mzbac commited on
Commit
f3a0032
·
verified ·
1 Parent(s): bb53e9f

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. config.json +86 -0
  2. model.safetensors +3 -0
  3. quantization.json +3423 -0
config.json ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "acoustic_vae_dim": 64,
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
+ "encoder_depths": "3-3-3-3-3-3-8",
21
+ "encoder_n_filters": 32,
22
+ "encoder_ratios": [
23
+ 8,
24
+ 5,
25
+ 5,
26
+ 4,
27
+ 2,
28
+ 2
29
+ ],
30
+ "fix_std": 0.5,
31
+ "layer_scale_init_value": 1e-06,
32
+ "layernorm": "RMSNorm",
33
+ "layernorm_elementwise_affine": true,
34
+ "layernorm_eps": 1e-05,
35
+ "mixer_layer": "depthwise_conv",
36
+ "model_type": "vibevoice_acoustic_tokenizer",
37
+ "pad_mode": "constant",
38
+ "std_dist_type": "gaussian",
39
+ "vae_dim": 64,
40
+ "weight_init_value": 0.01
41
+ },
42
+ "architectures": [
43
+ "VibeVoiceStreamingForConditionalGenerationInference"
44
+ ],
45
+ "decoder_config": {
46
+ "attention_dropout": 0.0,
47
+ "hidden_act": "silu",
48
+ "hidden_size": 896,
49
+ "initializer_range": 0.02,
50
+ "intermediate_size": 4864,
51
+ "max_position_embeddings": 8192,
52
+ "max_window_layers": 24,
53
+ "model_type": "qwen2",
54
+ "num_attention_heads": 14,
55
+ "num_hidden_layers": 24,
56
+ "num_key_value_heads": 2,
57
+ "rms_norm_eps": 1e-06,
58
+ "rope_scaling": null,
59
+ "rope_theta": 1000000.0,
60
+ "sliding_window": null,
61
+ "tie_word_embeddings": false,
62
+ "torch_dtype": "bfloat16",
63
+ "use_cache": true,
64
+ "use_sliding_window": false,
65
+ "vocab_size": 151936
66
+ },
67
+ "diffusion_head_config": {
68
+ "ddpm_batch_mul": 4,
69
+ "ddpm_beta_schedule": "cosine",
70
+ "ddpm_num_inference_steps": 20,
71
+ "ddpm_num_steps": 1000,
72
+ "diffusion_type": "ddpm",
73
+ "head_ffn_ratio": 3.0,
74
+ "head_layers": 4,
75
+ "hidden_size": 896,
76
+ "latent_size": 64,
77
+ "model_type": "vibevoice_diffusion_head",
78
+ "prediction_type": "v_prediction",
79
+ "rms_norm_eps": 1e-05,
80
+ "speech_vae_dim": 64
81
+ },
82
+ "model_type": "vibevoice_streaming",
83
+ "torch_dtype": "bfloat16",
84
+ "transformers_version": "4.51.3",
85
+ "tts_backbone_num_hidden_layers": 20
86
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78af36493494fd0c914bda22119249b059e33224f812c6e584d559dd2b84a521
3
+ size 1508326329
quantization.json ADDED
@@ -0,0 +1,3423 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bits" : 8,
3
+ "group_size" : 32,
4
+ "layers" : [
5
+ {
6
+ "bits" : 8,
7
+ "file" : "model.safetensors",
8
+ "group_size" : 32,
9
+ "in_dim" : 896,
10
+ "mode" : "affine",
11
+ "name" : "model.tts_language_model.layers.1.mlp.gate_proj",
12
+ "out_dim" : 4864,
13
+ "quant_file" : "model.safetensors",
14
+ "shape" : [
15
+ 4864,
16
+ 896
17
+ ]
18
+ },
19
+ {
20
+ "bits" : 8,
21
+ "file" : "model.safetensors",
22
+ "group_size" : 32,
23
+ "in_dim" : 896,
24
+ "mode" : "affine",
25
+ "name" : "model.tts_language_model.layers.3.mlp.gate_proj",
26
+ "out_dim" : 4864,
27
+ "quant_file" : "model.safetensors",
28
+ "shape" : [
29
+ 4864,
30
+ 896
31
+ ]
32
+ },
33
+ {
34
+ "bits" : 8,
35
+ "file" : "model.safetensors",
36
+ "group_size" : 32,
37
+ "in_dim" : 896,
38
+ "mode" : "affine",
39
+ "name" : "model.tts_language_model.layers.13.self_attn.k_proj",
40
+ "out_dim" : 128,
41
+ "quant_file" : "model.safetensors",
42
+ "shape" : [
43
+ 128,
44
+ 896
45
+ ]
46
+ },
47
+ {
48
+ "bits" : 8,
49
+ "file" : "model.safetensors",
50
+ "group_size" : 32,
51
+ "in_dim" : 4864,
52
+ "mode" : "affine",
53
+ "name" : "model.tts_language_model.layers.18.mlp.down_proj",
54
+ "out_dim" : 896,
55
+ "quant_file" : "model.safetensors",
56
+ "shape" : [
57
+ 896,
58
+ 4864
59
+ ]
60
+ },
61
+ {
62
+ "bits" : 8,
63
+ "file" : "model.safetensors",
64
+ "group_size" : 32,
65
+ "in_dim" : 1024,
66
+ "mode" : "affine",
67
+ "name" : "model.acoustic_tokenizer.decoder.stages.1.2.ffn.linear1",
68
+ "out_dim" : 4096,
69
+ "quant_file" : "model.safetensors",
70
+ "shape" : [
71
+ 4096,
72
+ 1024
73
+ ]
74
+ },
75
+ {
76
+ "bits" : 8,
77
+ "file" : "model.safetensors",
78
+ "group_size" : 32,
79
+ "in_dim" : 896,
80
+ "mode" : "affine",
81
+ "name" : "model.language_model.layers.2.self_attn.o_proj",
82
+ "out_dim" : 896,
83
+ "quant_file" : "model.safetensors",
84
+ "shape" : [
85
+ 896,
86
+ 896
87
+ ]
88
+ },
89
+ {
90
+ "bits" : 8,
91
+ "file" : "model.safetensors",
92
+ "group_size" : 32,
93
+ "in_dim" : 896,
94
+ "mode" : "affine",
95
+ "name" : "model.tts_language_model.layers.11.self_attn.k_proj",
96
+ "out_dim" : 128,
97
+ "quant_file" : "model.safetensors",
98
+ "shape" : [
99
+ 128,
100
+ 896
101
+ ]
102
+ },
103
+ {
104
+ "bits" : 8,
105
+ "file" : "model.safetensors",
106
+ "group_size" : 32,
107
+ "in_dim" : 896,
108
+ "mode" : "affine",
109
+ "name" : "model.language_model.layers.1.self_attn.k_proj",
110
+ "out_dim" : 128,
111
+ "quant_file" : "model.safetensors",
112
+ "shape" : [
113
+ 128,
114
+ 896
115
+ ]
116
+ },
117
+ {
118
+ "bits" : 8,
119
+ "file" : "model.safetensors",
120
+ "group_size" : 32,
121
+ "in_dim" : 512,
122
+ "mode" : "affine",
123
+ "name" : "model.acoustic_tokenizer.decoder.stages.4.2.ffn.linear2",
124
+ "out_dim" : 128,
125
+ "quant_file" : "model.safetensors",
126
+ "shape" : [
127
+ 128,
128
+ 512
129
+ ]
130
+ },
131
+ {
132
+ "bits" : 8,
133
+ "file" : "model.safetensors",
134
+ "group_size" : 32,
135
+ "in_dim" : 896,
136
+ "mode" : "affine",
137
+ "name" : "model.tts_language_model.layers.8.self_attn.v_proj",
138
+ "out_dim" : 128,
139
+ "quant_file" : "model.safetensors",
140
+ "shape" : [
141
+ 128,
142
+ 896
143
+ ]
144
+ },
145
+ {
146
+ "bits" : 8,
147
+ "file" : "model.safetensors",
148
+ "group_size" : 32,
149
+ "in_dim" : 896,
150
+ "mode" : "affine",
151
+ "name" : "model.tts_language_model.layers.15.mlp.up_proj",
152
+ "out_dim" : 4864,
153
+ "quant_file" : "model.safetensors",
154
+ "shape" : [
155
+ 4864,
156
+ 896
157
+ ]
158
+ },
159
+ {
160
+ "bits" : 8,
161
+ "file" : "model.safetensors",
162
+ "group_size" : 32,
163
+ "in_dim" : 896,
164
+ "mode" : "affine",
165
+ "name" : "model.prediction_head.layers.0.adaLN_modulation.1",
166
+ "out_dim" : 2688,
167
+ "quant_file" : "model.safetensors",
168
+ "shape" : [
169
+ 2688,
170
+ 896
171
+ ]
172
+ },
173
+ {
174
+ "bits" : 8,
175
+ "file" : "model.safetensors",
176
+ "group_size" : 32,
177
+ "in_dim" : 896,
178
+ "mode" : "affine",
179
+ "name" : "model.tts_language_model.layers.13.self_attn.o_proj",
180
+ "out_dim" : 896,
181
+ "quant_file" : "model.safetensors",
182
+ "shape" : [
183
+ 896,
184
+ 896
185
+ ]
186
+ },
187
+ {
188
+ "bits" : 8,
189
+ "file" : "model.safetensors",
190
+ "group_size" : 32,
191
+ "in_dim" : 128,
192
+ "mode" : "affine",
193
+ "name" : "model.acoustic_tokenizer.decoder.stages.6.0.ffn.linear2",
194
+ "out_dim" : 32,
195
+ "quant_file" : "model.safetensors",
196
+ "shape" : [
197
+ 32,
198
+ 128
199
+ ]
200
+ },
201
+ {
202
+ "bits" : 8,
203
+ "file" : "model.safetensors",
204
+ "group_size" : 32,
205
+ "in_dim" : 896,
206
+ "mode" : "affine",
207
+ "name" : "model.tts_language_model.layers.16.self_attn.q_proj",
208
+ "out_dim" : 896,
209
+ "quant_file" : "model.safetensors",
210
+ "shape" : [
211
+ 896,
212
+ 896
213
+ ]
214
+ },
215
+ {
216
+ "bits" : 8,
217
+ "file" : "model.safetensors",
218
+ "group_size" : 32,
219
+ "in_dim" : 896,
220
+ "mode" : "affine",
221
+ "name" : "model.tts_language_model.layers.5.self_attn.k_proj",
222
+ "out_dim" : 128,
223
+ "quant_file" : "model.safetensors",
224
+ "shape" : [
225
+ 128,
226
+ 896
227
+ ]
228
+ },
229
+ {
230
+ "bits" : 8,
231
+ "file" : "model.safetensors",
232
+ "group_size" : 32,
233
+ "in_dim" : 896,
234
+ "mode" : "affine",
235
+ "name" : "model.tts_language_model.layers.19.mlp.gate_proj",
236
+ "out_dim" : 4864,
237
+ "quant_file" : "model.safetensors",
238
+ "shape" : [
239
+ 4864,
240
+ 896
241
+ ]
242
+ },
243
+ {
244
+ "bits" : 8,
245
+ "file" : "model.safetensors",
246
+ "group_size" : 32,
247
+ "in_dim" : 1024,
248
+ "mode" : "affine",
249
+ "name" : "model.acoustic_tokenizer.decoder.stages.1.0.ffn.linear1",
250
+ "out_dim" : 4096,
251
+ "quant_file" : "model.safetensors",
252
+ "shape" : [
253
+ 4096,
254
+ 1024
255
+ ]
256
+ },
257
+ {
258
+ "bits" : 8,
259
+ "file" : "model.safetensors",
260
+ "group_size" : 32,
261
+ "in_dim" : 896,
262
+ "mode" : "affine",
263
+ "name" : "model.tts_language_model.layers.4.mlp.up_proj",
264
+ "out_dim" : 4864,
265
+ "quant_file" : "model.safetensors",
266
+ "shape" : [
267
+ 4864,
268
+ 896
269
+ ]
270
+ },
271
+ {
272
+ "bits" : 8,
273
+ "file" : "model.safetensors",
274
+ "group_size" : 32,
275
+ "in_dim" : 896,
276
+ "mode" : "affine",
277
+ "name" : "model.tts_language_model.layers.14.self_attn.o_proj",
278
+ "out_dim" : 896,
279
+ "quant_file" : "model.safetensors",
280
+ "shape" : [
281
+ 896,
282
+ 896
283
+ ]
284
+ },
285
+ {
286
+ "bits" : 8,
287
+ "file" : "model.safetensors",
288
+ "group_size" : 32,
289
+ "in_dim" : 512,
290
+ "mode" : "affine",
291
+ "name" : "model.acoustic_tokenizer.decoder.stages.4.1.ffn.linear2",
292
+ "out_dim" : 128,
293
+ "quant_file" : "model.safetensors",
294
+ "shape" : [
295
+ 128,
296
+ 512
297
+ ]
298
+ },
299
+ {
300
+ "bits" : 8,
301
+ "file" : "model.safetensors",
302
+ "group_size" : 32,
303
+ "in_dim" : 64,
304
+ "mode" : "affine",
305
+ "name" : "model.acoustic_tokenizer.decoder.stages.5.2.ffn.linear1",
306
+ "out_dim" : 256,
307
+ "quant_file" : "model.safetensors",
308
+ "shape" : [
309
+ 256,
310
+ 64
311
+ ]
312
+ },
313
+ {
314
+ "bits" : 8,
315
+ "file" : "model.safetensors",
316
+ "group_size" : 32,
317
+ "in_dim" : 896,
318
+ "mode" : "affine",
319
+ "name" : "model.tts_language_model.layers.11.mlp.up_proj",
320
+ "out_dim" : 4864,
321
+ "quant_file" : "model.safetensors",
322
+ "shape" : [
323
+ 4864,
324
+ 896
325
+ ]
326
+ },
327
+ {
328
+ "bits" : 8,
329
+ "file" : "model.safetensors",
330
+ "group_size" : 32,
331
+ "in_dim" : 896,
332
+ "mode" : "affine",
333
+ "name" : "model.prediction_head.layers.1.ffn.gate_proj",
334
+ "out_dim" : 2688,
335
+ "quant_file" : "model.safetensors",
336
+ "shape" : [
337
+ 2688,
338
+ 896
339
+ ]
340
+ },
341
+ {
342
+ "bits" : 8,
343
+ "file" : "model.safetensors",
344
+ "group_size" : 32,
345
+ "in_dim" : 896,
346
+ "mode" : "affine",
347
+ "name" : "model.tts_language_model.layers.14.self_attn.q_proj",
348
+ "out_dim" : 896,
349
+ "quant_file" : "model.safetensors",
350
+ "shape" : [
351
+ 896,
352
+ 896
353
+ ]
354
+ },
355
+ {
356
+ "bits" : 8,
357
+ "file" : "model.safetensors",
358
+ "group_size" : 32,
359
+ "in_dim" : 896,
360
+ "mode" : "affine",
361
+ "name" : "model.tts_language_model.layers.16.mlp.gate_proj",
362
+ "out_dim" : 4864,
363
+ "quant_file" : "model.safetensors",
364
+ "shape" : [
365
+ 4864,
366
+ 896
367
+ ]
368
+ },
369
+ {
370
+ "bits" : 8,
371
+ "file" : "model.safetensors",
372
+ "group_size" : 32,
373
+ "in_dim" : 896,
374
+ "mode" : "affine",
375
+ "name" : "model.tts_language_model.layers.6.self_attn.k_proj",
376
+ "out_dim" : 128,
377
+ "quant_file" : "model.safetensors",
378
+ "shape" : [
379
+ 128,
380
+ 896
381
+ ]
382
+ },
383
+ {
384
+ "bits" : 8,
385
+ "file" : "model.safetensors",
386
+ "group_size" : 32,
387
+ "in_dim" : 896,
388
+ "mode" : "affine",
389
+ "name" : "model.prediction_head.layers.2.adaLN_modulation.1",
390
+ "out_dim" : 2688,
391
+ "quant_file" : "model.safetensors",
392
+ "shape" : [
393
+ 2688,
394
+ 896
395
+ ]
396
+ },
397
+ {
398
+ "bits" : 8,
399
+ "file" : "model.safetensors",
400
+ "group_size" : 32,
401
+ "in_dim" : 896,
402
+ "mode" : "affine",
403
+ "name" : "model.tts_language_model.layers.6.self_attn.v_proj",
404
+ "out_dim" : 128,
405
+ "quant_file" : "model.safetensors",
406
+ "shape" : [
407
+ 128,
408
+ 896
409
+ ]
410
+ },
411
+ {
412
+ "bits" : 8,
413
+ "file" : "model.safetensors",
414
+ "group_size" : 32,
415
+ "in_dim" : 512,
416
+ "mode" : "affine",
417
+ "name" : "model.acoustic_tokenizer.decoder.stages.4.0.ffn.linear2",
418
+ "out_dim" : 128,
419
+ "quant_file" : "model.safetensors",
420
+ "shape" : [
421
+ 128,
422
+ 512
423
+ ]
424
+ },
425
+ {
426
+ "bits" : 8,
427
+ "file" : "model.safetensors",
428
+ "group_size" : 32,
429
+ "in_dim" : 8192,
430
+ "mode" : "affine",
431
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.4.ffn.linear2",
432
+ "out_dim" : 2048,
433
+ "quant_file" : "model.safetensors",
434
+ "shape" : [
435
+ 2048,
436
+ 8192
437
+ ]
438
+ },
439
+ {
440
+ "bits" : 8,
441
+ "file" : "model.safetensors",
442
+ "group_size" : 32,
443
+ "in_dim" : 896,
444
+ "mode" : "affine",
445
+ "name" : "model.tts_language_model.layers.7.mlp.up_proj",
446
+ "out_dim" : 4864,
447
+ "quant_file" : "model.safetensors",
448
+ "shape" : [
449
+ 4864,
450
+ 896
451
+ ]
452
+ },
453
+ {
454
+ "bits" : 8,
455
+ "file" : "model.safetensors",
456
+ "group_size" : 32,
457
+ "in_dim" : 896,
458
+ "mode" : "affine",
459
+ "name" : "model.tts_language_model.layers.0.self_attn.k_proj",
460
+ "out_dim" : 128,
461
+ "quant_file" : "model.safetensors",
462
+ "shape" : [
463
+ 128,
464
+ 896
465
+ ]
466
+ },
467
+ {
468
+ "bits" : 8,
469
+ "file" : "model.safetensors",
470
+ "group_size" : 32,
471
+ "in_dim" : 896,
472
+ "mode" : "affine",
473
+ "name" : "model.tts_language_model.layers.19.self_attn.q_proj",
474
+ "out_dim" : 896,
475
+ "quant_file" : "model.safetensors",
476
+ "shape" : [
477
+ 896,
478
+ 896
479
+ ]
480
+ },
481
+ {
482
+ "bits" : 8,
483
+ "file" : "model.safetensors",
484
+ "group_size" : 32,
485
+ "in_dim" : 896,
486
+ "mode" : "affine",
487
+ "name" : "model.language_model.layers.2.mlp.gate_proj",
488
+ "out_dim" : 4864,
489
+ "quant_file" : "model.safetensors",
490
+ "shape" : [
491
+ 4864,
492
+ 896
493
+ ]
494
+ },
495
+ {
496
+ "bits" : 8,
497
+ "file" : "model.safetensors",
498
+ "group_size" : 32,
499
+ "in_dim" : 2048,
500
+ "mode" : "affine",
501
+ "name" : "model.acoustic_tokenizer.decoder.stages.2.2.ffn.linear2",
502
+ "out_dim" : 512,
503
+ "quant_file" : "model.safetensors",
504
+ "shape" : [
505
+ 512,
506
+ 2048
507
+ ]
508
+ },
509
+ {
510
+ "bits" : 8,
511
+ "file" : "model.safetensors",
512
+ "group_size" : 32,
513
+ "in_dim" : 512,
514
+ "mode" : "affine",
515
+ "name" : "model.acoustic_tokenizer.decoder.stages.2.2.ffn.linear1",
516
+ "out_dim" : 2048,
517
+ "quant_file" : "model.safetensors",
518
+ "shape" : [
519
+ 2048,
520
+ 512
521
+ ]
522
+ },
523
+ {
524
+ "bits" : 8,
525
+ "file" : "model.safetensors",
526
+ "group_size" : 32,
527
+ "in_dim" : 896,
528
+ "mode" : "affine",
529
+ "name" : "model.tts_language_model.layers.13.self_attn.v_proj",
530
+ "out_dim" : 128,
531
+ "quant_file" : "model.safetensors",
532
+ "shape" : [
533
+ 128,
534
+ 896
535
+ ]
536
+ },
537
+ {
538
+ "bits" : 8,
539
+ "file" : "model.safetensors",
540
+ "group_size" : 32,
541
+ "in_dim" : 896,
542
+ "mode" : "affine",
543
+ "name" : "model.language_model.layers.3.self_attn.k_proj",
544
+ "out_dim" : 128,
545
+ "quant_file" : "model.safetensors",
546
+ "shape" : [
547
+ 128,
548
+ 896
549
+ ]
550
+ },
551
+ {
552
+ "bits" : 8,
553
+ "file" : "model.safetensors",
554
+ "group_size" : 32,
555
+ "in_dim" : 896,
556
+ "mode" : "affine",
557
+ "name" : "model.tts_language_model.layers.7.self_attn.o_proj",
558
+ "out_dim" : 896,
559
+ "quant_file" : "model.safetensors",
560
+ "shape" : [
561
+ 896,
562
+ 896
563
+ ]
564
+ },
565
+ {
566
+ "bits" : 8,
567
+ "file" : "model.safetensors",
568
+ "group_size" : 32,
569
+ "in_dim" : 1024,
570
+ "mode" : "affine",
571
+ "name" : "model.acoustic_tokenizer.decoder.stages.3.1.ffn.linear2",
572
+ "out_dim" : 256,
573
+ "quant_file" : "model.safetensors",
574
+ "shape" : [
575
+ 256,
576
+ 1024
577
+ ]
578
+ },
579
+ {
580
+ "bits" : 8,
581
+ "file" : "model.safetensors",
582
+ "group_size" : 32,
583
+ "in_dim" : 896,
584
+ "mode" : "affine",
585
+ "name" : "model.tts_language_model.layers.12.self_attn.k_proj",
586
+ "out_dim" : 128,
587
+ "quant_file" : "model.safetensors",
588
+ "shape" : [
589
+ 128,
590
+ 896
591
+ ]
592
+ },
593
+ {
594
+ "bits" : 8,
595
+ "file" : "model.safetensors",
596
+ "group_size" : 32,
597
+ "in_dim" : 896,
598
+ "mode" : "affine",
599
+ "name" : "model.language_model.layers.1.self_attn.v_proj",
600
+ "out_dim" : 128,
601
+ "quant_file" : "model.safetensors",
602
+ "shape" : [
603
+ 128,
604
+ 896
605
+ ]
606
+ },
607
+ {
608
+ "bits" : 8,
609
+ "file" : "model.safetensors",
610
+ "group_size" : 32,
611
+ "in_dim" : 1024,
612
+ "mode" : "affine",
613
+ "name" : "model.acoustic_tokenizer.decoder.stages.3.0.ffn.linear2",
614
+ "out_dim" : 256,
615
+ "quant_file" : "model.safetensors",
616
+ "shape" : [
617
+ 256,
618
+ 1024
619
+ ]
620
+ },
621
+ {
622
+ "bits" : 8,
623
+ "file" : "model.safetensors",
624
+ "group_size" : 32,
625
+ "in_dim" : 896,
626
+ "mode" : "affine",
627
+ "name" : "model.tts_language_model.layers.10.mlp.up_proj",
628
+ "out_dim" : 4864,
629
+ "quant_file" : "model.safetensors",
630
+ "shape" : [
631
+ 4864,
632
+ 896
633
+ ]
634
+ },
635
+ {
636
+ "bits" : 8,
637
+ "file" : "model.safetensors",
638
+ "group_size" : 32,
639
+ "in_dim" : 896,
640
+ "mode" : "affine",
641
+ "name" : "model.tts_language_model.layers.4.mlp.gate_proj",
642
+ "out_dim" : 4864,
643
+ "quant_file" : "model.safetensors",
644
+ "shape" : [
645
+ 4864,
646
+ 896
647
+ ]
648
+ },
649
+ {
650
+ "bits" : 8,
651
+ "file" : "model.safetensors",
652
+ "group_size" : 32,
653
+ "in_dim" : 896,
654
+ "mode" : "affine",
655
+ "name" : "model.tts_language_model.layers.17.self_attn.q_proj",
656
+ "out_dim" : 896,
657
+ "quant_file" : "model.safetensors",
658
+ "shape" : [
659
+ 896,
660
+ 896
661
+ ]
662
+ },
663
+ {
664
+ "bits" : 8,
665
+ "file" : "model.safetensors",
666
+ "group_size" : 32,
667
+ "in_dim" : 896,
668
+ "mode" : "affine",
669
+ "name" : "model.tts_language_model.layers.10.self_attn.k_proj",
670
+ "out_dim" : 128,
671
+ "quant_file" : "model.safetensors",
672
+ "shape" : [
673
+ 128,
674
+ 896
675
+ ]
676
+ },
677
+ {
678
+ "bits" : 8,
679
+ "file" : "model.safetensors",
680
+ "group_size" : 32,
681
+ "in_dim" : 896,
682
+ "mode" : "affine",
683
+ "name" : "model.prediction_head.layers.3.adaLN_modulation.1",
684
+ "out_dim" : 2688,
685
+ "quant_file" : "model.safetensors",
686
+ "shape" : [
687
+ 2688,
688
+ 896
689
+ ]
690
+ },
691
+ {
692
+ "bits" : 8,
693
+ "file" : "model.safetensors",
694
+ "group_size" : 32,
695
+ "in_dim" : 896,
696
+ "mode" : "affine",
697
+ "name" : "model.language_model.layers.2.self_attn.k_proj",
698
+ "out_dim" : 128,
699
+ "quant_file" : "model.safetensors",
700
+ "shape" : [
701
+ 128,
702
+ 896
703
+ ]
704
+ },
705
+ {
706
+ "bits" : 8,
707
+ "file" : "model.safetensors",
708
+ "group_size" : 32,
709
+ "in_dim" : 896,
710
+ "mode" : "affine",
711
+ "name" : "model.tts_language_model.layers.9.mlp.gate_proj",
712
+ "out_dim" : 4864,
713
+ "quant_file" : "model.safetensors",
714
+ "shape" : [
715
+ 4864,
716
+ 896
717
+ ]
718
+ },
719
+ {
720
+ "bits" : 8,
721
+ "file" : "model.safetensors",
722
+ "group_size" : 32,
723
+ "in_dim" : 2048,
724
+ "mode" : "affine",
725
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.6.ffn.linear1",
726
+ "out_dim" : 8192,
727
+ "quant_file" : "model.safetensors",
728
+ "shape" : [
729
+ 8192,
730
+ 2048
731
+ ]
732
+ },
733
+ {
734
+ "bits" : 8,
735
+ "file" : "model.safetensors",
736
+ "group_size" : 32,
737
+ "in_dim" : 896,
738
+ "mode" : "affine",
739
+ "name" : "model.prediction_head.cond_proj",
740
+ "out_dim" : 896,
741
+ "quant_file" : "model.safetensors",
742
+ "shape" : [
743
+ 896,
744
+ 896
745
+ ]
746
+ },
747
+ {
748
+ "bits" : 8,
749
+ "file" : "model.safetensors",
750
+ "group_size" : 32,
751
+ "in_dim" : 896,
752
+ "mode" : "affine",
753
+ "name" : "model.tts_language_model.layers.2.self_attn.k_proj",
754
+ "out_dim" : 128,
755
+ "quant_file" : "model.safetensors",
756
+ "shape" : [
757
+ 128,
758
+ 896
759
+ ]
760
+ },
761
+ {
762
+ "bits" : 8,
763
+ "file" : "model.safetensors",
764
+ "group_size" : 32,
765
+ "in_dim" : 896,
766
+ "mode" : "affine",
767
+ "name" : "model.tts_language_model.layers.15.mlp.gate_proj",
768
+ "out_dim" : 4864,
769
+ "quant_file" : "model.safetensors",
770
+ "shape" : [
771
+ 4864,
772
+ 896
773
+ ]
774
+ },
775
+ {
776
+ "bits" : 8,
777
+ "file" : "model.safetensors",
778
+ "group_size" : 32,
779
+ "in_dim" : 896,
780
+ "mode" : "affine",
781
+ "name" : "model.tts_language_model.layers.1.self_attn.k_proj",
782
+ "out_dim" : 128,
783
+ "quant_file" : "model.safetensors",
784
+ "shape" : [
785
+ 128,
786
+ 896
787
+ ]
788
+ },
789
+ {
790
+ "bits" : 8,
791
+ "file" : "model.safetensors",
792
+ "group_size" : 32,
793
+ "in_dim" : 4864,
794
+ "mode" : "affine",
795
+ "name" : "model.tts_language_model.layers.7.mlp.down_proj",
796
+ "out_dim" : 896,
797
+ "quant_file" : "model.safetensors",
798
+ "shape" : [
799
+ 896,
800
+ 4864
801
+ ]
802
+ },
803
+ {
804
+ "bits" : 8,
805
+ "file" : "model.safetensors",
806
+ "group_size" : 32,
807
+ "in_dim" : 896,
808
+ "mode" : "affine",
809
+ "name" : "model.language_model.layers.1.self_attn.o_proj",
810
+ "out_dim" : 896,
811
+ "quant_file" : "model.safetensors",
812
+ "shape" : [
813
+ 896,
814
+ 896
815
+ ]
816
+ },
817
+ {
818
+ "bits" : 8,
819
+ "file" : "model.safetensors",
820
+ "group_size" : 32,
821
+ "in_dim" : 896,
822
+ "mode" : "affine",
823
+ "name" : "model.tts_language_model.layers.19.self_attn.o_proj",
824
+ "out_dim" : 896,
825
+ "quant_file" : "model.safetensors",
826
+ "shape" : [
827
+ 896,
828
+ 896
829
+ ]
830
+ },
831
+ {
832
+ "bits" : 8,
833
+ "file" : "model.safetensors",
834
+ "group_size" : 32,
835
+ "in_dim" : 256,
836
+ "mode" : "affine",
837
+ "name" : "model.acoustic_tokenizer.decoder.stages.3.0.ffn.linear1",
838
+ "out_dim" : 1024,
839
+ "quant_file" : "model.safetensors",
840
+ "shape" : [
841
+ 1024,
842
+ 256
843
+ ]
844
+ },
845
+ {
846
+ "bits" : 8,
847
+ "file" : "model.safetensors",
848
+ "group_size" : 32,
849
+ "in_dim" : 896,
850
+ "mode" : "affine",
851
+ "name" : "model.tts_language_model.layers.11.self_attn.o_proj",
852
+ "out_dim" : 896,
853
+ "quant_file" : "model.safetensors",
854
+ "shape" : [
855
+ 896,
856
+ 896
857
+ ]
858
+ },
859
+ {
860
+ "bits" : 8,
861
+ "file" : "model.safetensors",
862
+ "group_size" : 32,
863
+ "in_dim" : 896,
864
+ "mode" : "affine",
865
+ "name" : "model.tts_language_model.layers.12.self_attn.o_proj",
866
+ "out_dim" : 896,
867
+ "quant_file" : "model.safetensors",
868
+ "shape" : [
869
+ 896,
870
+ 896
871
+ ]
872
+ },
873
+ {
874
+ "bits" : 8,
875
+ "file" : "model.safetensors",
876
+ "group_size" : 32,
877
+ "in_dim" : 896,
878
+ "mode" : "affine",
879
+ "name" : "model.tts_language_model.layers.9.mlp.up_proj",
880
+ "out_dim" : 4864,
881
+ "quant_file" : "model.safetensors",
882
+ "shape" : [
883
+ 4864,
884
+ 896
885
+ ]
886
+ },
887
+ {
888
+ "bits" : 8,
889
+ "file" : "model.safetensors",
890
+ "group_size" : 32,
891
+ "in_dim" : 896,
892
+ "mode" : "affine",
893
+ "name" : "model.tts_language_model.layers.0.mlp.gate_proj",
894
+ "out_dim" : 4864,
895
+ "quant_file" : "model.safetensors",
896
+ "shape" : [
897
+ 4864,
898
+ 896
899
+ ]
900
+ },
901
+ {
902
+ "bits" : 8,
903
+ "file" : "model.safetensors",
904
+ "group_size" : 32,
905
+ "in_dim" : 896,
906
+ "mode" : "affine",
907
+ "name" : "model.tts_language_model.layers.2.self_attn.v_proj",
908
+ "out_dim" : 128,
909
+ "quant_file" : "model.safetensors",
910
+ "shape" : [
911
+ 128,
912
+ 896
913
+ ]
914
+ },
915
+ {
916
+ "bits" : 8,
917
+ "file" : "model.safetensors",
918
+ "group_size" : 32,
919
+ "in_dim" : 2048,
920
+ "mode" : "affine",
921
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.7.ffn.linear1",
922
+ "out_dim" : 8192,
923
+ "quant_file" : "model.safetensors",
924
+ "shape" : [
925
+ 8192,
926
+ 2048
927
+ ]
928
+ },
929
+ {
930
+ "bits" : 8,
931
+ "file" : "model.safetensors",
932
+ "group_size" : 32,
933
+ "in_dim" : 256,
934
+ "mode" : "affine",
935
+ "name" : "model.acoustic_tokenizer.decoder.stages.5.0.ffn.linear2",
936
+ "out_dim" : 64,
937
+ "quant_file" : "model.safetensors",
938
+ "shape" : [
939
+ 64,
940
+ 256
941
+ ]
942
+ },
943
+ {
944
+ "bits" : 8,
945
+ "file" : "model.safetensors",
946
+ "group_size" : 32,
947
+ "in_dim" : 4864,
948
+ "mode" : "affine",
949
+ "name" : "model.tts_language_model.layers.16.mlp.down_proj",
950
+ "out_dim" : 896,
951
+ "quant_file" : "model.safetensors",
952
+ "shape" : [
953
+ 896,
954
+ 4864
955
+ ]
956
+ },
957
+ {
958
+ "bits" : 8,
959
+ "file" : "model.safetensors",
960
+ "group_size" : 32,
961
+ "in_dim" : 896,
962
+ "mode" : "affine",
963
+ "name" : "model.prediction_head.layers.0.ffn.up_proj",
964
+ "out_dim" : 2688,
965
+ "quant_file" : "model.safetensors",
966
+ "shape" : [
967
+ 2688,
968
+ 896
969
+ ]
970
+ },
971
+ {
972
+ "bits" : 8,
973
+ "file" : "model.safetensors",
974
+ "group_size" : 32,
975
+ "in_dim" : 896,
976
+ "mode" : "affine",
977
+ "name" : "model.tts_language_model.layers.3.self_attn.q_proj",
978
+ "out_dim" : 896,
979
+ "quant_file" : "model.safetensors",
980
+ "shape" : [
981
+ 896,
982
+ 896
983
+ ]
984
+ },
985
+ {
986
+ "bits" : 8,
987
+ "file" : "model.safetensors",
988
+ "group_size" : 32,
989
+ "in_dim" : 896,
990
+ "mode" : "affine",
991
+ "name" : "model.language_model.layers.3.mlp.gate_proj",
992
+ "out_dim" : 4864,
993
+ "quant_file" : "model.safetensors",
994
+ "shape" : [
995
+ 4864,
996
+ 896
997
+ ]
998
+ },
999
+ {
1000
+ "bits" : 8,
1001
+ "file" : "model.safetensors",
1002
+ "group_size" : 32,
1003
+ "in_dim" : 2048,
1004
+ "mode" : "affine",
1005
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.2.ffn.linear1",
1006
+ "out_dim" : 8192,
1007
+ "quant_file" : "model.safetensors",
1008
+ "shape" : [
1009
+ 8192,
1010
+ 2048
1011
+ ]
1012
+ },
1013
+ {
1014
+ "bits" : 8,
1015
+ "file" : "model.safetensors",
1016
+ "group_size" : 32,
1017
+ "in_dim" : 64,
1018
+ "mode" : "affine",
1019
+ "name" : "model.prediction_head.noisy_images_proj",
1020
+ "out_dim" : 896,
1021
+ "quant_file" : "model.safetensors",
1022
+ "shape" : [
1023
+ 896,
1024
+ 64
1025
+ ]
1026
+ },
1027
+ {
1028
+ "bits" : 8,
1029
+ "file" : "model.safetensors",
1030
+ "group_size" : 32,
1031
+ "in_dim" : 896,
1032
+ "mode" : "affine",
1033
+ "name" : "model.tts_language_model.layers.1.self_attn.q_proj",
1034
+ "out_dim" : 896,
1035
+ "quant_file" : "model.safetensors",
1036
+ "shape" : [
1037
+ 896,
1038
+ 896
1039
+ ]
1040
+ },
1041
+ {
1042
+ "bits" : 8,
1043
+ "file" : "model.safetensors",
1044
+ "group_size" : 32,
1045
+ "in_dim" : 896,
1046
+ "mode" : "affine",
1047
+ "name" : "model.tts_language_model.layers.0.self_attn.q_proj",
1048
+ "out_dim" : 896,
1049
+ "quant_file" : "model.safetensors",
1050
+ "shape" : [
1051
+ 896,
1052
+ 896
1053
+ ]
1054
+ },
1055
+ {
1056
+ "bits" : 8,
1057
+ "file" : "model.safetensors",
1058
+ "group_size" : 32,
1059
+ "in_dim" : 2048,
1060
+ "mode" : "affine",
1061
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.5.ffn.linear1",
1062
+ "out_dim" : 8192,
1063
+ "quant_file" : "model.safetensors",
1064
+ "shape" : [
1065
+ 8192,
1066
+ 2048
1067
+ ]
1068
+ },
1069
+ {
1070
+ "bits" : 8,
1071
+ "file" : "model.safetensors",
1072
+ "group_size" : 32,
1073
+ "in_dim" : 896,
1074
+ "mode" : "affine",
1075
+ "name" : "model.tts_language_model.layers.18.self_attn.q_proj",
1076
+ "out_dim" : 896,
1077
+ "quant_file" : "model.safetensors",
1078
+ "shape" : [
1079
+ 896,
1080
+ 896
1081
+ ]
1082
+ },
1083
+ {
1084
+ "bits" : 8,
1085
+ "file" : "model.safetensors",
1086
+ "group_size" : 32,
1087
+ "in_dim" : 1024,
1088
+ "mode" : "affine",
1089
+ "name" : "model.acoustic_tokenizer.decoder.stages.1.1.ffn.linear1",
1090
+ "out_dim" : 4096,
1091
+ "quant_file" : "model.safetensors",
1092
+ "shape" : [
1093
+ 4096,
1094
+ 1024
1095
+ ]
1096
+ },
1097
+ {
1098
+ "bits" : 8,
1099
+ "file" : "model.safetensors",
1100
+ "group_size" : 32,
1101
+ "in_dim" : 2688,
1102
+ "mode" : "affine",
1103
+ "name" : "model.prediction_head.layers.0.ffn.down_proj",
1104
+ "out_dim" : 896,
1105
+ "quant_file" : "model.safetensors",
1106
+ "shape" : [
1107
+ 896,
1108
+ 2688
1109
+ ]
1110
+ },
1111
+ {
1112
+ "bits" : 8,
1113
+ "file" : "model.safetensors",
1114
+ "group_size" : 32,
1115
+ "in_dim" : 896,
1116
+ "mode" : "affine",
1117
+ "name" : "model.language_model.layers.0.self_attn.o_proj",
1118
+ "out_dim" : 896,
1119
+ "quant_file" : "model.safetensors",
1120
+ "shape" : [
1121
+ 896,
1122
+ 896
1123
+ ]
1124
+ },
1125
+ {
1126
+ "bits" : 8,
1127
+ "file" : "model.safetensors",
1128
+ "group_size" : 32,
1129
+ "in_dim" : 896,
1130
+ "mode" : "affine",
1131
+ "name" : "model.tts_language_model.layers.5.self_attn.o_proj",
1132
+ "out_dim" : 896,
1133
+ "quant_file" : "model.safetensors",
1134
+ "shape" : [
1135
+ 896,
1136
+ 896
1137
+ ]
1138
+ },
1139
+ {
1140
+ "bits" : 8,
1141
+ "file" : "model.safetensors",
1142
+ "group_size" : 32,
1143
+ "in_dim" : 896,
1144
+ "mode" : "affine",
1145
+ "name" : "model.tts_language_model.layers.7.mlp.gate_proj",
1146
+ "out_dim" : 4864,
1147
+ "quant_file" : "model.safetensors",
1148
+ "shape" : [
1149
+ 4864,
1150
+ 896
1151
+ ]
1152
+ },
1153
+ {
1154
+ "bits" : 8,
1155
+ "file" : "model.safetensors",
1156
+ "group_size" : 32,
1157
+ "in_dim" : 4864,
1158
+ "mode" : "affine",
1159
+ "name" : "model.tts_language_model.layers.14.mlp.down_proj",
1160
+ "out_dim" : 896,
1161
+ "quant_file" : "model.safetensors",
1162
+ "shape" : [
1163
+ 896,
1164
+ 4864
1165
+ ]
1166
+ },
1167
+ {
1168
+ "bits" : 8,
1169
+ "file" : "model.safetensors",
1170
+ "group_size" : 32,
1171
+ "in_dim" : 896,
1172
+ "mode" : "affine",
1173
+ "name" : "model.tts_language_model.layers.13.self_attn.q_proj",
1174
+ "out_dim" : 896,
1175
+ "quant_file" : "model.safetensors",
1176
+ "shape" : [
1177
+ 896,
1178
+ 896
1179
+ ]
1180
+ },
1181
+ {
1182
+ "bits" : 8,
1183
+ "file" : "model.safetensors",
1184
+ "group_size" : 32,
1185
+ "in_dim" : 896,
1186
+ "mode" : "affine",
1187
+ "name" : "model.language_model.layers.3.mlp.up_proj",
1188
+ "out_dim" : 4864,
1189
+ "quant_file" : "model.safetensors",
1190
+ "shape" : [
1191
+ 4864,
1192
+ 896
1193
+ ]
1194
+ },
1195
+ {
1196
+ "bits" : 8,
1197
+ "file" : "model.safetensors",
1198
+ "group_size" : 32,
1199
+ "in_dim" : 4864,
1200
+ "mode" : "affine",
1201
+ "name" : "model.tts_language_model.layers.2.mlp.down_proj",
1202
+ "out_dim" : 896,
1203
+ "quant_file" : "model.safetensors",
1204
+ "shape" : [
1205
+ 896,
1206
+ 4864
1207
+ ]
1208
+ },
1209
+ {
1210
+ "bits" : 8,
1211
+ "file" : "model.safetensors",
1212
+ "group_size" : 32,
1213
+ "in_dim" : 896,
1214
+ "mode" : "affine",
1215
+ "name" : "model.tts_language_model.layers.8.self_attn.k_proj",
1216
+ "out_dim" : 128,
1217
+ "quant_file" : "model.safetensors",
1218
+ "shape" : [
1219
+ 128,
1220
+ 896
1221
+ ]
1222
+ },
1223
+ {
1224
+ "bits" : 8,
1225
+ "file" : "model.safetensors",
1226
+ "group_size" : 32,
1227
+ "in_dim" : 4864,
1228
+ "mode" : "affine",
1229
+ "name" : "model.tts_language_model.layers.19.mlp.down_proj",
1230
+ "out_dim" : 896,
1231
+ "quant_file" : "model.safetensors",
1232
+ "shape" : [
1233
+ 896,
1234
+ 4864
1235
+ ]
1236
+ },
1237
+ {
1238
+ "bits" : 8,
1239
+ "file" : "model.safetensors",
1240
+ "group_size" : 32,
1241
+ "in_dim" : 896,
1242
+ "mode" : "affine",
1243
+ "name" : "model.tts_language_model.layers.12.self_attn.q_proj",
1244
+ "out_dim" : 896,
1245
+ "quant_file" : "model.safetensors",
1246
+ "shape" : [
1247
+ 896,
1248
+ 896
1249
+ ]
1250
+ },
1251
+ {
1252
+ "bits" : 8,
1253
+ "file" : "model.safetensors",
1254
+ "group_size" : 32,
1255
+ "in_dim" : 256,
1256
+ "mode" : "affine",
1257
+ "name" : "model.acoustic_tokenizer.decoder.stages.3.1.ffn.linear1",
1258
+ "out_dim" : 1024,
1259
+ "quant_file" : "model.safetensors",
1260
+ "shape" : [
1261
+ 1024,
1262
+ 256
1263
+ ]
1264
+ },
1265
+ {
1266
+ "bits" : 8,
1267
+ "file" : "model.safetensors",
1268
+ "group_size" : 32,
1269
+ "in_dim" : 4864,
1270
+ "mode" : "affine",
1271
+ "name" : "model.tts_language_model.layers.9.mlp.down_proj",
1272
+ "out_dim" : 896,
1273
+ "quant_file" : "model.safetensors",
1274
+ "shape" : [
1275
+ 896,
1276
+ 4864
1277
+ ]
1278
+ },
1279
+ {
1280
+ "bits" : 8,
1281
+ "file" : "model.safetensors",
1282
+ "group_size" : 32,
1283
+ "in_dim" : 896,
1284
+ "mode" : "affine",
1285
+ "name" : "model.tts_language_model.layers.17.mlp.gate_proj",
1286
+ "out_dim" : 4864,
1287
+ "quant_file" : "model.safetensors",
1288
+ "shape" : [
1289
+ 4864,
1290
+ 896
1291
+ ]
1292
+ },
1293
+ {
1294
+ "bits" : 8,
1295
+ "file" : "model.safetensors",
1296
+ "group_size" : 32,
1297
+ "in_dim" : 2048,
1298
+ "mode" : "affine",
1299
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.3.ffn.linear1",
1300
+ "out_dim" : 8192,
1301
+ "quant_file" : "model.safetensors",
1302
+ "shape" : [
1303
+ 8192,
1304
+ 2048
1305
+ ]
1306
+ },
1307
+ {
1308
+ "bits" : 8,
1309
+ "file" : "model.safetensors",
1310
+ "group_size" : 32,
1311
+ "in_dim" : 896,
1312
+ "mode" : "affine",
1313
+ "name" : "model.tts_language_model.layers.18.self_attn.o_proj",
1314
+ "out_dim" : 896,
1315
+ "quant_file" : "model.safetensors",
1316
+ "shape" : [
1317
+ 896,
1318
+ 896
1319
+ ]
1320
+ },
1321
+ {
1322
+ "bits" : 8,
1323
+ "file" : "model.safetensors",
1324
+ "group_size" : 32,
1325
+ "in_dim" : 896,
1326
+ "mode" : "affine",
1327
+ "name" : "model.language_model.layers.0.self_attn.q_proj",
1328
+ "out_dim" : 896,
1329
+ "quant_file" : "model.safetensors",
1330
+ "shape" : [
1331
+ 896,
1332
+ 896
1333
+ ]
1334
+ },
1335
+ {
1336
+ "bits" : 8,
1337
+ "file" : "model.safetensors",
1338
+ "group_size" : 32,
1339
+ "in_dim" : 4864,
1340
+ "mode" : "affine",
1341
+ "name" : "model.tts_language_model.layers.3.mlp.down_proj",
1342
+ "out_dim" : 896,
1343
+ "quant_file" : "model.safetensors",
1344
+ "shape" : [
1345
+ 896,
1346
+ 4864
1347
+ ]
1348
+ },
1349
+ {
1350
+ "bits" : 8,
1351
+ "file" : "model.safetensors",
1352
+ "group_size" : 32,
1353
+ "in_dim" : 896,
1354
+ "mode" : "affine",
1355
+ "name" : "model.language_model.layers.0.mlp.up_proj",
1356
+ "out_dim" : 4864,
1357
+ "quant_file" : "model.safetensors",
1358
+ "shape" : [
1359
+ 4864,
1360
+ 896
1361
+ ]
1362
+ },
1363
+ {
1364
+ "bits" : 8,
1365
+ "file" : "model.safetensors",
1366
+ "group_size" : 32,
1367
+ "in_dim" : 4864,
1368
+ "mode" : "affine",
1369
+ "name" : "model.tts_language_model.layers.5.mlp.down_proj",
1370
+ "out_dim" : 896,
1371
+ "quant_file" : "model.safetensors",
1372
+ "shape" : [
1373
+ 896,
1374
+ 4864
1375
+ ]
1376
+ },
1377
+ {
1378
+ "bits" : 8,
1379
+ "file" : "model.safetensors",
1380
+ "group_size" : 32,
1381
+ "in_dim" : 896,
1382
+ "mode" : "affine",
1383
+ "name" : "model.tts_language_model.layers.7.self_attn.k_proj",
1384
+ "out_dim" : 128,
1385
+ "quant_file" : "model.safetensors",
1386
+ "shape" : [
1387
+ 128,
1388
+ 896
1389
+ ]
1390
+ },
1391
+ {
1392
+ "bits" : 8,
1393
+ "file" : "model.safetensors",
1394
+ "group_size" : 32,
1395
+ "in_dim" : 896,
1396
+ "mode" : "affine",
1397
+ "name" : "model.tts_language_model.layers.10.self_attn.o_proj",
1398
+ "out_dim" : 896,
1399
+ "quant_file" : "model.safetensors",
1400
+ "shape" : [
1401
+ 896,
1402
+ 896
1403
+ ]
1404
+ },
1405
+ {
1406
+ "bits" : 8,
1407
+ "file" : "model.safetensors",
1408
+ "group_size" : 32,
1409
+ "in_dim" : 896,
1410
+ "mode" : "affine",
1411
+ "name" : "model.tts_language_model.layers.8.self_attn.o_proj",
1412
+ "out_dim" : 896,
1413
+ "quant_file" : "model.safetensors",
1414
+ "shape" : [
1415
+ 896,
1416
+ 896
1417
+ ]
1418
+ },
1419
+ {
1420
+ "bits" : 8,
1421
+ "file" : "model.safetensors",
1422
+ "group_size" : 32,
1423
+ "in_dim" : 896,
1424
+ "mode" : "affine",
1425
+ "name" : "model.prediction_head.layers.2.ffn.gate_proj",
1426
+ "out_dim" : 2688,
1427
+ "quant_file" : "model.safetensors",
1428
+ "shape" : [
1429
+ 2688,
1430
+ 896
1431
+ ]
1432
+ },
1433
+ {
1434
+ "bits" : 8,
1435
+ "file" : "model.safetensors",
1436
+ "group_size" : 32,
1437
+ "in_dim" : 896,
1438
+ "mode" : "affine",
1439
+ "name" : "model.tts_language_model.layers.4.self_attn.k_proj",
1440
+ "out_dim" : 128,
1441
+ "quant_file" : "model.safetensors",
1442
+ "shape" : [
1443
+ 128,
1444
+ 896
1445
+ ]
1446
+ },
1447
+ {
1448
+ "bits" : 8,
1449
+ "file" : "model.safetensors",
1450
+ "group_size" : 32,
1451
+ "in_dim" : 128,
1452
+ "mode" : "affine",
1453
+ "name" : "model.acoustic_tokenizer.decoder.stages.4.0.ffn.linear1",
1454
+ "out_dim" : 512,
1455
+ "quant_file" : "model.safetensors",
1456
+ "shape" : [
1457
+ 512,
1458
+ 128
1459
+ ]
1460
+ },
1461
+ {
1462
+ "bits" : 8,
1463
+ "file" : "model.safetensors",
1464
+ "group_size" : 32,
1465
+ "in_dim" : 2048,
1466
+ "mode" : "affine",
1467
+ "name" : "model.acoustic_tokenizer.decoder.stages.2.1.ffn.linear2",
1468
+ "out_dim" : 512,
1469
+ "quant_file" : "model.safetensors",
1470
+ "shape" : [
1471
+ 512,
1472
+ 2048
1473
+ ]
1474
+ },
1475
+ {
1476
+ "bits" : 8,
1477
+ "file" : "model.safetensors",
1478
+ "group_size" : 32,
1479
+ "in_dim" : 4864,
1480
+ "mode" : "affine",
1481
+ "name" : "model.tts_language_model.layers.11.mlp.down_proj",
1482
+ "out_dim" : 896,
1483
+ "quant_file" : "model.safetensors",
1484
+ "shape" : [
1485
+ 896,
1486
+ 4864
1487
+ ]
1488
+ },
1489
+ {
1490
+ "bits" : 8,
1491
+ "file" : "model.safetensors",
1492
+ "group_size" : 32,
1493
+ "in_dim" : 896,
1494
+ "mode" : "affine",
1495
+ "name" : "model.language_model.layers.0.mlp.gate_proj",
1496
+ "out_dim" : 4864,
1497
+ "quant_file" : "model.safetensors",
1498
+ "shape" : [
1499
+ 4864,
1500
+ 896
1501
+ ]
1502
+ },
1503
+ {
1504
+ "bits" : 8,
1505
+ "file" : "model.safetensors",
1506
+ "group_size" : 32,
1507
+ "in_dim" : 4864,
1508
+ "mode" : "affine",
1509
+ "name" : "model.tts_language_model.layers.6.mlp.down_proj",
1510
+ "out_dim" : 896,
1511
+ "quant_file" : "model.safetensors",
1512
+ "shape" : [
1513
+ 896,
1514
+ 4864
1515
+ ]
1516
+ },
1517
+ {
1518
+ "bits" : 8,
1519
+ "file" : "model.safetensors",
1520
+ "group_size" : 32,
1521
+ "in_dim" : 8192,
1522
+ "mode" : "affine",
1523
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.0.ffn.linear2",
1524
+ "out_dim" : 2048,
1525
+ "quant_file" : "model.safetensors",
1526
+ "shape" : [
1527
+ 2048,
1528
+ 8192
1529
+ ]
1530
+ },
1531
+ {
1532
+ "bits" : 8,
1533
+ "file" : "model.safetensors",
1534
+ "group_size" : 32,
1535
+ "in_dim" : 4864,
1536
+ "mode" : "affine",
1537
+ "name" : "model.tts_language_model.layers.1.mlp.down_proj",
1538
+ "out_dim" : 896,
1539
+ "quant_file" : "model.safetensors",
1540
+ "shape" : [
1541
+ 896,
1542
+ 4864
1543
+ ]
1544
+ },
1545
+ {
1546
+ "bits" : 8,
1547
+ "file" : "model.safetensors",
1548
+ "group_size" : 32,
1549
+ "in_dim" : 896,
1550
+ "mode" : "affine",
1551
+ "name" : "model.tts_language_model.layers.18.mlp.gate_proj",
1552
+ "out_dim" : 4864,
1553
+ "quant_file" : "model.safetensors",
1554
+ "shape" : [
1555
+ 4864,
1556
+ 896
1557
+ ]
1558
+ },
1559
+ {
1560
+ "bits" : 8,
1561
+ "file" : "model.safetensors",
1562
+ "group_size" : 32,
1563
+ "in_dim" : 2048,
1564
+ "mode" : "affine",
1565
+ "name" : "model.acoustic_tokenizer.decoder.stages.2.0.ffn.linear2",
1566
+ "out_dim" : 512,
1567
+ "quant_file" : "model.safetensors",
1568
+ "shape" : [
1569
+ 512,
1570
+ 2048
1571
+ ]
1572
+ },
1573
+ {
1574
+ "bits" : 8,
1575
+ "file" : "model.safetensors",
1576
+ "group_size" : 32,
1577
+ "in_dim" : 896,
1578
+ "mode" : "affine",
1579
+ "name" : "model.prediction_head.layers.1.ffn.up_proj",
1580
+ "out_dim" : 2688,
1581
+ "quant_file" : "model.safetensors",
1582
+ "shape" : [
1583
+ 2688,
1584
+ 896
1585
+ ]
1586
+ },
1587
+ {
1588
+ "bits" : 8,
1589
+ "file" : "model.safetensors",
1590
+ "group_size" : 32,
1591
+ "in_dim" : 896,
1592
+ "mode" : "affine",
1593
+ "name" : "model.tts_language_model.layers.1.self_attn.v_proj",
1594
+ "out_dim" : 128,
1595
+ "quant_file" : "model.safetensors",
1596
+ "shape" : [
1597
+ 128,
1598
+ 896
1599
+ ]
1600
+ },
1601
+ {
1602
+ "bits" : 8,
1603
+ "file" : "model.safetensors",
1604
+ "group_size" : 32,
1605
+ "in_dim" : 896,
1606
+ "mode" : "affine",
1607
+ "name" : "model.tts_language_model.layers.0.mlp.up_proj",
1608
+ "out_dim" : 4864,
1609
+ "quant_file" : "model.safetensors",
1610
+ "shape" : [
1611
+ 4864,
1612
+ 896
1613
+ ]
1614
+ },
1615
+ {
1616
+ "bits" : 8,
1617
+ "file" : "model.safetensors",
1618
+ "group_size" : 32,
1619
+ "in_dim" : 896,
1620
+ "mode" : "affine",
1621
+ "name" : "model.language_model.layers.3.self_attn.o_proj",
1622
+ "out_dim" : 896,
1623
+ "quant_file" : "model.safetensors",
1624
+ "shape" : [
1625
+ 896,
1626
+ 896
1627
+ ]
1628
+ },
1629
+ {
1630
+ "bits" : 8,
1631
+ "file" : "model.safetensors",
1632
+ "group_size" : 32,
1633
+ "in_dim" : 896,
1634
+ "mode" : "affine",
1635
+ "name" : "model.tts_language_model.layers.7.self_attn.v_proj",
1636
+ "out_dim" : 128,
1637
+ "quant_file" : "model.safetensors",
1638
+ "shape" : [
1639
+ 128,
1640
+ 896
1641
+ ]
1642
+ },
1643
+ {
1644
+ "bits" : 8,
1645
+ "file" : "model.safetensors",
1646
+ "group_size" : 32,
1647
+ "in_dim" : 896,
1648
+ "mode" : "affine",
1649
+ "name" : "model.tts_language_model.layers.12.mlp.up_proj",
1650
+ "out_dim" : 4864,
1651
+ "quant_file" : "model.safetensors",
1652
+ "shape" : [
1653
+ 4864,
1654
+ 896
1655
+ ]
1656
+ },
1657
+ {
1658
+ "bits" : 8,
1659
+ "file" : "model.safetensors",
1660
+ "group_size" : 32,
1661
+ "in_dim" : 896,
1662
+ "mode" : "affine",
1663
+ "name" : "model.tts_language_model.layers.19.self_attn.k_proj",
1664
+ "out_dim" : 128,
1665
+ "quant_file" : "model.safetensors",
1666
+ "shape" : [
1667
+ 128,
1668
+ 896
1669
+ ]
1670
+ },
1671
+ {
1672
+ "bits" : 8,
1673
+ "file" : "model.safetensors",
1674
+ "group_size" : 32,
1675
+ "in_dim" : 2048,
1676
+ "mode" : "affine",
1677
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.1.ffn.linear1",
1678
+ "out_dim" : 8192,
1679
+ "quant_file" : "model.safetensors",
1680
+ "shape" : [
1681
+ 8192,
1682
+ 2048
1683
+ ]
1684
+ },
1685
+ {
1686
+ "bits" : 8,
1687
+ "file" : "model.safetensors",
1688
+ "group_size" : 32,
1689
+ "in_dim" : 4096,
1690
+ "mode" : "affine",
1691
+ "name" : "model.acoustic_tokenizer.decoder.stages.1.1.ffn.linear2",
1692
+ "out_dim" : 1024,
1693
+ "quant_file" : "model.safetensors",
1694
+ "shape" : [
1695
+ 1024,
1696
+ 4096
1697
+ ]
1698
+ },
1699
+ {
1700
+ "bits" : 8,
1701
+ "file" : "model.safetensors",
1702
+ "group_size" : 32,
1703
+ "in_dim" : 64,
1704
+ "mode" : "affine",
1705
+ "name" : "model.acoustic_tokenizer.decoder.stages.5.0.ffn.linear1",
1706
+ "out_dim" : 256,
1707
+ "quant_file" : "model.safetensors",
1708
+ "shape" : [
1709
+ 256,
1710
+ 64
1711
+ ]
1712
+ },
1713
+ {
1714
+ "bits" : 8,
1715
+ "file" : "model.safetensors",
1716
+ "group_size" : 32,
1717
+ "in_dim" : 896,
1718
+ "mode" : "affine",
1719
+ "name" : "model.tts_language_model.layers.13.mlp.up_proj",
1720
+ "out_dim" : 4864,
1721
+ "quant_file" : "model.safetensors",
1722
+ "shape" : [
1723
+ 4864,
1724
+ 896
1725
+ ]
1726
+ },
1727
+ {
1728
+ "bits" : 8,
1729
+ "file" : "model.safetensors",
1730
+ "group_size" : 32,
1731
+ "in_dim" : 2688,
1732
+ "mode" : "affine",
1733
+ "name" : "model.prediction_head.layers.2.ffn.down_proj",
1734
+ "out_dim" : 896,
1735
+ "quant_file" : "model.safetensors",
1736
+ "shape" : [
1737
+ 896,
1738
+ 2688
1739
+ ]
1740
+ },
1741
+ {
1742
+ "bits" : 8,
1743
+ "file" : "model.safetensors",
1744
+ "group_size" : 32,
1745
+ "in_dim" : 896,
1746
+ "mode" : "affine",
1747
+ "name" : "model.tts_language_model.layers.10.mlp.gate_proj",
1748
+ "out_dim" : 4864,
1749
+ "quant_file" : "model.safetensors",
1750
+ "shape" : [
1751
+ 4864,
1752
+ 896
1753
+ ]
1754
+ },
1755
+ {
1756
+ "bits" : 8,
1757
+ "file" : "model.safetensors",
1758
+ "group_size" : 32,
1759
+ "in_dim" : 896,
1760
+ "mode" : "affine",
1761
+ "name" : "model.tts_language_model.layers.4.self_attn.o_proj",
1762
+ "out_dim" : 896,
1763
+ "quant_file" : "model.safetensors",
1764
+ "shape" : [
1765
+ 896,
1766
+ 896
1767
+ ]
1768
+ },
1769
+ {
1770
+ "bits" : 8,
1771
+ "file" : "model.safetensors",
1772
+ "group_size" : 32,
1773
+ "in_dim" : 896,
1774
+ "mode" : "affine",
1775
+ "name" : "model.tts_language_model.layers.12.mlp.gate_proj",
1776
+ "out_dim" : 4864,
1777
+ "quant_file" : "model.safetensors",
1778
+ "shape" : [
1779
+ 4864,
1780
+ 896
1781
+ ]
1782
+ },
1783
+ {
1784
+ "bits" : 8,
1785
+ "file" : "model.safetensors",
1786
+ "group_size" : 32,
1787
+ "in_dim" : 8192,
1788
+ "mode" : "affine",
1789
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.7.ffn.linear2",
1790
+ "out_dim" : 2048,
1791
+ "quant_file" : "model.safetensors",
1792
+ "shape" : [
1793
+ 2048,
1794
+ 8192
1795
+ ]
1796
+ },
1797
+ {
1798
+ "bits" : 8,
1799
+ "file" : "model.safetensors",
1800
+ "group_size" : 32,
1801
+ "in_dim" : 896,
1802
+ "mode" : "affine",
1803
+ "name" : "model.tts_language_model.layers.9.self_attn.q_proj",
1804
+ "out_dim" : 896,
1805
+ "quant_file" : "model.safetensors",
1806
+ "shape" : [
1807
+ 896,
1808
+ 896
1809
+ ]
1810
+ },
1811
+ {
1812
+ "bits" : 8,
1813
+ "file" : "model.safetensors",
1814
+ "group_size" : 32,
1815
+ "in_dim" : 896,
1816
+ "mode" : "affine",
1817
+ "name" : "model.tts_language_model.layers.15.self_attn.k_proj",
1818
+ "out_dim" : 128,
1819
+ "quant_file" : "model.safetensors",
1820
+ "shape" : [
1821
+ 128,
1822
+ 896
1823
+ ]
1824
+ },
1825
+ {
1826
+ "bits" : 8,
1827
+ "file" : "model.safetensors",
1828
+ "group_size" : 32,
1829
+ "in_dim" : 896,
1830
+ "mode" : "affine",
1831
+ "name" : "model.tts_language_model.layers.15.self_attn.o_proj",
1832
+ "out_dim" : 896,
1833
+ "quant_file" : "model.safetensors",
1834
+ "shape" : [
1835
+ 896,
1836
+ 896
1837
+ ]
1838
+ },
1839
+ {
1840
+ "bits" : 8,
1841
+ "file" : "model.safetensors",
1842
+ "group_size" : 32,
1843
+ "in_dim" : 896,
1844
+ "mode" : "affine",
1845
+ "name" : "model.tts_language_model.layers.16.self_attn.k_proj",
1846
+ "out_dim" : 128,
1847
+ "quant_file" : "model.safetensors",
1848
+ "shape" : [
1849
+ 128,
1850
+ 896
1851
+ ]
1852
+ },
1853
+ {
1854
+ "bits" : 8,
1855
+ "file" : "model.safetensors",
1856
+ "group_size" : 32,
1857
+ "in_dim" : 896,
1858
+ "mode" : "affine",
1859
+ "name" : "model.tts_language_model.layers.5.self_attn.q_proj",
1860
+ "out_dim" : 896,
1861
+ "quant_file" : "model.safetensors",
1862
+ "shape" : [
1863
+ 896,
1864
+ 896
1865
+ ]
1866
+ },
1867
+ {
1868
+ "bits" : 8,
1869
+ "file" : "model.safetensors",
1870
+ "group_size" : 32,
1871
+ "in_dim" : 896,
1872
+ "mode" : "affine",
1873
+ "name" : "model.tts_language_model.layers.4.self_attn.q_proj",
1874
+ "out_dim" : 896,
1875
+ "quant_file" : "model.safetensors",
1876
+ "shape" : [
1877
+ 896,
1878
+ 896
1879
+ ]
1880
+ },
1881
+ {
1882
+ "bits" : 8,
1883
+ "file" : "model.safetensors",
1884
+ "group_size" : 32,
1885
+ "in_dim" : 4864,
1886
+ "mode" : "affine",
1887
+ "name" : "model.tts_language_model.layers.15.mlp.down_proj",
1888
+ "out_dim" : 896,
1889
+ "quant_file" : "model.safetensors",
1890
+ "shape" : [
1891
+ 896,
1892
+ 4864
1893
+ ]
1894
+ },
1895
+ {
1896
+ "bits" : 8,
1897
+ "file" : "model.safetensors",
1898
+ "group_size" : 32,
1899
+ "in_dim" : 896,
1900
+ "mode" : "affine",
1901
+ "name" : "model.tts_language_model.layers.16.mlp.up_proj",
1902
+ "out_dim" : 4864,
1903
+ "quant_file" : "model.safetensors",
1904
+ "shape" : [
1905
+ 4864,
1906
+ 896
1907
+ ]
1908
+ },
1909
+ {
1910
+ "bits" : 8,
1911
+ "file" : "model.safetensors",
1912
+ "group_size" : 32,
1913
+ "in_dim" : 896,
1914
+ "mode" : "affine",
1915
+ "name" : "model.tts_language_model.layers.9.self_attn.v_proj",
1916
+ "out_dim" : 128,
1917
+ "quant_file" : "model.safetensors",
1918
+ "shape" : [
1919
+ 128,
1920
+ 896
1921
+ ]
1922
+ },
1923
+ {
1924
+ "bits" : 8,
1925
+ "file" : "model.safetensors",
1926
+ "group_size" : 32,
1927
+ "in_dim" : 32,
1928
+ "mode" : "affine",
1929
+ "name" : "model.acoustic_tokenizer.decoder.stages.6.2.ffn.linear1",
1930
+ "out_dim" : 128,
1931
+ "quant_file" : "model.safetensors",
1932
+ "shape" : [
1933
+ 128,
1934
+ 32
1935
+ ]
1936
+ },
1937
+ {
1938
+ "bits" : 8,
1939
+ "file" : "model.safetensors",
1940
+ "group_size" : 32,
1941
+ "in_dim" : 896,
1942
+ "mode" : "affine",
1943
+ "name" : "model.tts_language_model.layers.15.self_attn.v_proj",
1944
+ "out_dim" : 128,
1945
+ "quant_file" : "model.safetensors",
1946
+ "shape" : [
1947
+ 128,
1948
+ 896
1949
+ ]
1950
+ },
1951
+ {
1952
+ "bits" : 8,
1953
+ "file" : "model.safetensors",
1954
+ "group_size" : 32,
1955
+ "in_dim" : 128,
1956
+ "mode" : "affine",
1957
+ "name" : "model.acoustic_tokenizer.decoder.stages.4.2.ffn.linear1",
1958
+ "out_dim" : 512,
1959
+ "quant_file" : "model.safetensors",
1960
+ "shape" : [
1961
+ 512,
1962
+ 128
1963
+ ]
1964
+ },
1965
+ {
1966
+ "bits" : 8,
1967
+ "file" : "model.safetensors",
1968
+ "group_size" : 32,
1969
+ "in_dim" : 896,
1970
+ "mode" : "affine",
1971
+ "name" : "model.tts_language_model.layers.5.mlp.up_proj",
1972
+ "out_dim" : 4864,
1973
+ "quant_file" : "model.safetensors",
1974
+ "shape" : [
1975
+ 4864,
1976
+ 896
1977
+ ]
1978
+ },
1979
+ {
1980
+ "bits" : 8,
1981
+ "file" : "model.safetensors",
1982
+ "group_size" : 32,
1983
+ "in_dim" : 896,
1984
+ "mode" : "affine",
1985
+ "name" : "model.tts_language_model.layers.4.self_attn.v_proj",
1986
+ "out_dim" : 128,
1987
+ "quant_file" : "model.safetensors",
1988
+ "shape" : [
1989
+ 128,
1990
+ 896
1991
+ ]
1992
+ },
1993
+ {
1994
+ "bits" : 8,
1995
+ "file" : "model.safetensors",
1996
+ "group_size" : 32,
1997
+ "in_dim" : 896,
1998
+ "mode" : "affine",
1999
+ "name" : "model.acoustic_connector.fc2",
2000
+ "out_dim" : 896,
2001
+ "quant_file" : "model.safetensors",
2002
+ "shape" : [
2003
+ 896,
2004
+ 896
2005
+ ]
2006
+ },
2007
+ {
2008
+ "bits" : 8,
2009
+ "file" : "model.safetensors",
2010
+ "group_size" : 32,
2011
+ "in_dim" : 896,
2012
+ "mode" : "affine",
2013
+ "name" : "model.tts_language_model.layers.8.mlp.up_proj",
2014
+ "out_dim" : 4864,
2015
+ "quant_file" : "model.safetensors",
2016
+ "shape" : [
2017
+ 4864,
2018
+ 896
2019
+ ]
2020
+ },
2021
+ {
2022
+ "bits" : 8,
2023
+ "file" : "model.safetensors",
2024
+ "group_size" : 32,
2025
+ "in_dim" : 896,
2026
+ "mode" : "affine",
2027
+ "name" : "model.language_model.layers.2.mlp.up_proj",
2028
+ "out_dim" : 4864,
2029
+ "quant_file" : "model.safetensors",
2030
+ "shape" : [
2031
+ 4864,
2032
+ 896
2033
+ ]
2034
+ },
2035
+ {
2036
+ "bits" : 8,
2037
+ "file" : "model.safetensors",
2038
+ "group_size" : 32,
2039
+ "in_dim" : 4864,
2040
+ "mode" : "affine",
2041
+ "name" : "model.tts_language_model.layers.17.mlp.down_proj",
2042
+ "out_dim" : 896,
2043
+ "quant_file" : "model.safetensors",
2044
+ "shape" : [
2045
+ 896,
2046
+ 4864
2047
+ ]
2048
+ },
2049
+ {
2050
+ "bits" : 8,
2051
+ "file" : "model.safetensors",
2052
+ "group_size" : 32,
2053
+ "in_dim" : 896,
2054
+ "mode" : "affine",
2055
+ "name" : "model.language_model.layers.2.self_attn.q_proj",
2056
+ "out_dim" : 896,
2057
+ "quant_file" : "model.safetensors",
2058
+ "shape" : [
2059
+ 896,
2060
+ 896
2061
+ ]
2062
+ },
2063
+ {
2064
+ "bits" : 8,
2065
+ "file" : "model.safetensors",
2066
+ "group_size" : 32,
2067
+ "in_dim" : 896,
2068
+ "mode" : "affine",
2069
+ "name" : "model.tts_language_model.layers.13.mlp.gate_proj",
2070
+ "out_dim" : 4864,
2071
+ "quant_file" : "model.safetensors",
2072
+ "shape" : [
2073
+ 4864,
2074
+ 896
2075
+ ]
2076
+ },
2077
+ {
2078
+ "bits" : 8,
2079
+ "file" : "model.safetensors",
2080
+ "group_size" : 32,
2081
+ "in_dim" : 896,
2082
+ "mode" : "affine",
2083
+ "name" : "model.tts_language_model.layers.1.self_attn.o_proj",
2084
+ "out_dim" : 896,
2085
+ "quant_file" : "model.safetensors",
2086
+ "shape" : [
2087
+ 896,
2088
+ 896
2089
+ ]
2090
+ },
2091
+ {
2092
+ "bits" : 8,
2093
+ "file" : "model.safetensors",
2094
+ "group_size" : 32,
2095
+ "in_dim" : 896,
2096
+ "mode" : "affine",
2097
+ "name" : "model.tts_language_model.layers.7.self_attn.q_proj",
2098
+ "out_dim" : 896,
2099
+ "quant_file" : "model.safetensors",
2100
+ "shape" : [
2101
+ 896,
2102
+ 896
2103
+ ]
2104
+ },
2105
+ {
2106
+ "bits" : 8,
2107
+ "file" : "model.safetensors",
2108
+ "group_size" : 32,
2109
+ "in_dim" : 896,
2110
+ "mode" : "affine",
2111
+ "name" : "model.tts_language_model.layers.19.mlp.up_proj",
2112
+ "out_dim" : 4864,
2113
+ "quant_file" : "model.safetensors",
2114
+ "shape" : [
2115
+ 4864,
2116
+ 896
2117
+ ]
2118
+ },
2119
+ {
2120
+ "bits" : 8,
2121
+ "file" : "model.safetensors",
2122
+ "group_size" : 32,
2123
+ "in_dim" : 2688,
2124
+ "mode" : "affine",
2125
+ "name" : "model.prediction_head.layers.3.ffn.down_proj",
2126
+ "out_dim" : 896,
2127
+ "quant_file" : "model.safetensors",
2128
+ "shape" : [
2129
+ 896,
2130
+ 2688
2131
+ ]
2132
+ },
2133
+ {
2134
+ "bits" : 8,
2135
+ "file" : "model.safetensors",
2136
+ "group_size" : 32,
2137
+ "in_dim" : 896,
2138
+ "mode" : "affine",
2139
+ "name" : "model.tts_language_model.layers.14.self_attn.v_proj",
2140
+ "out_dim" : 128,
2141
+ "quant_file" : "model.safetensors",
2142
+ "shape" : [
2143
+ 128,
2144
+ 896
2145
+ ]
2146
+ },
2147
+ {
2148
+ "bits" : 8,
2149
+ "file" : "model.safetensors",
2150
+ "group_size" : 32,
2151
+ "in_dim" : 4864,
2152
+ "mode" : "affine",
2153
+ "name" : "model.language_model.layers.2.mlp.down_proj",
2154
+ "out_dim" : 896,
2155
+ "quant_file" : "model.safetensors",
2156
+ "shape" : [
2157
+ 896,
2158
+ 4864
2159
+ ]
2160
+ },
2161
+ {
2162
+ "bits" : 8,
2163
+ "file" : "model.safetensors",
2164
+ "group_size" : 32,
2165
+ "in_dim" : 896,
2166
+ "mode" : "affine",
2167
+ "name" : "model.language_model.layers.1.mlp.gate_proj",
2168
+ "out_dim" : 4864,
2169
+ "quant_file" : "model.safetensors",
2170
+ "shape" : [
2171
+ 4864,
2172
+ 896
2173
+ ]
2174
+ },
2175
+ {
2176
+ "bits" : 8,
2177
+ "file" : "model.safetensors",
2178
+ "group_size" : 32,
2179
+ "in_dim" : 128,
2180
+ "mode" : "affine",
2181
+ "name" : "model.acoustic_tokenizer.decoder.stages.6.2.ffn.linear2",
2182
+ "out_dim" : 32,
2183
+ "quant_file" : "model.safetensors",
2184
+ "shape" : [
2185
+ 32,
2186
+ 128
2187
+ ]
2188
+ },
2189
+ {
2190
+ "bits" : 8,
2191
+ "file" : "model.safetensors",
2192
+ "group_size" : 32,
2193
+ "in_dim" : 896,
2194
+ "mode" : "affine",
2195
+ "name" : "model.prediction_head.layers.3.ffn.up_proj",
2196
+ "out_dim" : 2688,
2197
+ "quant_file" : "model.safetensors",
2198
+ "shape" : [
2199
+ 2688,
2200
+ 896
2201
+ ]
2202
+ },
2203
+ {
2204
+ "bits" : 8,
2205
+ "file" : "model.safetensors",
2206
+ "group_size" : 32,
2207
+ "in_dim" : 896,
2208
+ "mode" : "affine",
2209
+ "name" : "model.tts_language_model.layers.17.self_attn.o_proj",
2210
+ "out_dim" : 896,
2211
+ "quant_file" : "model.safetensors",
2212
+ "shape" : [
2213
+ 896,
2214
+ 896
2215
+ ]
2216
+ },
2217
+ {
2218
+ "bits" : 8,
2219
+ "file" : "model.safetensors",
2220
+ "group_size" : 32,
2221
+ "in_dim" : 896,
2222
+ "mode" : "affine",
2223
+ "name" : "model.prediction_head.layers.3.ffn.gate_proj",
2224
+ "out_dim" : 2688,
2225
+ "quant_file" : "model.safetensors",
2226
+ "shape" : [
2227
+ 2688,
2228
+ 896
2229
+ ]
2230
+ },
2231
+ {
2232
+ "bits" : 8,
2233
+ "file" : "model.safetensors",
2234
+ "group_size" : 32,
2235
+ "in_dim" : 896,
2236
+ "mode" : "affine",
2237
+ "name" : "model.tts_language_model.layers.11.mlp.gate_proj",
2238
+ "out_dim" : 4864,
2239
+ "quant_file" : "model.safetensors",
2240
+ "shape" : [
2241
+ 4864,
2242
+ 896
2243
+ ]
2244
+ },
2245
+ {
2246
+ "bits" : 8,
2247
+ "file" : "model.safetensors",
2248
+ "group_size" : 32,
2249
+ "in_dim" : 896,
2250
+ "mode" : "affine",
2251
+ "name" : "model.tts_language_model.layers.11.self_attn.v_proj",
2252
+ "out_dim" : 128,
2253
+ "quant_file" : "model.safetensors",
2254
+ "shape" : [
2255
+ 128,
2256
+ 896
2257
+ ]
2258
+ },
2259
+ {
2260
+ "bits" : 8,
2261
+ "file" : "model.safetensors",
2262
+ "group_size" : 32,
2263
+ "in_dim" : 896,
2264
+ "mode" : "affine",
2265
+ "name" : "model.language_model.layers.1.self_attn.q_proj",
2266
+ "out_dim" : 896,
2267
+ "quant_file" : "model.safetensors",
2268
+ "shape" : [
2269
+ 896,
2270
+ 896
2271
+ ]
2272
+ },
2273
+ {
2274
+ "bits" : 8,
2275
+ "file" : "model.safetensors",
2276
+ "group_size" : 32,
2277
+ "in_dim" : 896,
2278
+ "mode" : "affine",
2279
+ "name" : "model.tts_language_model.layers.11.self_attn.q_proj",
2280
+ "out_dim" : 896,
2281
+ "quant_file" : "model.safetensors",
2282
+ "shape" : [
2283
+ 896,
2284
+ 896
2285
+ ]
2286
+ },
2287
+ {
2288
+ "bits" : 8,
2289
+ "file" : "model.safetensors",
2290
+ "group_size" : 32,
2291
+ "in_dim" : 4864,
2292
+ "mode" : "affine",
2293
+ "name" : "model.tts_language_model.layers.10.mlp.down_proj",
2294
+ "out_dim" : 896,
2295
+ "quant_file" : "model.safetensors",
2296
+ "shape" : [
2297
+ 896,
2298
+ 4864
2299
+ ]
2300
+ },
2301
+ {
2302
+ "bits" : 8,
2303
+ "file" : "model.safetensors",
2304
+ "group_size" : 32,
2305
+ "in_dim" : 4096,
2306
+ "mode" : "affine",
2307
+ "name" : "model.acoustic_tokenizer.decoder.stages.1.0.ffn.linear2",
2308
+ "out_dim" : 1024,
2309
+ "quant_file" : "model.safetensors",
2310
+ "shape" : [
2311
+ 1024,
2312
+ 4096
2313
+ ]
2314
+ },
2315
+ {
2316
+ "bits" : 8,
2317
+ "file" : "model.safetensors",
2318
+ "group_size" : 32,
2319
+ "in_dim" : 896,
2320
+ "mode" : "affine",
2321
+ "name" : "model.tts_language_model.layers.9.self_attn.o_proj",
2322
+ "out_dim" : 896,
2323
+ "quant_file" : "model.safetensors",
2324
+ "shape" : [
2325
+ 896,
2326
+ 896
2327
+ ]
2328
+ },
2329
+ {
2330
+ "bits" : 8,
2331
+ "file" : "model.safetensors",
2332
+ "group_size" : 32,
2333
+ "in_dim" : 896,
2334
+ "mode" : "affine",
2335
+ "name" : "model.tts_language_model.layers.14.mlp.up_proj",
2336
+ "out_dim" : 4864,
2337
+ "quant_file" : "model.safetensors",
2338
+ "shape" : [
2339
+ 4864,
2340
+ 896
2341
+ ]
2342
+ },
2343
+ {
2344
+ "bits" : 8,
2345
+ "file" : "model.safetensors",
2346
+ "group_size" : 32,
2347
+ "in_dim" : 2048,
2348
+ "mode" : "affine",
2349
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.4.ffn.linear1",
2350
+ "out_dim" : 8192,
2351
+ "quant_file" : "model.safetensors",
2352
+ "shape" : [
2353
+ 8192,
2354
+ 2048
2355
+ ]
2356
+ },
2357
+ {
2358
+ "bits" : 8,
2359
+ "file" : "model.safetensors",
2360
+ "group_size" : 32,
2361
+ "in_dim" : 896,
2362
+ "mode" : "affine",
2363
+ "name" : "model.tts_language_model.layers.5.self_attn.v_proj",
2364
+ "out_dim" : 128,
2365
+ "quant_file" : "model.safetensors",
2366
+ "shape" : [
2367
+ 128,
2368
+ 896
2369
+ ]
2370
+ },
2371
+ {
2372
+ "bits" : 8,
2373
+ "file" : "model.safetensors",
2374
+ "group_size" : 32,
2375
+ "in_dim" : 896,
2376
+ "mode" : "affine",
2377
+ "name" : "model.tts_language_model.layers.5.mlp.gate_proj",
2378
+ "out_dim" : 4864,
2379
+ "quant_file" : "model.safetensors",
2380
+ "shape" : [
2381
+ 4864,
2382
+ 896
2383
+ ]
2384
+ },
2385
+ {
2386
+ "bits" : 8,
2387
+ "file" : "model.safetensors",
2388
+ "group_size" : 32,
2389
+ "in_dim" : 4864,
2390
+ "mode" : "affine",
2391
+ "name" : "model.tts_language_model.layers.0.mlp.down_proj",
2392
+ "out_dim" : 896,
2393
+ "quant_file" : "model.safetensors",
2394
+ "shape" : [
2395
+ 896,
2396
+ 4864
2397
+ ]
2398
+ },
2399
+ {
2400
+ "bits" : 8,
2401
+ "file" : "model.safetensors",
2402
+ "group_size" : 32,
2403
+ "in_dim" : 128,
2404
+ "mode" : "affine",
2405
+ "name" : "model.acoustic_tokenizer.decoder.stages.4.1.ffn.linear1",
2406
+ "out_dim" : 512,
2407
+ "quant_file" : "model.safetensors",
2408
+ "shape" : [
2409
+ 512,
2410
+ 128
2411
+ ]
2412
+ },
2413
+ {
2414
+ "bits" : 8,
2415
+ "file" : "model.safetensors",
2416
+ "group_size" : 32,
2417
+ "in_dim" : 128,
2418
+ "mode" : "affine",
2419
+ "name" : "model.acoustic_tokenizer.decoder.stages.6.1.ffn.linear2",
2420
+ "out_dim" : 32,
2421
+ "quant_file" : "model.safetensors",
2422
+ "shape" : [
2423
+ 32,
2424
+ 128
2425
+ ]
2426
+ },
2427
+ {
2428
+ "bits" : 8,
2429
+ "file" : "model.safetensors",
2430
+ "group_size" : 32,
2431
+ "in_dim" : 896,
2432
+ "mode" : "affine",
2433
+ "name" : "model.tts_language_model.layers.2.self_attn.o_proj",
2434
+ "out_dim" : 896,
2435
+ "quant_file" : "model.safetensors",
2436
+ "shape" : [
2437
+ 896,
2438
+ 896
2439
+ ]
2440
+ },
2441
+ {
2442
+ "bits" : 8,
2443
+ "file" : "model.safetensors",
2444
+ "group_size" : 32,
2445
+ "in_dim" : 896,
2446
+ "mode" : "affine",
2447
+ "name" : "model.tts_language_model.layers.3.self_attn.v_proj",
2448
+ "out_dim" : 128,
2449
+ "quant_file" : "model.safetensors",
2450
+ "shape" : [
2451
+ 128,
2452
+ 896
2453
+ ]
2454
+ },
2455
+ {
2456
+ "bits" : 8,
2457
+ "file" : "model.safetensors",
2458
+ "group_size" : 32,
2459
+ "in_dim" : 256,
2460
+ "mode" : "affine",
2461
+ "name" : "model.acoustic_tokenizer.decoder.stages.3.2.ffn.linear1",
2462
+ "out_dim" : 1024,
2463
+ "quant_file" : "model.safetensors",
2464
+ "shape" : [
2465
+ 1024,
2466
+ 256
2467
+ ]
2468
+ },
2469
+ {
2470
+ "bits" : 8,
2471
+ "file" : "model.safetensors",
2472
+ "group_size" : 32,
2473
+ "in_dim" : 896,
2474
+ "mode" : "affine",
2475
+ "name" : "model.tts_language_model.layers.2.self_attn.q_proj",
2476
+ "out_dim" : 896,
2477
+ "quant_file" : "model.safetensors",
2478
+ "shape" : [
2479
+ 896,
2480
+ 896
2481
+ ]
2482
+ },
2483
+ {
2484
+ "bits" : 8,
2485
+ "file" : "model.safetensors",
2486
+ "group_size" : 32,
2487
+ "in_dim" : 1024,
2488
+ "mode" : "affine",
2489
+ "name" : "model.acoustic_tokenizer.decoder.stages.3.2.ffn.linear2",
2490
+ "out_dim" : 256,
2491
+ "quant_file" : "model.safetensors",
2492
+ "shape" : [
2493
+ 256,
2494
+ 1024
2495
+ ]
2496
+ },
2497
+ {
2498
+ "bits" : 8,
2499
+ "file" : "model.safetensors",
2500
+ "group_size" : 32,
2501
+ "in_dim" : 896,
2502
+ "mode" : "affine",
2503
+ "name" : "model.tts_language_model.layers.6.mlp.gate_proj",
2504
+ "out_dim" : 4864,
2505
+ "quant_file" : "model.safetensors",
2506
+ "shape" : [
2507
+ 4864,
2508
+ 896
2509
+ ]
2510
+ },
2511
+ {
2512
+ "bits" : 8,
2513
+ "file" : "model.safetensors",
2514
+ "group_size" : 32,
2515
+ "in_dim" : 64,
2516
+ "mode" : "affine",
2517
+ "name" : "model.acoustic_connector.fc1",
2518
+ "out_dim" : 896,
2519
+ "quant_file" : "model.safetensors",
2520
+ "shape" : [
2521
+ 896,
2522
+ 64
2523
+ ]
2524
+ },
2525
+ {
2526
+ "bits" : 8,
2527
+ "file" : "model.safetensors",
2528
+ "group_size" : 32,
2529
+ "in_dim" : 4864,
2530
+ "mode" : "affine",
2531
+ "name" : "model.tts_language_model.layers.4.mlp.down_proj",
2532
+ "out_dim" : 896,
2533
+ "quant_file" : "model.safetensors",
2534
+ "shape" : [
2535
+ 896,
2536
+ 4864
2537
+ ]
2538
+ },
2539
+ {
2540
+ "bits" : 8,
2541
+ "file" : "model.safetensors",
2542
+ "group_size" : 32,
2543
+ "in_dim" : 896,
2544
+ "mode" : "affine",
2545
+ "name" : "model.tts_language_model.layers.16.self_attn.o_proj",
2546
+ "out_dim" : 896,
2547
+ "quant_file" : "model.safetensors",
2548
+ "shape" : [
2549
+ 896,
2550
+ 896
2551
+ ]
2552
+ },
2553
+ {
2554
+ "bits" : 8,
2555
+ "file" : "model.safetensors",
2556
+ "group_size" : 32,
2557
+ "in_dim" : 8192,
2558
+ "mode" : "affine",
2559
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.5.ffn.linear2",
2560
+ "out_dim" : 2048,
2561
+ "quant_file" : "model.safetensors",
2562
+ "shape" : [
2563
+ 2048,
2564
+ 8192
2565
+ ]
2566
+ },
2567
+ {
2568
+ "bits" : 8,
2569
+ "file" : "model.safetensors",
2570
+ "group_size" : 32,
2571
+ "in_dim" : 896,
2572
+ "mode" : "affine",
2573
+ "name" : "model.language_model.layers.0.self_attn.v_proj",
2574
+ "out_dim" : 128,
2575
+ "quant_file" : "model.safetensors",
2576
+ "shape" : [
2577
+ 128,
2578
+ 896
2579
+ ]
2580
+ },
2581
+ {
2582
+ "bits" : 8,
2583
+ "file" : "model.safetensors",
2584
+ "group_size" : 32,
2585
+ "in_dim" : 896,
2586
+ "mode" : "affine",
2587
+ "name" : "model.tts_language_model.layers.18.self_attn.v_proj",
2588
+ "out_dim" : 128,
2589
+ "quant_file" : "model.safetensors",
2590
+ "shape" : [
2591
+ 128,
2592
+ 896
2593
+ ]
2594
+ },
2595
+ {
2596
+ "bits" : 8,
2597
+ "file" : "model.safetensors",
2598
+ "group_size" : 32,
2599
+ "in_dim" : 896,
2600
+ "mode" : "affine",
2601
+ "name" : "model.tts_language_model.layers.10.self_attn.v_proj",
2602
+ "out_dim" : 128,
2603
+ "quant_file" : "model.safetensors",
2604
+ "shape" : [
2605
+ 128,
2606
+ 896
2607
+ ]
2608
+ },
2609
+ {
2610
+ "bits" : 8,
2611
+ "file" : "model.safetensors",
2612
+ "group_size" : 32,
2613
+ "in_dim" : 4864,
2614
+ "mode" : "affine",
2615
+ "name" : "model.language_model.layers.1.mlp.down_proj",
2616
+ "out_dim" : 896,
2617
+ "quant_file" : "model.safetensors",
2618
+ "shape" : [
2619
+ 896,
2620
+ 4864
2621
+ ]
2622
+ },
2623
+ {
2624
+ "bits" : 8,
2625
+ "file" : "model.safetensors",
2626
+ "group_size" : 32,
2627
+ "in_dim" : 256,
2628
+ "mode" : "affine",
2629
+ "name" : "model.acoustic_tokenizer.decoder.stages.5.1.ffn.linear2",
2630
+ "out_dim" : 64,
2631
+ "quant_file" : "model.safetensors",
2632
+ "shape" : [
2633
+ 64,
2634
+ 256
2635
+ ]
2636
+ },
2637
+ {
2638
+ "bits" : 8,
2639
+ "file" : "model.safetensors",
2640
+ "group_size" : 32,
2641
+ "in_dim" : 896,
2642
+ "mode" : "affine",
2643
+ "name" : "model.prediction_head.final_layer.linear",
2644
+ "out_dim" : 64,
2645
+ "quant_file" : "model.safetensors",
2646
+ "shape" : [
2647
+ 64,
2648
+ 896
2649
+ ]
2650
+ },
2651
+ {
2652
+ "bits" : 8,
2653
+ "file" : "model.safetensors",
2654
+ "group_size" : 32,
2655
+ "in_dim" : 896,
2656
+ "mode" : "affine",
2657
+ "name" : "model.tts_language_model.layers.6.self_attn.q_proj",
2658
+ "out_dim" : 896,
2659
+ "quant_file" : "model.safetensors",
2660
+ "shape" : [
2661
+ 896,
2662
+ 896
2663
+ ]
2664
+ },
2665
+ {
2666
+ "bits" : 8,
2667
+ "file" : "model.safetensors",
2668
+ "group_size" : 32,
2669
+ "in_dim" : 512,
2670
+ "mode" : "affine",
2671
+ "name" : "model.acoustic_tokenizer.decoder.stages.2.0.ffn.linear1",
2672
+ "out_dim" : 2048,
2673
+ "quant_file" : "model.safetensors",
2674
+ "shape" : [
2675
+ 2048,
2676
+ 512
2677
+ ]
2678
+ },
2679
+ {
2680
+ "bits" : 8,
2681
+ "file" : "model.safetensors",
2682
+ "group_size" : 32,
2683
+ "in_dim" : 896,
2684
+ "mode" : "affine",
2685
+ "name" : "model.tts_language_model.layers.17.self_attn.k_proj",
2686
+ "out_dim" : 128,
2687
+ "quant_file" : "model.safetensors",
2688
+ "shape" : [
2689
+ 128,
2690
+ 896
2691
+ ]
2692
+ },
2693
+ {
2694
+ "bits" : 8,
2695
+ "file" : "model.safetensors",
2696
+ "group_size" : 32,
2697
+ "in_dim" : 896,
2698
+ "mode" : "affine",
2699
+ "name" : "model.prediction_head.t_embedder.mlp.2",
2700
+ "out_dim" : 896,
2701
+ "quant_file" : "model.safetensors",
2702
+ "shape" : [
2703
+ 896,
2704
+ 896
2705
+ ]
2706
+ },
2707
+ {
2708
+ "bits" : 8,
2709
+ "file" : "model.safetensors",
2710
+ "group_size" : 32,
2711
+ "in_dim" : 2048,
2712
+ "mode" : "affine",
2713
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.0.ffn.linear1",
2714
+ "out_dim" : 8192,
2715
+ "quant_file" : "model.safetensors",
2716
+ "shape" : [
2717
+ 8192,
2718
+ 2048
2719
+ ]
2720
+ },
2721
+ {
2722
+ "bits" : 8,
2723
+ "file" : "model.safetensors",
2724
+ "group_size" : 32,
2725
+ "in_dim" : 896,
2726
+ "mode" : "affine",
2727
+ "name" : "model.tts_language_model.layers.0.self_attn.o_proj",
2728
+ "out_dim" : 896,
2729
+ "quant_file" : "model.safetensors",
2730
+ "shape" : [
2731
+ 896,
2732
+ 896
2733
+ ]
2734
+ },
2735
+ {
2736
+ "bits" : 8,
2737
+ "file" : "model.safetensors",
2738
+ "group_size" : 32,
2739
+ "in_dim" : 4096,
2740
+ "mode" : "affine",
2741
+ "name" : "model.acoustic_tokenizer.decoder.stages.1.2.ffn.linear2",
2742
+ "out_dim" : 1024,
2743
+ "quant_file" : "model.safetensors",
2744
+ "shape" : [
2745
+ 1024,
2746
+ 4096
2747
+ ]
2748
+ },
2749
+ {
2750
+ "bits" : 8,
2751
+ "file" : "model.safetensors",
2752
+ "group_size" : 32,
2753
+ "in_dim" : 2688,
2754
+ "mode" : "affine",
2755
+ "name" : "model.prediction_head.layers.1.ffn.down_proj",
2756
+ "out_dim" : 896,
2757
+ "quant_file" : "model.safetensors",
2758
+ "shape" : [
2759
+ 896,
2760
+ 2688
2761
+ ]
2762
+ },
2763
+ {
2764
+ "bits" : 8,
2765
+ "file" : "model.safetensors",
2766
+ "group_size" : 32,
2767
+ "in_dim" : 896,
2768
+ "mode" : "affine",
2769
+ "name" : "model.prediction_head.layers.2.ffn.up_proj",
2770
+ "out_dim" : 2688,
2771
+ "quant_file" : "model.safetensors",
2772
+ "shape" : [
2773
+ 2688,
2774
+ 896
2775
+ ]
2776
+ },
2777
+ {
2778
+ "bits" : 8,
2779
+ "file" : "model.safetensors",
2780
+ "group_size" : 32,
2781
+ "in_dim" : 896,
2782
+ "mode" : "affine",
2783
+ "name" : "model.tts_language_model.layers.3.self_attn.k_proj",
2784
+ "out_dim" : 128,
2785
+ "quant_file" : "model.safetensors",
2786
+ "shape" : [
2787
+ 128,
2788
+ 896
2789
+ ]
2790
+ },
2791
+ {
2792
+ "bits" : 8,
2793
+ "file" : "model.safetensors",
2794
+ "group_size" : 32,
2795
+ "in_dim" : 896,
2796
+ "mode" : "affine",
2797
+ "name" : "model.tts_language_model.layers.9.self_attn.k_proj",
2798
+ "out_dim" : 128,
2799
+ "quant_file" : "model.safetensors",
2800
+ "shape" : [
2801
+ 128,
2802
+ 896
2803
+ ]
2804
+ },
2805
+ {
2806
+ "bits" : 8,
2807
+ "file" : "model.safetensors",
2808
+ "group_size" : 32,
2809
+ "in_dim" : 896,
2810
+ "mode" : "affine",
2811
+ "name" : "model.tts_language_model.layers.6.mlp.up_proj",
2812
+ "out_dim" : 4864,
2813
+ "quant_file" : "model.safetensors",
2814
+ "shape" : [
2815
+ 4864,
2816
+ 896
2817
+ ]
2818
+ },
2819
+ {
2820
+ "bits" : 8,
2821
+ "file" : "model.safetensors",
2822
+ "group_size" : 32,
2823
+ "in_dim" : 896,
2824
+ "mode" : "affine",
2825
+ "name" : "model.tts_language_model.layers.14.mlp.gate_proj",
2826
+ "out_dim" : 4864,
2827
+ "quant_file" : "model.safetensors",
2828
+ "shape" : [
2829
+ 4864,
2830
+ 896
2831
+ ]
2832
+ },
2833
+ {
2834
+ "bits" : 8,
2835
+ "file" : "model.safetensors",
2836
+ "group_size" : 32,
2837
+ "in_dim" : 32,
2838
+ "mode" : "affine",
2839
+ "name" : "model.acoustic_tokenizer.decoder.stages.6.1.ffn.linear1",
2840
+ "out_dim" : 128,
2841
+ "quant_file" : "model.safetensors",
2842
+ "shape" : [
2843
+ 128,
2844
+ 32
2845
+ ]
2846
+ },
2847
+ {
2848
+ "bits" : 8,
2849
+ "file" : "model.safetensors",
2850
+ "group_size" : 32,
2851
+ "in_dim" : 32,
2852
+ "mode" : "affine",
2853
+ "name" : "model.acoustic_tokenizer.decoder.stages.6.0.ffn.linear1",
2854
+ "out_dim" : 128,
2855
+ "quant_file" : "model.safetensors",
2856
+ "shape" : [
2857
+ 128,
2858
+ 32
2859
+ ]
2860
+ },
2861
+ {
2862
+ "bits" : 8,
2863
+ "file" : "model.safetensors",
2864
+ "group_size" : 32,
2865
+ "in_dim" : 896,
2866
+ "mode" : "affine",
2867
+ "name" : "model.tts_language_model.layers.14.self_attn.k_proj",
2868
+ "out_dim" : 128,
2869
+ "quant_file" : "model.safetensors",
2870
+ "shape" : [
2871
+ 128,
2872
+ 896
2873
+ ]
2874
+ },
2875
+ {
2876
+ "bits" : 8,
2877
+ "file" : "model.safetensors",
2878
+ "group_size" : 32,
2879
+ "in_dim" : 256,
2880
+ "mode" : "affine",
2881
+ "name" : "model.acoustic_tokenizer.decoder.stages.5.2.ffn.linear2",
2882
+ "out_dim" : 64,
2883
+ "quant_file" : "model.safetensors",
2884
+ "shape" : [
2885
+ 64,
2886
+ 256
2887
+ ]
2888
+ },
2889
+ {
2890
+ "bits" : 8,
2891
+ "file" : "model.safetensors",
2892
+ "group_size" : 32,
2893
+ "in_dim" : 896,
2894
+ "mode" : "affine",
2895
+ "name" : "model.language_model.layers.2.self_attn.v_proj",
2896
+ "out_dim" : 128,
2897
+ "quant_file" : "model.safetensors",
2898
+ "shape" : [
2899
+ 128,
2900
+ 896
2901
+ ]
2902
+ },
2903
+ {
2904
+ "bits" : 8,
2905
+ "file" : "model.safetensors",
2906
+ "group_size" : 32,
2907
+ "in_dim" : 896,
2908
+ "mode" : "affine",
2909
+ "name" : "model.tts_language_model.layers.18.self_attn.k_proj",
2910
+ "out_dim" : 128,
2911
+ "quant_file" : "model.safetensors",
2912
+ "shape" : [
2913
+ 128,
2914
+ 896
2915
+ ]
2916
+ },
2917
+ {
2918
+ "bits" : 8,
2919
+ "file" : "model.safetensors",
2920
+ "group_size" : 32,
2921
+ "in_dim" : 4864,
2922
+ "mode" : "affine",
2923
+ "name" : "model.tts_language_model.layers.8.mlp.down_proj",
2924
+ "out_dim" : 896,
2925
+ "quant_file" : "model.safetensors",
2926
+ "shape" : [
2927
+ 896,
2928
+ 4864
2929
+ ]
2930
+ },
2931
+ {
2932
+ "bits" : 8,
2933
+ "file" : "model.safetensors",
2934
+ "group_size" : 32,
2935
+ "in_dim" : 896,
2936
+ "mode" : "affine",
2937
+ "name" : "model.tts_language_model.layers.2.mlp.up_proj",
2938
+ "out_dim" : 4864,
2939
+ "quant_file" : "model.safetensors",
2940
+ "shape" : [
2941
+ 4864,
2942
+ 896
2943
+ ]
2944
+ },
2945
+ {
2946
+ "bits" : 8,
2947
+ "file" : "model.safetensors",
2948
+ "group_size" : 32,
2949
+ "in_dim" : 896,
2950
+ "mode" : "affine",
2951
+ "name" : "model.tts_language_model.layers.18.mlp.up_proj",
2952
+ "out_dim" : 4864,
2953
+ "quant_file" : "model.safetensors",
2954
+ "shape" : [
2955
+ 4864,
2956
+ 896
2957
+ ]
2958
+ },
2959
+ {
2960
+ "bits" : 8,
2961
+ "file" : "model.safetensors",
2962
+ "group_size" : 32,
2963
+ "in_dim" : 4864,
2964
+ "mode" : "affine",
2965
+ "name" : "model.language_model.layers.0.mlp.down_proj",
2966
+ "out_dim" : 896,
2967
+ "quant_file" : "model.safetensors",
2968
+ "shape" : [
2969
+ 896,
2970
+ 4864
2971
+ ]
2972
+ },
2973
+ {
2974
+ "bits" : 8,
2975
+ "file" : "model.safetensors",
2976
+ "group_size" : 32,
2977
+ "in_dim" : 896,
2978
+ "mode" : "affine",
2979
+ "name" : "model.tts_language_model.layers.1.mlp.up_proj",
2980
+ "out_dim" : 4864,
2981
+ "quant_file" : "model.safetensors",
2982
+ "shape" : [
2983
+ 4864,
2984
+ 896
2985
+ ]
2986
+ },
2987
+ {
2988
+ "bits" : 8,
2989
+ "file" : "model.safetensors",
2990
+ "group_size" : 32,
2991
+ "in_dim" : 896,
2992
+ "mode" : "affine",
2993
+ "name" : "model.tts_language_model.layers.3.self_attn.o_proj",
2994
+ "out_dim" : 896,
2995
+ "quant_file" : "model.safetensors",
2996
+ "shape" : [
2997
+ 896,
2998
+ 896
2999
+ ]
3000
+ },
3001
+ {
3002
+ "bits" : 8,
3003
+ "file" : "model.safetensors",
3004
+ "group_size" : 32,
3005
+ "in_dim" : 896,
3006
+ "mode" : "affine",
3007
+ "name" : "model.tts_language_model.layers.6.self_attn.o_proj",
3008
+ "out_dim" : 896,
3009
+ "quant_file" : "model.safetensors",
3010
+ "shape" : [
3011
+ 896,
3012
+ 896
3013
+ ]
3014
+ },
3015
+ {
3016
+ "bits" : 8,
3017
+ "file" : "model.safetensors",
3018
+ "group_size" : 32,
3019
+ "in_dim" : 4864,
3020
+ "mode" : "affine",
3021
+ "name" : "model.tts_language_model.layers.13.mlp.down_proj",
3022
+ "out_dim" : 896,
3023
+ "quant_file" : "model.safetensors",
3024
+ "shape" : [
3025
+ 896,
3026
+ 4864
3027
+ ]
3028
+ },
3029
+ {
3030
+ "bits" : 8,
3031
+ "file" : "model.safetensors",
3032
+ "group_size" : 32,
3033
+ "in_dim" : 512,
3034
+ "mode" : "affine",
3035
+ "name" : "model.acoustic_tokenizer.decoder.stages.2.1.ffn.linear1",
3036
+ "out_dim" : 2048,
3037
+ "quant_file" : "model.safetensors",
3038
+ "shape" : [
3039
+ 2048,
3040
+ 512
3041
+ ]
3042
+ },
3043
+ {
3044
+ "bits" : 8,
3045
+ "file" : "model.safetensors",
3046
+ "group_size" : 32,
3047
+ "in_dim" : 896,
3048
+ "mode" : "affine",
3049
+ "name" : "model.tts_language_model.layers.0.self_attn.v_proj",
3050
+ "out_dim" : 128,
3051
+ "quant_file" : "model.safetensors",
3052
+ "shape" : [
3053
+ 128,
3054
+ 896
3055
+ ]
3056
+ },
3057
+ {
3058
+ "bits" : 8,
3059
+ "file" : "model.safetensors",
3060
+ "group_size" : 32,
3061
+ "in_dim" : 896,
3062
+ "mode" : "affine",
3063
+ "name" : "model.tts_language_model.layers.3.mlp.up_proj",
3064
+ "out_dim" : 4864,
3065
+ "quant_file" : "model.safetensors",
3066
+ "shape" : [
3067
+ 4864,
3068
+ 896
3069
+ ]
3070
+ },
3071
+ {
3072
+ "bits" : 8,
3073
+ "file" : "model.safetensors",
3074
+ "group_size" : 32,
3075
+ "in_dim" : 896,
3076
+ "mode" : "affine",
3077
+ "name" : "model.language_model.layers.3.self_attn.v_proj",
3078
+ "out_dim" : 128,
3079
+ "quant_file" : "model.safetensors",
3080
+ "shape" : [
3081
+ 128,
3082
+ 896
3083
+ ]
3084
+ },
3085
+ {
3086
+ "bits" : 8,
3087
+ "file" : "model.safetensors",
3088
+ "group_size" : 32,
3089
+ "in_dim" : 896,
3090
+ "mode" : "affine",
3091
+ "name" : "model.language_model.layers.0.self_attn.k_proj",
3092
+ "out_dim" : 128,
3093
+ "quant_file" : "model.safetensors",
3094
+ "shape" : [
3095
+ 128,
3096
+ 896
3097
+ ]
3098
+ },
3099
+ {
3100
+ "bits" : 8,
3101
+ "file" : "model.safetensors",
3102
+ "group_size" : 32,
3103
+ "in_dim" : 4864,
3104
+ "mode" : "affine",
3105
+ "name" : "model.language_model.layers.3.mlp.down_proj",
3106
+ "out_dim" : 896,
3107
+ "quant_file" : "model.safetensors",
3108
+ "shape" : [
3109
+ 896,
3110
+ 4864
3111
+ ]
3112
+ },
3113
+ {
3114
+ "bits" : 8,
3115
+ "file" : "model.safetensors",
3116
+ "group_size" : 32,
3117
+ "in_dim" : 896,
3118
+ "mode" : "affine",
3119
+ "name" : "model.tts_language_model.layers.19.self_attn.v_proj",
3120
+ "out_dim" : 128,
3121
+ "quant_file" : "model.safetensors",
3122
+ "shape" : [
3123
+ 128,
3124
+ 896
3125
+ ]
3126
+ },
3127
+ {
3128
+ "bits" : 8,
3129
+ "file" : "model.safetensors",
3130
+ "group_size" : 32,
3131
+ "in_dim" : 896,
3132
+ "mode" : "affine",
3133
+ "name" : "model.language_model.layers.1.mlp.up_proj",
3134
+ "out_dim" : 4864,
3135
+ "quant_file" : "model.safetensors",
3136
+ "shape" : [
3137
+ 4864,
3138
+ 896
3139
+ ]
3140
+ },
3141
+ {
3142
+ "bits" : 8,
3143
+ "file" : "model.safetensors",
3144
+ "group_size" : 32,
3145
+ "in_dim" : 896,
3146
+ "mode" : "affine",
3147
+ "name" : "model.tts_language_model.layers.12.self_attn.v_proj",
3148
+ "out_dim" : 128,
3149
+ "quant_file" : "model.safetensors",
3150
+ "shape" : [
3151
+ 128,
3152
+ 896
3153
+ ]
3154
+ },
3155
+ {
3156
+ "bits" : 8,
3157
+ "file" : "model.safetensors",
3158
+ "group_size" : 32,
3159
+ "in_dim" : 896,
3160
+ "mode" : "affine",
3161
+ "name" : "model.tts_language_model.layers.17.mlp.up_proj",
3162
+ "out_dim" : 4864,
3163
+ "quant_file" : "model.safetensors",
3164
+ "shape" : [
3165
+ 4864,
3166
+ 896
3167
+ ]
3168
+ },
3169
+ {
3170
+ "bits" : 8,
3171
+ "file" : "model.safetensors",
3172
+ "group_size" : 32,
3173
+ "in_dim" : 896,
3174
+ "mode" : "affine",
3175
+ "name" : "model.prediction_head.final_layer.adaLN_modulation.1",
3176
+ "out_dim" : 1792,
3177
+ "quant_file" : "model.safetensors",
3178
+ "shape" : [
3179
+ 1792,
3180
+ 896
3181
+ ]
3182
+ },
3183
+ {
3184
+ "bits" : 8,
3185
+ "file" : "model.safetensors",
3186
+ "group_size" : 32,
3187
+ "in_dim" : 8192,
3188
+ "mode" : "affine",
3189
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.3.ffn.linear2",
3190
+ "out_dim" : 2048,
3191
+ "quant_file" : "model.safetensors",
3192
+ "shape" : [
3193
+ 2048,
3194
+ 8192
3195
+ ]
3196
+ },
3197
+ {
3198
+ "bits" : 8,
3199
+ "file" : "model.safetensors",
3200
+ "group_size" : 32,
3201
+ "in_dim" : 896,
3202
+ "mode" : "affine",
3203
+ "name" : "model.tts_language_model.layers.17.self_attn.v_proj",
3204
+ "out_dim" : 128,
3205
+ "quant_file" : "model.safetensors",
3206
+ "shape" : [
3207
+ 128,
3208
+ 896
3209
+ ]
3210
+ },
3211
+ {
3212
+ "bits" : 8,
3213
+ "file" : "model.safetensors",
3214
+ "group_size" : 32,
3215
+ "in_dim" : 896,
3216
+ "mode" : "affine",
3217
+ "name" : "model.tts_language_model.layers.8.self_attn.q_proj",
3218
+ "out_dim" : 896,
3219
+ "quant_file" : "model.safetensors",
3220
+ "shape" : [
3221
+ 896,
3222
+ 896
3223
+ ]
3224
+ },
3225
+ {
3226
+ "bits" : 8,
3227
+ "file" : "model.safetensors",
3228
+ "group_size" : 32,
3229
+ "in_dim" : 8192,
3230
+ "mode" : "affine",
3231
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.6.ffn.linear2",
3232
+ "out_dim" : 2048,
3233
+ "quant_file" : "model.safetensors",
3234
+ "shape" : [
3235
+ 2048,
3236
+ 8192
3237
+ ]
3238
+ },
3239
+ {
3240
+ "bits" : 8,
3241
+ "file" : "model.safetensors",
3242
+ "group_size" : 32,
3243
+ "in_dim" : 64,
3244
+ "mode" : "affine",
3245
+ "name" : "model.acoustic_tokenizer.decoder.stages.5.1.ffn.linear1",
3246
+ "out_dim" : 256,
3247
+ "quant_file" : "model.safetensors",
3248
+ "shape" : [
3249
+ 256,
3250
+ 64
3251
+ ]
3252
+ },
3253
+ {
3254
+ "bits" : 8,
3255
+ "file" : "model.safetensors",
3256
+ "group_size" : 32,
3257
+ "in_dim" : 896,
3258
+ "mode" : "affine",
3259
+ "name" : "model.language_model.layers.3.self_attn.q_proj",
3260
+ "out_dim" : 896,
3261
+ "quant_file" : "model.safetensors",
3262
+ "shape" : [
3263
+ 896,
3264
+ 896
3265
+ ]
3266
+ },
3267
+ {
3268
+ "bits" : 8,
3269
+ "file" : "model.safetensors",
3270
+ "group_size" : 32,
3271
+ "in_dim" : 896,
3272
+ "mode" : "affine",
3273
+ "name" : "model.tts_language_model.layers.2.mlp.gate_proj",
3274
+ "out_dim" : 4864,
3275
+ "quant_file" : "model.safetensors",
3276
+ "shape" : [
3277
+ 4864,
3278
+ 896
3279
+ ]
3280
+ },
3281
+ {
3282
+ "bits" : 8,
3283
+ "file" : "model.safetensors",
3284
+ "group_size" : 32,
3285
+ "in_dim" : 896,
3286
+ "mode" : "affine",
3287
+ "name" : "model.prediction_head.layers.1.adaLN_modulation.1",
3288
+ "out_dim" : 2688,
3289
+ "quant_file" : "model.safetensors",
3290
+ "shape" : [
3291
+ 2688,
3292
+ 896
3293
+ ]
3294
+ },
3295
+ {
3296
+ "bits" : 8,
3297
+ "file" : "model.safetensors",
3298
+ "group_size" : 32,
3299
+ "in_dim" : 4864,
3300
+ "mode" : "affine",
3301
+ "name" : "model.tts_language_model.layers.12.mlp.down_proj",
3302
+ "out_dim" : 896,
3303
+ "quant_file" : "model.safetensors",
3304
+ "shape" : [
3305
+ 896,
3306
+ 4864
3307
+ ]
3308
+ },
3309
+ {
3310
+ "bits" : 8,
3311
+ "file" : "model.safetensors",
3312
+ "group_size" : 32,
3313
+ "in_dim" : 8192,
3314
+ "mode" : "affine",
3315
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.2.ffn.linear2",
3316
+ "out_dim" : 2048,
3317
+ "quant_file" : "model.safetensors",
3318
+ "shape" : [
3319
+ 2048,
3320
+ 8192
3321
+ ]
3322
+ },
3323
+ {
3324
+ "bits" : 8,
3325
+ "file" : "model.safetensors",
3326
+ "group_size" : 32,
3327
+ "in_dim" : 896,
3328
+ "mode" : "affine",
3329
+ "name" : "model.tts_language_model.layers.15.self_attn.q_proj",
3330
+ "out_dim" : 896,
3331
+ "quant_file" : "model.safetensors",
3332
+ "shape" : [
3333
+ 896,
3334
+ 896
3335
+ ]
3336
+ },
3337
+ {
3338
+ "bits" : 8,
3339
+ "file" : "model.safetensors",
3340
+ "group_size" : 32,
3341
+ "in_dim" : 256,
3342
+ "mode" : "affine",
3343
+ "name" : "model.prediction_head.t_embedder.mlp.0",
3344
+ "out_dim" : 896,
3345
+ "quant_file" : "model.safetensors",
3346
+ "shape" : [
3347
+ 896,
3348
+ 256
3349
+ ]
3350
+ },
3351
+ {
3352
+ "bits" : 8,
3353
+ "file" : "model.safetensors",
3354
+ "group_size" : 32,
3355
+ "in_dim" : 896,
3356
+ "mode" : "affine",
3357
+ "name" : "model.tts_language_model.layers.8.mlp.gate_proj",
3358
+ "out_dim" : 4864,
3359
+ "quant_file" : "model.safetensors",
3360
+ "shape" : [
3361
+ 4864,
3362
+ 896
3363
+ ]
3364
+ },
3365
+ {
3366
+ "bits" : 8,
3367
+ "file" : "model.safetensors",
3368
+ "group_size" : 32,
3369
+ "in_dim" : 896,
3370
+ "mode" : "affine",
3371
+ "name" : "model.tts_language_model.layers.16.self_attn.v_proj",
3372
+ "out_dim" : 128,
3373
+ "quant_file" : "model.safetensors",
3374
+ "shape" : [
3375
+ 128,
3376
+ 896
3377
+ ]
3378
+ },
3379
+ {
3380
+ "bits" : 8,
3381
+ "file" : "model.safetensors",
3382
+ "group_size" : 32,
3383
+ "in_dim" : 896,
3384
+ "mode" : "affine",
3385
+ "name" : "model.prediction_head.layers.0.ffn.gate_proj",
3386
+ "out_dim" : 2688,
3387
+ "quant_file" : "model.safetensors",
3388
+ "shape" : [
3389
+ 2688,
3390
+ 896
3391
+ ]
3392
+ },
3393
+ {
3394
+ "bits" : 8,
3395
+ "file" : "model.safetensors",
3396
+ "group_size" : 32,
3397
+ "in_dim" : 896,
3398
+ "mode" : "affine",
3399
+ "name" : "model.tts_language_model.layers.10.self_attn.q_proj",
3400
+ "out_dim" : 896,
3401
+ "quant_file" : "model.safetensors",
3402
+ "shape" : [
3403
+ 896,
3404
+ 896
3405
+ ]
3406
+ },
3407
+ {
3408
+ "bits" : 8,
3409
+ "file" : "model.safetensors",
3410
+ "group_size" : 32,
3411
+ "in_dim" : 8192,
3412
+ "mode" : "affine",
3413
+ "name" : "model.acoustic_tokenizer.decoder.stages.0.1.ffn.linear2",
3414
+ "out_dim" : 2048,
3415
+ "quant_file" : "model.safetensors",
3416
+ "shape" : [
3417
+ 2048,
3418
+ 8192
3419
+ ]
3420
+ }
3421
+ ],
3422
+ "mode" : "affine"
3423
+ }