1.feed_forward_w2.weight": "model.safetensors",
144
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.1.feed_forward_w3.weight": "model.safetensors",
145
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.1.ffn_norm.weight": "model.safetensors",
146
+ "audio_tokenizer_decoder.decoder_blocks.5.layers.1.ffn_scale": "model.safetensors",
147
+ "audio_tokenizer_decoder.decoder_blocks.6.conv.original0": "model.safetensors",
148
+ "audio_tokenizer_decoder.decoder_blocks.6.conv.original1": "model.safetensors",
149
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.0.attention.k_norm.weight": "model.safetensors",
150
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.0.attention.q_norm.weight": "model.safetensors",
151
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.0.attention.wk.weight": "model.safetensors",
152
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.0.attention.wo.weight": "model.safetensors",
153
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.0.attention.wq.weight": "model.safetensors",
154
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.0.attention.wv.weight": "model.safetensors",
155
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.0.attention_norm.weight": "model.safetensors",
156
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.0.attention_scale": "model.safetensors",
157
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.0.feed_forward_w1.weight": "model.safetensors",
158
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.0.feed_forward_w2.weight": "model.safetensors",
159
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.0.feed_forward_w3.weight": "model.safetensors",
160
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.0.ffn_norm.weight": "model.safetensors",
161
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.0.ffn_scale": "model.safetensors",
162
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.1.attention.k_norm.weight": "model.safetensors",
163
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.1.attention.q_norm.weight": "model.safetensors",
164
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.1.attention.wk.weight": "model.safetensors",
165
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.1.attention.wo.weight": "model.safetensors",
166
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.1.attention.wq.weight": "model.safetensors",
167
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.1.attention.wv.weight": "model.safetensors",
168
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.1.attention_norm.weight": "model.safetensors",
169
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.1.attention_scale": "model.safetensors",
170
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.1.feed_forward_w1.weight": "model.safetensors",
171
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.1.feed_forward_w2.weight": "model.safetensors",
172
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.1.feed_forward_w3.weight": "model.safetensors",
173
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.1.ffn_norm.weight": "model.safetensors",
174
+ "audio_tokenizer_decoder.decoder_blocks.7.layers.1.ffn_scale": "model.safetensors",
175
+ "audio_tokenizer_decoder.output_proj.conv.original0": "model.safetensors",
176
+ "audio_tokenizer_decoder.output_proj.conv.original1": "model.safetensors",
177
+ "audio_tokenizer_decoder.quantizer_semantic_codebook.cluster_usage": "model.safetensors",
178
+ "audio_tokenizer_decoder.quantizer_semantic_codebook.embedding_sum": "model.safetensors",
179
+ "llm.audio_codebook_embeddings.weight": "model.safetensors",
180
+ "llm.layers.0.attention.wk.scales": "model.safetensors",
181
+ "llm.layers.0.attention.wk.weight": "model.safetensors",
182
+ "llm.layers.0.attention.wo.scales": "model.safetensors",
183
+ "llm.layers.0.attention.wo.weight": "model.safetensors",
184
+ "llm.layers.0.attention.wq.scales": "model.safetensors",
185
+ "llm.layers.0.attention.wq.weight": "model.safetensors",
186
+ "llm.layers.0.attention.wv.scales": "model.safetensors",
187
+ "llm.layers.0.attention.wv.weight": "model.safetensors",
188
+ "llm.layers.0.attention_norm.weight": "model.safetensors",
189
+ "llm.layers.0.feed_forward_w1.scales": "model.safetensors",
190
+ "llm.layers.0.feed_forward_w1.weight": "model.safetensors",
191
+ "llm.layers.0.feed_forward_w2.scales": "model.safetensors",
192
+ "llm.layers.0.feed_forward_w2.weight": "model.safetensors",
193
+ "llm.layers.0.feed_forward_w3.scales": "model.safetensors",
194
+ "llm.layers.0.feed_forward_w3.weight": "model.safetensors",
195
+ "llm.layers.0.ffn_norm.weight": "model.safetensors",
196
+ "llm.layers.1.attention.wk.scales": "model.safetensors",
197
+ "llm.layers.1.attention.wk.weight": "model.safetensors",
198
+ "llm.layers.1.attention.wo.scales": "model.safetensors",
199
+ "llm.layers.1.attention.wo.weight": "model.safetensors",
200
+ "llm.layers.1.attention.wq.scales": "model.safetensors",
201
+ "llm.layers.1.attention.wq.weight": "model.safetensors",
202
+ "llm.layers.1.attention.wv.scales": "model.safetensors",
203
+ "llm.layers.1.attention.wv.weight": "model.safetensors",
204
+ "llm.layers.1.attention_norm.weight": "model.safetensors",
205
+ "llm.layers.1.feed_forward_w1.scales": "model.safetensors",
206
+ "llm.layers.1.feed_forward_w1.weight": "model.safetensors",
207
+ "llm.layers.1.feed_forward_w2.scales": "model.safetensors",
208
+ "llm.layers.1.feed_forward_w2.weight": "model.safetensors",
209
+ "llm.layers.1.feed_forward_w3.scales": "model.safetensors",
210
+ "llm.layers.1.feed_forward_w3.weight": "model.safetensors",
211
+ "llm.layers.1.ffn_norm.weight": "model.safetensors",
212
+ "llm.layers.10.attention.wk.scales": "model.safetensors",
213
+ "llm.layers.10.attention.wk.weight": "model.safetensors",
214
+ "llm.layers.10.attention.wo.scales": "model.safetensors",
215
+ "llm.layers.10.attention.wo.weight": "model.safetensors",
216
+ "llm.layers.10.attention.wq.scales": "model.safetensors",
217
+ "llm.layers.10.attention.wq.weight": "model.safetensors",
218
+ "llm.layers.10.attention.wv.scales": "model.safetensors",
219
+ "llm.layers.10.attention.wv.weight": "model.safetensors",
220
+ "llm.layers.10.attention_norm.weight": "model.safetensors",
221
+ "llm.layers.10.feed_forward_w1.scales": "model.safetensors",
222
+ "llm.layers.10.feed_forward_w1.weight": "model.safetensors",
223
+ "llm.layers.10.feed_forward_w2.scales": "model.safetensors",
224
+ "llm.layers.10.feed_forward_w2.weight": "model.safetensors",
225
+ "llm.layers.10.feed_forward_w3.scales": "model.safetensors",
226
+ "llm.layers.10.feed_forward_w3.weight": "model.safetensors",
227
+ "llm.layers.10.ffn_norm.weight": "model.safetensors",
228
+ "llm.layers.11.attention.wk.scales": "model.safetensors",
229
+ "llm.layers.11.attention.wk.weight": "model.safetensors",
230
+ "llm.layers.11.attention.wo.scales": "model.safetensors",
231
+ "llm.layers.11.attention.wo.weight": "model.safetensors",
232
+ "llm.layers.11.attention.wq.scales": "model.safetensors",
233
+ "llm.layers.11.attention.wq.weight": "model.safetensors",
234
+ "llm.layers.11.attention.wv.scales": "model.safetensors",
235
+ "llm.layers.11.attention.wv.weight": "model.safetensors",
236
+ "llm.layers.11.attention_norm.weight": "model.safetensors",
237
+ "llm.layers.11.feed_forward_w1.scales": "model.safetensors",
238
+ "llm.layers.11.feed_forward_w1.weight": "model.safetensors",
239
+ "llm.layers.11.feed_forward_w2.scales": "model.safetensors",
240
+ "llm.layers.11.feed_forward_w2.weight": "model.safetensors",
241
+ "llm.layers.11.feed_forward_w3.scales": "model.safetensors",
242
+ "llm.layers.11.feed_forward_w3.weight": "model.safetensors",
243
+ "llm.layers.11.ffn_norm.weight": "model.safetensors",
244
+ "llm.layers.12.attention.wk.scales": "model.safetensors",
245
+ "llm.layers.12.attention.wk.weight": "model.safetensors",
246
+ "llm.layers.12.attention.wo.scales": "model.safetensors",
247
+ "llm.layers.12.attention.wo.weight": "model.safetensors",
248
+ "llm.layers.12.attention.wq.scales": "model.safetensors",
249
+ "llm.layers.12.attention.wq.weight": "model.safetensors",
250
+ "llm.layers.12.attention.wv.scales": "model.safetensors",
251
+ "llm.layers.12.attention.wv.weight": "model.safetensors",
252
+ "llm.layers.12.attention_norm.weight": "model.safetensors",
253
+ "llm.layers.12.feed_forward_w1.scales": "model.safetensors",
254
+ "llm.layers.12.feed_forward_w1.weight": "model.safetensors",
255
+ "llm.layers.12.feed_forward_w2.scales": "model.safetensors",
256
+ "llm.layers.12.feed_forward_w2.weight": "model.safetensors",
257
+ "llm.layers.12.feed_forward_w3.scales": "model.safetensors",
258
+ "llm.layers.12.feed_forward_w3.weight": "model.safetensors",
259
+ "llm.layers.12.ffn_norm.weight": "model.safetensors",
260
+ "llm.layers.13.attention.wk.scales": "model.safetensors",
261
+ "llm.layers.13.attention.wk.weight": "model.safetensors",
262
+ "llm.layers.13.attention.wo.scales": "model.safetensors",
263
+ "llm.layers.13.attention.wo.weight": "model.safetensors",
264
+ "llm.layers.13.attention.wq.scales": "model.safetensors",
265
+ "llm.layers.13.attention.wq.weight": "model.safetensors",
266
+ "llm.layers.13.attention.wv.scales": "model.safetensors",
267
+ "llm.layers.13.attention.wv.weight": "model.safetensors",
268
+ "llm.layers.13.attention_norm.weight": "model.safetensors",
269
+ "llm.layers.13.feed_forward_w1.scales": "model.safetensors",
270
+ "llm.layers.13.feed_forward_w1.weight": "model.safetensors",
271
+ "llm.layers.13.feed_forward_w2.scales": "model.safetensors",
272
+ "llm.layers.13.feed_forward_w2.weight": "model.safetensors",
273
+ "llm.layers.13.feed_forward_w3.scales": "model.safetensors",
274
+ "llm.layers.13.feed_forward_w3.weight": "model.safetensors",
275
+ "llm.layers.13.ffn_norm.weight": "model.safetensors",
276
+ "llm.layers.14.attention.wk.scales": "model.safetensors",
277
+ "llm.layers.14.attention.wk.weight": "model.safetensors",
278
+ "llm.layers.14.attention.wo.scales": "model.safetensors",
279
+ "llm.layers.14.attention.wo.weight": "model.safetensors",
280
+ "llm.layers.14.attention.wq.scales": "model.safetensors",
281
+ "llm.layers.14.attention.wq.weight": "model.safetensors",
282
+ "llm.layers.14.attention.wv.scales": "model.safetensors",
283
+ "llm.layers.14.attention.wv.weight": "model.safetensors",
284
+ "llm.layers.14.attention_norm.weight": "model.safetensors",
285
+ "llm.layers.14.feed_forward_w1.scales": "model.safetensors",
286
+ "llm.layers.14.feed_forward_w1.weight": "model.safetensors",
287
+ "llm.layers.14.feed_forward_w2.scales": "model.safetensors",
288
+ "llm.layers.14.feed_forward_w2.weight": "model.safetensors",
289
+ "llm.layers.14.feed_forward_w3.scales": "model.safetensors",
290
+ "llm.layers.14.feed_forward_w3.weight": "model.safetensors",
291
+ "llm.layers.14.ffn_norm.weight": "model.safetensors",
292
+ "llm.layers.15.attention.wk.scales": "model.safetensors",
293
+ "llm.layers.15.attention.wk.weight": "model.safetensors",
294
+ "llm.layers.15.attention.wo.scales": "model.safetensors",
295
+ "llm.layers.15.attention.wo.weight": "model.safetensors",
296
+ "llm.layers.15.attention.wq.scales": "model.safetensors",
297
+ "llm.layers.15.attention.wq.weight": "model.safetensors",
298
+ "llm.layers.15.attention.wv.scales": "model.safetensors",
299
+ "llm.layers.15.attention.wv.weight": "model.safetensors",
300
+ "llm.layers.15.attention_norm.weight": "model.safetensors",
301
+ "llm.layers.15.feed_forward_w1.scales": "model.safetensors",
302
+ "llm.layers.15.feed_forward_w1.weight": "model.safetensors",
303
+ "llm.layers.15.feed_forward_w2.scales": "model.safetensors",
304
+ "llm.layers.15.feed_forward_w2.weight": "model.safetensors",
305
+ "llm.layers.15.feed_forward_w3.scales": "model.safetensors",
306
+ "llm.layers.15.feed_forward_w3.weight": "model.safetensors",
307
+ "llm.layers.15.ffn_norm.weight": "model.safetensors",
308
+ "llm.layers.16.attention.wk.scales": "model.safetensors",
309
+ "llm.layers.16.attention.wk.weight": "model.safetensors",
310
+ "llm.layers.16.attention.wo.scales": "model.safetensors",
311
+ "llm.layers.16.attention.wo.weight": "model.safetensors",
312
+ "llm.layers.16.attention.wq.scales": "model.safetensors",
313
+ "llm.layers.16.attention.wq.weight": "model.safetensors",
314
+ "llm.layers.16.attention.wv.scales": "model.safetensors",
315
+ "llm.layers.16.attention.wv.weight": "model.safetensors",
316
+ "llm.layers.16.attention_norm.weight": "model.safetensors",
317
+ "llm.layers.16.feed_forward_w1.scales": "model.safetensors",
318
+ "llm.layers.16.feed_forward_w1.weight": "model.safetensors",
319
+ "llm.layers.16.feed_forward_w2.scales": "model.safetensors",
320
+ "llm.layers.16.feed_forward_w2.weight": "model.safetensors",
321
+ "llm.layers.16.feed_forward_w3.scales": "model.safetensors",
322
+ "llm.layers.16.feed_forward_w3.weight": "model.safetensors",
323
+ "llm.layers.16.ffn_norm.weight": "model.safetensors",
324
+ "llm.layers.17.attention.wk.scales": "model.safetensors",
325
+ "llm.layers.17.attention.wk.weight": "model.safetensors",
326
+ "llm.layers.17.attention.wo.scales": "model.safetensors",
327
+ "llm.layers.17.attention.wo.weight": "model.safetensors",
328
+ "llm.layers.17.attention.wq.scales": "model.safetensors",
329
+ "llm.layers.17.attention.wq.weight": "model.safetensors",
330
+ "llm.layers.17.attention.wv.scales": "model.safetensors",
331
+ "llm.layers.17.attention.wv.weight": "model.safetensors",
332
+ "llm.layers.17.attention_norm.weight": "model.safetensors",
333
+ "llm.layers.17.feed_forward_w1.scales": "model.safetensors",
334
+ "llm.layers.17.feed_forward_w1.weight": "model.safetensors",
335
+ "llm.layers.17.feed_forward_w2.scales": "model.safetensors",
336
+ "llm.layers.17.feed_forward_w2.weight": "model.safetensors",
337
+ "llm.layers.17.feed_forward_w3.scales": "model.safetensors",
338
+ "llm.layers.17.feed_forward_w3.weight": "model.safetensors",
339
+ "llm.layers.17.ffn_norm.weight": "model.safetensors",
340
+ "llm.layers.18.attention.wk.scales": "model.safetensors",
341
+ "llm.layers.18.attention.wk.weight": "model.safetensors",
342
+ "llm.layers.18.attention.wo.scales": "model.safetensors",
343
+ "llm.layers.18.attention.wo.weight": "model.safetensors",
344
+ "llm.layers.18.attention.wq.scales": "model.safetensors",
345
+ "llm.layers.18.attention.wq.weight": "model.safetensors",
346
+ "llm.layers.18.attention.wv.scales": "model.safetensors",
347
+ "llm.layers.18.attention.wv.weight": "model.safetensors",
348
+ "llm.layers.18.attention_norm.weight": "model.safetensors",
349
+ "llm.layers.18.feed_forward_w1.scales": "model.safetensors",
350
+ "llm.layers.18.feed_forward_w1.weight": "model.safetensors",
351
+ "llm.layers.18.feed_forward_w2.scales": "model.safetensors",
352
+ "llm.layers.18.feed_forward_w2.weight": "model.safetensors",
353
+ "llm.layers.18.feed_forward_w3.scales": "model.safetensors",
354
+ "llm.layers.18.feed_forward_w3.weight": "model.safetensors",
355
+ "llm.layers.18.ffn_norm.weight": "model.safetensors",
356
+ "llm.layers.19.attention.wk.scales": "model.safetensors",
357
+ "llm.layers.19.attention.wk.weight": "model.safetensors",
358
+ "llm.layers.19.attention.wo.scales": "model.safetensors",
359
+ "llm.layers.19.attention.wo.weight": "model.safetensors",
360
+ "llm.layers.19.attention.wq.scales": "model.safetensors",
361
+ "llm.layers.19.attention.wq.weight": "model.safetensors",
362
+ "llm.layers.19.attention.wv.scales": "model.safetensors",
363
+ "llm.layers.19.attention.wv.weight": "model.safetensors",
364
+ "llm.layers.19.attention_norm.weight": "model.safetensors",
365
+ "llm.layers.19.feed_forward_w1.scales": "model.safetensors",
366
+ "llm.layers.19.feed_forward_w1.weight": "model.safetensors",
367
+ "llm.layers.19.feed_forward_w2.scales": "model.safetensors",
368
+ "llm.layers.19.feed_forward_w2.weight": "model.safetensors",
369
+ "llm.layers.19.feed_forward_w3.scales": "model.safetensors",
370
+ "llm.layers.19.feed_forward_w3.weight": "model.safetensors",
371
+ "llm.layers.19.ffn_norm.weight": "model.safetensors",
372
+ "llm.layers.2.attention.wk.scales": "model.safetensors",
373
+ "llm.layers.2.attention.wk.weight": "model.safetensors",
374
+ "llm.layers.2.attention.wo.scales": "model.safetensors",
375
+ "llm.layers.2.attention.wo.weight": "model.safetensors",
376
+ "llm.layers.2.attention.wq.scales": "model.safetensors",
377
+ "llm.layers.2.attention.wq.weight": "model.safetensors",
378
+ "llm.layers.2.attention.wv.scales": "model.safetensors",
379
+ "llm.layers.2.attention.wv.weight": "model.safetensors",
380
+ "llm.layers.2.attention_norm.weight": "model.safetensors",
381
+ "llm.layers.2.feed_forward_w1.scales": "model.safetensors",
382
+ "llm.layers.2.feed_forward_w1.weight": "model.safetensors",
383
+ "llm.layers.2.feed_forward_w2.scales": "model.safetensors",
384
+ "llm.layers.2.feed_forward_w2.weight": "model.safetensors",
385
+ "llm.layers.2.feed_forward_w3.scales": "model.safetensors",
386
+ "llm.layers.2.feed_forward_w3.weight": "model.safetensors",
387
+ "llm.layers.2.ffn_norm.weight": "model.safetensors",
388
+ "llm.layers.20.attention.wk.scales": "model.safetensors",
389
+ "llm.layers.20.attention.wk.weight": "model.safetensors",
390
+ "llm.layers.20.attention.wo.scales": "model.safetensors",
391
+ "llm.layers.20.attention.wo.weight": "model.safetensors",
392
+ "llm.layers.20.attention.wq.scales": "model.safetensors",
393
+ "llm.layers.20.attention.wq.weight": "model.safetensors",
394
+ "llm.layers.20.attention.wv.scales": "model.safetensors",
395
+ "llm.layers.20.attention.wv.weight": "model.safetensors",
396
+ "llm.layers.20.attention_norm.weight": "model.safetensors",
397
+ "llm.layers.20.feed_forward_w1.scales": "model.safetensors",
398
+ "llm.layers.20.feed_forward_w1.weight": "model.safetensors",
399
+ "llm.layers.20.feed_forward_w2.scales": "model.safetensors",
400
+ "llm.layers.20.feed_forward_w2.weight": "model.safetensors",
401
+ "llm.layers.20.feed_forward_w3.scales": "model.safetensors",
402
+ "llm.layers.20.feed_forward_w3.weight": "model.safetensors",
403
+ "llm.layers.20.ffn_norm.weight": "model.safetensors",
404
+ "llm.layers.21.attention.wk.scales": "model.safetensors",
405
+ "llm.layers.21.attention.wk.weight": "model.safetensors",
406
+ "llm.layers.21.attention.wo.scales": "model.safetensors",
407
+ "llm.layers.21.attention.wo.weight": "model.safetensors",
408
+ "llm.layers.21.attention.wq.scales": "model.safetensors",
409
+ "llm.layers.21.attention.wq.weight": "model.safetensors",
410
+ "llm.layers.21.attention.wv.scales": "model.safetensors",
411
+ "llm.layers.21.attention.wv.weight": "model.safetensors",
412
+ "llm.layers.21.attention_norm.weight": "model.safetensors",
413
+ "llm.layers.21.feed_forward_w1.scales": "model.safetensors",
414
+ "llm.layers.21.feed_forward_w1.weight": "model.safetensors",
415
+ "llm.layers.21.feed_forward_w2.scales": "model.safetensors",
416
+ "llm.layers.21.feed_forward_w2.weight": "model.safetensors",
417
+ "llm.layers.21.feed_forward_w3.scales": "model.safetensors",
418
+ "llm.layers.21.feed_forward_w3.weight": "model.safetensors",
419
+ "llm.layers.21.ffn_norm.weight": "model.safetensors",
420
+ "llm.layers.22.attention.wk.scales": "model.safetensors",
421
+ "llm.layers.22.attention.wk.weight": "model.safetensors",
422
+ "llm.layers.22.attention.wo.scales": "model.safetensors",
423
+ "llm.layers.22.attention.wo.weight": "model.safetensors",
424
+ "llm.layers.22.attention.wq.scales": "model.safetensors",
425
+ "llm.layers.22.attention.wq.weight": "model.safetensors",
426
+ "llm.layers.22.attention.wv.scales": "model.safetensors",
427
+ "llm.layers.22.attention.wv.weight": "model.safetensors",
428
+ "llm.layers.22.attention_norm.weight": "model.safetensors",
429
+ "llm.layers.22.feed_forward_w1.scales": "model.safetensors",
430
+ "llm.layers.22.feed_forward_w1.weight": "model.safetensors",
431
+ "llm.layers.22.feed_forward_w2.scales": "model.safetensors",
432
+ "llm.layers.22.feed_forward_w2.weight": "model.safetensors",
433
+ "llm.layers.22.feed_forward_w3.scales": "model.safetensors",
434
+ "llm.layers.22.feed_forward_w3.weight": "model.safetensors",
435
+ "llm.layers.22.ffn_norm.weight": "model.safetensors",
436
+ "llm.layers.23.attention.wk.scales": "model.safetensors",
437
+ "llm.layers.23.attention.wk.weight": "model.safetensors",
438
+ "llm.layers.23.attention.wo.scales": "model.safetensors",
439
+ "llm.layers.23.attention.wo.weight": "model.safetensors",
440
+ "llm.layers.23.attention.wq.scales": "model.safetensors",
441
+ "llm.layers.23.attention.wq.weight": "model.safetensors",
442
+ "llm.layers.23.attention.wv.scales": "model.safetensors",
443
+ "llm.layers.23.attention.wv.weight": "model.safetensors",
444
+ "llm.layers.23.attention_norm.weight": "model.safetensors",
445
+ "llm.layers.23.feed_forward_w1.scales": "model.safetensors",
446
+ "llm.layers.23.feed_forward_w1.weight": "model.safetensors",
447
+ "llm.layers.23.feed_forward_w2.scales": "model.safetensors",
448
+ "llm.layers.23.feed_forward_w2.weight": "model.safetensors",
449
+ "llm.layers.23.feed_forward_w3.scales": "model.safetensors",
450
+ "llm.layers.23.feed_forward_w3.weight": "model.safetensors",
451
+ "llm.layers.23.ffn_norm.weight": "model.safetensors",
452
+ "llm.layers.24.attention.wk.scales": "model.safetensors",
453
+ "llm.layers.24.attention.wk.weight": "model.safetensors",
454
+ "llm.layers.24.attention.wo.scales": "model.safetensors",
455
+ "llm.layers.24.attention.wo.weight": "model.safetensors",
456
+ "llm.layers.24.attention.wq.scales": "model.safetensors",
457
+ "llm.layers.24.attention.wq.weight": "model.safetensors",
458
+ "llm.layers.24.attention.wv.scales": "model.safetensors",
459
+ "llm.layers.24.attention.wv.weight": "model.safetensors",
460
+ "llm.layers.24.attention_norm.weight": "model.safetensors",
461
+ "llm.layers.24.feed_forward_w1.scales": "model.safetensors",
462
+ "llm.layers.24.feed_forward_w1.weight": "model.safetensors",
463
+ "llm.layers.24.feed_forward_w2.scales": "model.safetensors",
464
+ "llm.layers.24.feed_forward_w2.weight": "model.safetensors",
465
+ "llm.layers.24.feed_forward_w3.scales": "model.safetensors",
466
+ "llm.layers.24.feed_forward_w3.weight": "model.safetensors",
467
+ "llm.layers.24.ffn_norm.weight": "model.safetensors",
468
+ "llm.layers.25.attention.wk.scales": "model.safetensors",
469
+ "llm.layers.25.attention.wk.weight": "model.safetensors",
470
+ "llm.layers.25.attention.wo.scales": "model.safetensors",
471
+ "llm.layers.25.attention.wo.weight": "model.safetensors",
472
+ "llm.layers.25.attention.wq.scales": "model.safetensors",
473
+ "llm.layers.25.attention.wq.weight": "model.safetensors",
474
+ "llm.layers.25.attention.wv.scales": "model.safetensors",
475
+ "llm.layers.25.attention.wv.weight": "model.safetensors",
476
+ "llm.layers.25.attention_norm.weight": "model.safetensors",
477
+ "llm.layers.25.feed_forward_w1.scales": "model.safetensors",
478
+ "llm.layers.25.feed_forward_w1.weight": "model.safetensors",
479
+ "llm.layers.25.feed_forward_w2.scales": "model.safetensors",
480
+ "llm.layers.25.feed_forward_w2.weight": "model.safetensors",
481
+ "llm.layers.25.feed_forward_w3.scales": "model.safetensors",
482
+ "llm.layers.25.feed_forward_w3.weight": "model.safetensors",
483
+ "llm.layers.25.ffn_norm.weight": "model.safetensors",
484
+ "llm.layers.3.attention.wk.scales": "model.safetensors",
485
+ "llm.layers.3.attention.wk.weight": "model.safetensors",
486
+ "llm.layers.3.attention.wo.scales": "model.safetensors",
487
+ "llm.layers.3.attention.wo.weight": "model.safetensors",
488
+ "llm.layers.3.attention.wq.scales": "model.safetensors",
489
+ "llm.layers.3.attention.wq.weight": "model.safetensors",
490
+ "llm.layers.3.attention.wv.scales": "model.safetensors",
491
+ "llm.layers.3.attention.wv.weight": "model.safetensors",
492
+ "llm.layers.3.attention_norm.weight": "model.safetensors",
493
+ "llm.layers.3.feed_forward_w1.scales": "model.safetensors",
494
+ "llm.layers.3.feed_forward_w1.weight": "model.safetensors",
495
+ "llm.layers.3.feed_forward_w2.scales": "model.safetensors",
496
+ "llm.layers.3.feed_forward_w2.weight": "model.safetensors",
497
+ "llm.layers.3.feed_forward_w3.scales": "model.safetensors",
498
+ "llm.layers.3.feed_forward_w3.weight": "model.safetensors",
499
+ "llm.layers.3.ffn_norm.weight": "model.safetensors",
500
+ "llm.layers.4.attention.wk.scales": "model.safetensors",
501
+ "llm.layers.4.attention.wk.weight": "model.safetensors",
502
+ "llm.layers.4.attention.wo.scales": "model.safetensors",
503
+ "llm.layers.4.attention.wo.weight": "model.safetensors",
504
+ "llm.layers.4.attention.wq.scales": "model.safetensors",
505
+ "llm.layers.4.attention.wq.weight": "model.safetensors",
506
+ "llm.layers.4.attention.wv.scales": "model.safetensors",
507
+ "llm.layers.4.attention.wv.weight": "model.safetensors",
508
+ "llm.layers.4.attention_norm.weight": "model.safetensors",
509
+ "llm.layers.4.feed_forward_w1.scales": "model.safetensors",
510
+ "llm.layers.4.feed_forward_w1.weight": "model.safetensors",
511
+ "llm.layers.4.feed_forward_w2.scales": "model.safetensors",
512
+ "llm.layers.4.feed_forward_w2.weight": "model.safetensors",
513
+ "llm.layers.4.feed_forward_w3.scales": "model.safetensors",
514
+ "llm.layers.4.feed_forward_w3.weight": "model.safetensors",
515
+ "llm.layers.4.ffn_norm.weight": "model.safetensors",
516
+ "llm.layers.5.attention.wk.scales": "model.safetensors",
517
+ "llm.layers.5.attention.wk.weight": "model.safetensors",
518
+ "llm.layers.5.attention.wo.scales": "model.safetensors",
519
+ "llm.layers.5.attention.wo.weight": "model.safetensors",
520
+ "llm.layers.5.attention.wq.scales": "model.safetensors",
521
+ "llm.layers.5.attention.wq.weight": "model.safetensors",
522
+ "llm.layers.5.attention.wv.scales": "model.safetensors",
523
+ "llm.layers.5.attention.wv.weight": "model.safetensors",
524
+ "llm.layers.5.attention_norm.weight": "model.safetensors",
525
+ "llm.layers.5.feed_forward_w1.scales": "model.safetensors",
526
+ "llm.layers.5.feed_forward_w1.weight": "model.safetensors",
527
+ "llm.layers.5.feed_forward_w2.scales": "model.safetensors",
528
+ "llm.layers.5.feed_forward_w2.weight": "model.safetensors",
529
+ "llm.layers.5.feed_forward_w3.scales": "model.safetensors",
530
+ "llm.layers.5.feed_forward_w3.weight": "model.safetensors",
531
+ "llm.layers.5.ffn_norm.weight": "model.safetensors",
532
+ "llm.layers.6.attention.wk.scales": "model.safetensors",
533
+ "llm.layers.6.attention.wk.weight": "model.safetensors",
534
+ "llm.layers.6.attention.wo.scales": "model.safetensors",
535
+ "llm.layers.6.attention.wo.weight": "model.safetensors",
536
+ "llm.layers.6.attention.wq.scales": "model.safetensors",
537
+ "llm.layers.6.attention.wq.weight": "model.safetensors",
538
+ "llm.layers.6.attention.wv.scales": "model.safetensors",
539
+ "llm.layers.6.attention.wv.weight": "model.safetensors",
540
+ "llm.layers.6.attention_norm.weight": "model.safetensors",
541
+ "llm.layers.6.feed_forward_w1.scales": "model.safetensors",
542
+ "llm.layers.6.feed_forward_w1.weight": "model.safetensors",
543
+ "llm.layers.6.feed_forward_w2.scales": "model.safetensors",
544
+ "llm.layers.6.feed_forward_w2.weight": "model.safetensors",
545
+ "llm.layers.6.feed_forward_w3.scales": "model.safetensors",
546
+ "llm.layers.6.feed_forward_w3.weight": "model.safetensors",
547
+ "llm.layers.6.ffn_norm.weight": "model.safetensors",
548
+ "llm.layers.7.attention.wk.scales": "model.safetensors",
549
+ "llm.layers.7.attention.wk.weight": "model.safetensors",
550
+ "llm.layers.7.attention.wo.scales": "model.safetensors",
551
+ "llm.layers.7.attention.wo.weight": "model.safetensors",
552
+ "llm.layers.7.attention.wq.scales": "model.safetensors",
553
+ "llm.layers.7.attention.wq.weight": "model.safetensors",
554
+ "llm.layers.7.attention.wv.scales": "model.safetensors",
555
+ "llm.layers.7.attention.wv.weight": "model.safetensors",
556
+ "llm.layers.7.attention_norm.weight": "model.safetensors",
557
+ "llm.layers.7.feed_forward_w1.scales": "model.safetensors",
558
+ "llm.layers.7.feed_forward_w1.weight": "model.safetensors",
559
+ "llm.layers.7.feed_forward_w2.scales": "model.safetensors",
560
+ "llm.layers.7.feed_forward_w2.weight": "model.safetensors",
561
+ "llm.layers.7.feed_forward_w3.scales": "model.safetensors",
562
+ "llm.layers.7.feed_forward_w3.weight": "model.safetensors",
563
+ "llm.layers.7.ffn_norm.weight": "model.safetensors",
564
+ "llm.layers.8.attention.wk.scales": "model.safetensors",
565
+ "llm.layers.8.attention.wk.weight": "model.safetensors",
566
+ "llm.layers.8.attention.wo.scales": "model.safetensors",
567
+ "llm.layers.8.attention.wo.weight": "model.safetensors",
568
+ "llm.layers.8.attention.wq.scales": "model.safetensors",
569
+ "llm.layers.8.attention.wq.weight": "model.safetensors",
570
+ "llm.layers.8.attention.wv.scales": "model.safetensors",
571
+ "llm.layers.8.attention.wv.weight": "model.safetensors",
572
+ "llm.layers.8.attention_norm.weight": "model.safetensors",
573
+ "llm.layers.8.feed_forward_w1.scales": "model.safetensors",
574
+ "llm.layers.8.feed_forward_w1.weight": "model.safetensors",
575
+ "llm.layers.8.feed_forward_w2.scales": "model.safetensors",
576
+ "llm.layers.8.feed_forward_w2.weight": "model.safetensors",
577
+ "llm.layers.8.feed_forward_w3.scales": "model.safetensors",
578
+ "llm.layers.8.feed_forward_w3.weight": "model.safetensors",
579
+ "llm.layers.8.ffn_norm.weight": "model.safetensors",
580
+ "llm.layers.9.attention.wk.scales": "model.safetensors",
581
+ "llm.layers.9.attention.wk.weight": "model.safetensors",
582
+ "llm.layers.9.attention.wo.scales": "model.safetensors",
583
+ "llm.layers.9.attention.wo.weight": "model.safetensors",
584
+ "llm.layers.9.attention.wq.scales": "model.safetensors",
585
+ "llm.layers.9.attention.wq.weight": "model.safetensors",
586
+ "llm.layers.9.attention.wv.scales": "model.safetensors",
587
+ "llm.layers.9.attention.wv.weight": "model.safetensors",
588
+ "llm.layers.9.attention_norm.weight": "model.safetensors",
589
+ "llm.layers.9.feed_forward_w1.scales": "model.safetensors",
590
+ "llm.layers.9.feed_forward_w1.weight": "model.safetensors",
591
+ "llm.layers.9.feed_forward_w2.scales": "model.safetensors",
592
+ "llm.layers.9.feed_forward_w2.weight": "model.safetensors",
593
+ "llm.layers.9.feed_forward_w3.scales": "model.safetensors",
594
+ "llm.layers.9.feed_forward_w3.weight": "model.safetensors",
595
+ "llm.layers.9.ffn_norm.weight": "model.safetensors",
596
+ "llm.norm.weight": "model.safetensors",
597
+ "llm.tok_embeddings.weight": "model.safetensors"
598
+ }
599
+ }
params.json ADDED
@@ -0,0 +1,130 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dim": 3072,
3
+ "n_layers": 26,
4
+ "head_dim": 128,
5
+ "hidden_dim": 9216,
6
+ "n_heads": 32,
7
+ "n_kv_heads": 8,
8
+ "fp8_matmul": false,
9
+ "use_biases": false,
10
+ "causal": true,
11
+ "rope_theta": 1000000.0,
12
+ "norm_eps": 1e-05,
13
+ "init": "NO_INIT",
14
+ "dropout": 0.0,
15
+ "vocab_size": 131072,
16
+ "model_parallel": 1,
17
+ "is_sequence_parallel": false,
18
+ "context_parallel": 1,
19
+ "tied_embeddings": true,
20
+ "shard_on_vocab_dim": false,
21
+ "model_pipelining": 1,
22
+ "virtual_model_pipelining": 1,
23
+ "fused_rms_norm": true,
24
+ "checkpoint": false,
25
+ "use_cache": false,
26
+ "max_concurrent_tokens": 65536,
27
+ "learnable_sinks": false,
28
+ "rms_norm": "PRE",
29
+ "cust_bwd": false,
30
+ "recompute_w1_every": 0,
31
+ "recompute_w3_every": 0,
32
+ "recompute_attn_every": 0,
33
+ "freeze_nonembedding": false,
34
+ "fsdp2": true,
35
+ "dp_replicate_size": 1,
36
+ "zero2": true,
37
+ "fsdp_optimize_backward_concat_if_pp": true,
38
+ "attention_type": "FLASH_ATTN_3",
39
+ "multimodal": {
40
+ "bos_token_id": 1,
41
+ "audio_model_args": {
42
+ "semantic_codebook_size": 8192,
43
+ "acoustic_codebook_size": 21,
44
+ "n_acoustic_codebook": 36,
45
+ "audio_encoding_args": {
46
+ "codebook_pattern": "parallel",
47
+ "interleave_audio_tokens_per_segment": 8192,
48
+ "interleave_text_tokens_per_segment": 8192,
49
+ "single_trailing_segment": false,
50
+ "num_codebooks": 37,
51
+ "sampling_rate": 24000,
52
+ "frame_rate": 12.5
53
+ },
54
+ "audio_token_id": 24,
55
+ "begin_audio_token_id": 25,
56
+ "input_embedding_concat_type": "sum",
57
+ "acoustic_transformer_args": {
58
+ "input_dim": 3072,
59
+ "dim": 3072,
60
+ "n_layers": 3,
61
+ "head_dim": 128,
62
+ "hidden_dim": 9216,
63
+ "n_heads": 32,
64
+ "n_kv_heads": 8,
65
+ "use_biases": false,
66
+ "rope_theta": 10000.0,
67
+ "sigma": 1e-05,
68
+ "sigma_max": 1.0
69
+ },
70
+ "p_uncond": 0.0,
71
+ "text_feature_bugged": false,
72
+ "condition_dropped_token_id": 42
73
+ },
74
+ "audio_tokenizer_args": {
75
+ "channels": 1,
76
+ "sampling_rate": 24000,
77
+ "pretransform_patch_size": 240,
78
+ "patch_proj_kernel_size": 7,
79
+ "semantic_codebook_size": 8192,
80
+ "semantic_dim": 256,
81
+ "acoustic_codebook_size": 21,
82
+ "acoustic_dim": 36,
83
+ "conv_weight_norm": true,
84
+ "causal": true,
85
+ "attn_sliding_window_size": 16,
86
+ "half_attn_window_upon_downsampling": true,
87
+ "dim": 1024,
88
+ "hidden_dim": 4096,
89
+ "head_dim": 128,
90
+ "n_heads": 8,
91
+ "n_kv_heads": 8,
92
+ "qk_norm_eps": 1e-06,
93
+ "qk_norm": true,
94
+ "use_biases": false,
95
+ "norm_eps": 0.01,
96
+ "layer_scale": true,
97
+ "layer_scale_init": 0.01,
98
+ "decoder_transformer_lengths_str": "2,2,2,2",
99
+ "decoder_convs_kernels_str": "3,4,4,4",
100
+ "decoder_convs_strides_str": "1,2,2,2",
101
+ "voice": {
102
+ "casual_female": 0,
103
+ "casual_male": 1,
104
+ "cheerful_female": 2,
105
+ "neutral_female": 3,
106
+ "neutral_male": 4,
107
+ "pt_male": 5,
108
+ "pt_female": 6,
109
+ "nl_male": 7,
110
+ "nl_female": 8,
111
+ "it_male": 9,
112
+ "it_female": 10,
113
+ "fr_male": 11,
114
+ "fr_female": 12,
115
+ "es_male": 13,
116
+ "es_female": 14,
117
+ "de_male": 15,
118
+ "de_female": 16,
119
+ "ar_male": 17,
120
+ "hi_male": 18,
121
+ "hi_female": 19
122
+ }
123
+ }
124
+ },
125
+ "torch_compile_swiglu_noncust_bwd": false,
126
+ "override_parameters_str": "",
127
+ "max_seq_len": 65536,
128
+ "model_type": "voxtral_tts",
129
+ "max_position_embeddings": 128000
130
+ }
tekken.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:587989c9f56676b35e7d16d6fc61461301e402d908392a8ce16f0349f61b56d7
3
+ size 14894731