Update config.json
Browse files- config.json +28 -9
config.json
CHANGED
@@ -1,5 +1,6 @@
|
|
1 |
{
|
2 |
-
"
|
|
|
3 |
"alignment_head": [
|
4 |
2,
|
5 |
null,
|
@@ -17,7 +18,7 @@
|
|
17 |
2,
|
18 |
2
|
19 |
],
|
20 |
-
"blocks":
|
21 |
"bos_token_id": 50256,
|
22 |
"c_res": 1,
|
23 |
"cond_c_res": [
|
@@ -62,9 +63,14 @@
|
|
62 |
"emb_dropout": 0.1,
|
63 |
"eos_token_id": 50256,
|
64 |
"fp16_params": true,
|
|
|
|
|
|
|
|
|
|
|
65 |
"hop_length": 256,
|
66 |
"init_scale": [
|
67 |
-
0.
|
68 |
1,
|
69 |
1
|
70 |
],
|
@@ -80,7 +86,7 @@
|
|
80 |
false,
|
81 |
false
|
82 |
],
|
83 |
-
"min_duration":
|
84 |
"mlp_init_scale": 0.02,
|
85 |
"model_type": "jukebox",
|
86 |
"multispec_loss_hop_length": [
|
@@ -107,13 +113,14 @@
|
|
107 |
"n_embd": 768,
|
108 |
"n_head": 12,
|
109 |
"n_heads": [
|
110 |
-
|
111 |
1,
|
112 |
1
|
113 |
],
|
114 |
"n_inner": null,
|
115 |
"n_layer": 12,
|
116 |
"n_positions": 1024,
|
|
|
117 |
"n_tokens": [
|
118 |
384,
|
119 |
0,
|
@@ -174,8 +181,8 @@
|
|
174 |
"res_scale": false,
|
175 |
"resid_dropout": 0.0,
|
176 |
"sample_hop_length": 30000,
|
177 |
-
"sample_length":
|
178 |
-
"sample_length_in_seconds":
|
179 |
"scale_attn_by_inverse_layer_idx": false,
|
180 |
"scale_attn_weights": true,
|
181 |
"single_enc_dec": [
|
@@ -185,7 +192,7 @@
|
|
185 |
],
|
186 |
"spectral": 0.0,
|
187 |
"spread": null,
|
188 |
-
"sr":
|
189 |
"strides_t": [
|
190 |
2,
|
191 |
2,
|
@@ -197,7 +204,8 @@
|
|
197 |
"summary_type": "cls_index",
|
198 |
"summary_use_proj": true,
|
199 |
"t_bins": 64,
|
200 |
-
"
|
|
|
201 |
"use_cache": true,
|
202 |
"use_nonrelative_specloss": true,
|
203 |
"use_tokens": [
|
@@ -234,6 +242,17 @@
|
|
234 |
2
|
235 |
],
|
236 |
"vq_vae_width": 32,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
237 |
"width": [
|
238 |
2048,
|
239 |
1920,
|
|
|
1 |
{
|
2 |
+
"_name_or_path": "ArthurZ/jukebox-1b-lyrics",
|
3 |
+
"activation_function": "quick_gelu",
|
4 |
"alignment_head": [
|
5 |
2,
|
6 |
null,
|
|
|
18 |
2,
|
19 |
2
|
20 |
],
|
21 |
+
"blocks": 64,
|
22 |
"bos_token_id": 50256,
|
23 |
"c_res": 1,
|
24 |
"cond_c_res": [
|
|
|
63 |
"emb_dropout": 0.1,
|
64 |
"eos_token_id": 50256,
|
65 |
"fp16_params": true,
|
66 |
+
"hop_fraction": [
|
67 |
+
0.125,
|
68 |
+
0.5,
|
69 |
+
0.5
|
70 |
+
],
|
71 |
"hop_length": 256,
|
72 |
"init_scale": [
|
73 |
+
0.2,
|
74 |
1,
|
75 |
1
|
76 |
],
|
|
|
86 |
false,
|
87 |
false
|
88 |
],
|
89 |
+
"min_duration": 17.84,
|
90 |
"mlp_init_scale": 0.02,
|
91 |
"model_type": "jukebox",
|
92 |
"multispec_loss_hop_length": [
|
|
|
113 |
"n_embd": 768,
|
114 |
"n_head": 12,
|
115 |
"n_heads": [
|
116 |
+
2,
|
117 |
1,
|
118 |
1
|
119 |
],
|
120 |
"n_inner": null,
|
121 |
"n_layer": 12,
|
122 |
"n_positions": 1024,
|
123 |
+
"n_samples": 1,
|
124 |
"n_tokens": [
|
125 |
384,
|
126 |
0,
|
|
|
181 |
"res_scale": false,
|
182 |
"resid_dropout": 0.0,
|
183 |
"sample_hop_length": 30000,
|
184 |
+
"sample_length": 2645888,
|
185 |
+
"sample_length_in_seconds": 60,
|
186 |
"scale_attn_by_inverse_layer_idx": false,
|
187 |
"scale_attn_weights": true,
|
188 |
"single_enc_dec": [
|
|
|
192 |
],
|
193 |
"spectral": 0.0,
|
194 |
"spread": null,
|
195 |
+
"sr": 44100,
|
196 |
"strides_t": [
|
197 |
2,
|
198 |
2,
|
|
|
204 |
"summary_type": "cls_index",
|
205 |
"summary_use_proj": true,
|
206 |
"t_bins": 64,
|
207 |
+
"total_sample_length_in_seconds": 180,
|
208 |
+
"transformers_version": "4.21.0.dev0",
|
209 |
"use_cache": true,
|
210 |
"use_nonrelative_specloss": true,
|
211 |
"use_tokens": [
|
|
|
242 |
2
|
243 |
],
|
244 |
"vq_vae_width": 32,
|
245 |
+
"vqvae_z_shapes": [
|
246 |
+
[
|
247 |
+
8192
|
248 |
+
],
|
249 |
+
[
|
250 |
+
32768
|
251 |
+
],
|
252 |
+
[
|
253 |
+
131072
|
254 |
+
]
|
255 |
+
],
|
256 |
"width": [
|
257 |
2048,
|
258 |
1920,
|