Hugging Face's logo

ArthurZ
/
jukebox-1b-lyrics

Feature Extraction
Transformers
PyTorch
jukebox
MusicGeneration
Inference Endpoints
Model card Files Community
13
ArthurZ HF staff commited on
Commit
6662b48
•
1 Parent(s): bb54e3b

add config

Browse files
Files changed (1) hide show
  1. config.json +243 -0
config.json ADDED
@@ -0,0 +1,243 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation_function": "gelu_new",
3
+ "alignment_head": [
4
+ 2,
5
+ null,
6
+ null
7
+ ],
8
+ "alignment_layer": [
9
+ 68,
10
+ null,
11
+ null
12
+ ],
13
+ "attn_dropout": 0.0,
14
+ "attn_init_scale": 1.0,
15
+ "attn_order": [
16
+ 10,
17
+ 2,
18
+ 2
19
+ ],
20
+ "blocks": 16,
21
+ "bos_token_id": 50256,
22
+ "c_res": 1,
23
+ "cond_c_res": [
24
+ 0,
25
+ 1,
26
+ 1
27
+ ],
28
+ "cond_depth": [
29
+ 3,
30
+ 16,
31
+ 16
32
+ ],
33
+ "cond_dilation_cycle": [
34
+ null,
35
+ 8,
36
+ 8
37
+ ],
38
+ "cond_dilation_growth_rate": [
39
+ 1,
40
+ 3,
41
+ 3
42
+ ],
43
+ "cond_m_conv": 1,
44
+ "cond_res_scale": false,
45
+ "cond_width": [
46
+ 128,
47
+ 1024,
48
+ 1024
49
+ ],
50
+ "cond_zero_out": false,
51
+ "copy_input": false,
52
+ "depth": [
53
+ 79,
54
+ 72,
55
+ 72
56
+ ],
57
+ "downs_t": [
58
+ 3,
59
+ 2,
60
+ 2
61
+ ],
62
+ "emb_dropout": 0.1,
63
+ "eos_token_id": 50256,
64
+ "fp16_params": true,
65
+ "hop_length": 256,
66
+ "init_scale": [
67
+ 0.7,
68
+ 1,
69
+ 1
70
+ ],
71
+ "initializer_range": 0.02,
72
+ "l_bins": 256,
73
+ "labels": false,
74
+ "layer_norm_epsilon": 1e-05,
75
+ "m_attn": 0.25,
76
+ "max_bow_genre_size": 1,
77
+ "max_duration": 600.0,
78
+ "merged_decoder": [
79
+ true,
80
+ false,
81
+ false
82
+ ],
83
+ "min_duration": 1,
84
+ "mlp_init_scale": 0.02,
85
+ "model_type": "jukebox",
86
+ "multispec_loss_hop_length": [
87
+ 240,
88
+ 120,
89
+ 50
90
+ ],
91
+ "multispec_loss_n_fft": [
92
+ 2048,
93
+ 1024,
94
+ 512
95
+ ],
96
+ "multispec_loss_window_size": [
97
+ 1200,
98
+ 600,
99
+ 240
100
+ ],
101
+ "multispectral": 1.0,
102
+ "n_ctx": 8192,
103
+ "n_embd": 768,
104
+ "n_head": 12,
105
+ "n_heads": [
106
+ 8,
107
+ 1,
108
+ 1
109
+ ],
110
+ "n_inner": null,
111
+ "n_layer": 12,
112
+ "n_positions": 1024,
113
+ "n_tokens": [
114
+ 512,
115
+ 0,
116
+ 0
117
+ ],
118
+ "n_vocab": 80,
119
+ "name": "AudioSamples",
120
+ "nb_priors": 3,
121
+ "pos_init": false,
122
+ "prime_attn_dropout": 0.0,
123
+ "prime_attn_order": [
124
+ 2,
125
+ 0,
126
+ 0
127
+ ],
128
+ "prime_blocks": 32,
129
+ "prime_c_res": 1,
130
+ "prime_cond_c_res": [
131
+ 0,
132
+ 1,
133
+ 1
134
+ ],
135
+ "prime_depth": [
136
+ 18,
137
+ 3,
138
+ 3
139
+ ],
140
+ "prime_emb_dropout": 0.0,
141
+ "prime_heads": 4,
142
+ "prime_init_scale": [
143
+ 0.1,
144
+ 0.4,
145
+ 0.4
146
+ ],
147
+ "prime_loss_fraction": [
148
+ 0.4,
149
+ 0.0,
150
+ 0.0
151
+ ],
152
+ "prime_m_attn": 0.25,
153
+ "prime_m_mlp": 1.0,
154
+ "prime_pos_init": false,
155
+ "prime_res_scale": false,
156
+ "prime_resid_dropout": 0.0,
157
+ "prime_spread": null,
158
+ "prime_width": [
159
+ 128,
160
+ 128,
161
+ 128
162
+ ],
163
+ "prime_zero_out": false,
164
+ "priors_width": [
165
+ 4096,
166
+ 2048,
167
+ 1024
168
+ ],
169
+ "reorder_and_upcast_attn": false,
170
+ "res_scale": false,
171
+ "resid_dropout": 0.0,
172
+ "sample_hop_length": 30000,
173
+ "sample_length": null,
174
+ "sample_length_in_seconds": 1,
175
+ "scale_attn_by_inverse_layer_idx": false,
176
+ "scale_attn_weights": true,
177
+ "single_enc_dec": [
178
+ true,
179
+ false,
180
+ false
181
+ ],
182
+ "spectral": 0.0,
183
+ "spread": null,
184
+ "sr": 16000,
185
+ "strides_t": [
186
+ 2,
187
+ 2,
188
+ 2
189
+ ],
190
+ "summary_activation": null,
191
+ "summary_first_dropout": 0.1,
192
+ "summary_proj_to_labels": true,
193
+ "summary_type": "cls_index",
194
+ "summary_use_proj": true,
195
+ "t_bins": 128,
196
+ "transformers_version": "4.20.0.dev0",
197
+ "use_cache": true,
198
+ "use_nonrelative_specloss": true,
199
+ "use_tokens": [
200
+ true,
201
+ false,
202
+ false
203
+ ],
204
+ "vocab_size": 50257,
205
+ "vq_vae_codebook_dimension": 2048,
206
+ "vq_vae_commit": 0.02,
207
+ "vq_vae_conv_block_depth": 4,
208
+ "vq_vae_conv_block_width": 64,
209
+ "vq_vae_depth": 4,
210
+ "vq_vae_dilation_cycle": null,
211
+ "vq_vae_dilation_growth_rate": 3,
212
+ "vq_vae_downs_t": [
213
+ 3,
214
+ 2,
215
+ 2
216
+ ],
217
+ "vq_vae_emmbedding_width": 2048,
218
+ "vq_vae_levels": 3,
219
+ "vq_vae_lmu": 0.99,
220
+ "vq_vae_m_conv": 1,
221
+ "vq_vae_multipliers": [
222
+ 2,
223
+ 1,
224
+ 1
225
+ ],
226
+ "vq_vae_reverse_decoder_dilation": 1,
227
+ "vq_vae_strides_t": [
228
+ 2,
229
+ 2,
230
+ 2
231
+ ],
232
+ "vq_vae_width": 64,
233
+ "width": [
234
+ 4800,
235
+ 1920,
236
+ 128
237
+ ],
238
+ "y_bins": [
239
+ 120,
240
+ 4111
241
+ ],
242
+ "zero_out": false
243
+ }