Upload config.json
Browse files- config.json +30 -16
config.json
CHANGED
@@ -13,7 +13,7 @@
|
|
13 |
"attn_dropout": 0.0,
|
14 |
"attn_init_scale": 1.0,
|
15 |
"attn_order": [
|
16 |
-
|
17 |
2,
|
18 |
2
|
19 |
],
|
@@ -50,7 +50,7 @@
|
|
50 |
"cond_zero_out": false,
|
51 |
"copy_input": false,
|
52 |
"depth": [
|
53 |
-
|
54 |
72,
|
55 |
72
|
56 |
],
|
@@ -69,8 +69,8 @@
|
|
69 |
1
|
70 |
],
|
71 |
"initializer_range": 0.02,
|
72 |
-
"l_bins":
|
73 |
-
"labels":
|
74 |
"layer_norm_epsilon": 1e-05,
|
75 |
"m_attn": 0.25,
|
76 |
"max_bow_genre_size": 1,
|
@@ -99,7 +99,11 @@
|
|
99 |
240
|
100 |
],
|
101 |
"multispectral": 1.0,
|
102 |
-
"n_ctx":
|
|
|
|
|
|
|
|
|
103 |
"n_embd": 768,
|
104 |
"n_head": 12,
|
105 |
"n_heads": [
|
@@ -111,11 +115,11 @@
|
|
111 |
"n_layer": 12,
|
112 |
"n_positions": 1024,
|
113 |
"n_tokens": [
|
114 |
-
|
115 |
0,
|
116 |
0
|
117 |
],
|
118 |
-
"n_vocab":
|
119 |
"name": "AudioSamples",
|
120 |
"nb_priors": 3,
|
121 |
"pos_init": false,
|
@@ -162,7 +166,7 @@
|
|
162 |
],
|
163 |
"prime_zero_out": false,
|
164 |
"priors_width": [
|
165 |
-
|
166 |
2048,
|
167 |
1024
|
168 |
],
|
@@ -192,7 +196,7 @@
|
|
192 |
"summary_proj_to_labels": true,
|
193 |
"summary_type": "cls_index",
|
194 |
"summary_use_proj": true,
|
195 |
-
"t_bins":
|
196 |
"transformers_version": "4.20.0.dev0",
|
197 |
"use_cache": true,
|
198 |
"use_nonrelative_specloss": true,
|
@@ -205,7 +209,7 @@
|
|
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":
|
209 |
"vq_vae_depth": 4,
|
210 |
"vq_vae_dilation_cycle": null,
|
211 |
"vq_vae_dilation_growth_rate": 3,
|
@@ -214,7 +218,7 @@
|
|
214 |
2,
|
215 |
2
|
216 |
],
|
217 |
-
"vq_vae_emmbedding_width":
|
218 |
"vq_vae_levels": 3,
|
219 |
"vq_vae_lmu": 0.99,
|
220 |
"vq_vae_m_conv": 1,
|
@@ -229,15 +233,25 @@
|
|
229 |
2,
|
230 |
2
|
231 |
],
|
232 |
-
"vq_vae_width":
|
233 |
"width": [
|
234 |
-
|
235 |
1920,
|
236 |
-
|
237 |
],
|
238 |
"y_bins": [
|
239 |
-
|
240 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
241 |
],
|
242 |
"zero_out": false
|
243 |
}
|
|
|
13 |
"attn_dropout": 0.0,
|
14 |
"attn_init_scale": 1.0,
|
15 |
"attn_order": [
|
16 |
+
12,
|
17 |
2,
|
18 |
2
|
19 |
],
|
|
|
50 |
"cond_zero_out": false,
|
51 |
"copy_input": false,
|
52 |
"depth": [
|
53 |
+
72,
|
54 |
72,
|
55 |
72
|
56 |
],
|
|
|
69 |
1
|
70 |
],
|
71 |
"initializer_range": 0.02,
|
72 |
+
"l_bins": 2048,
|
73 |
+
"labels": true,
|
74 |
"layer_norm_epsilon": 1e-05,
|
75 |
"m_attn": 0.25,
|
76 |
"max_bow_genre_size": 1,
|
|
|
99 |
240
|
100 |
],
|
101 |
"multispectral": 1.0,
|
102 |
+
"n_ctx": [
|
103 |
+
6144,
|
104 |
+
8192,
|
105 |
+
8192
|
106 |
+
],
|
107 |
"n_embd": 768,
|
108 |
"n_head": 12,
|
109 |
"n_heads": [
|
|
|
115 |
"n_layer": 12,
|
116 |
"n_positions": 1024,
|
117 |
"n_tokens": [
|
118 |
+
384,
|
119 |
0,
|
120 |
0
|
121 |
],
|
122 |
+
"n_vocab": 79,
|
123 |
"name": "AudioSamples",
|
124 |
"nb_priors": 3,
|
125 |
"pos_init": false,
|
|
|
166 |
],
|
167 |
"prime_zero_out": false,
|
168 |
"priors_width": [
|
169 |
+
1920,
|
170 |
2048,
|
171 |
1024
|
172 |
],
|
|
|
196 |
"summary_proj_to_labels": true,
|
197 |
"summary_type": "cls_index",
|
198 |
"summary_use_proj": true,
|
199 |
+
"t_bins": 64,
|
200 |
"transformers_version": "4.20.0.dev0",
|
201 |
"use_cache": true,
|
202 |
"use_nonrelative_specloss": true,
|
|
|
209 |
"vq_vae_codebook_dimension": 2048,
|
210 |
"vq_vae_commit": 0.02,
|
211 |
"vq_vae_conv_block_depth": 4,
|
212 |
+
"vq_vae_conv_block_width": 32,
|
213 |
"vq_vae_depth": 4,
|
214 |
"vq_vae_dilation_cycle": null,
|
215 |
"vq_vae_dilation_growth_rate": 3,
|
|
|
218 |
2,
|
219 |
2
|
220 |
],
|
221 |
+
"vq_vae_emmbedding_width": 64,
|
222 |
"vq_vae_levels": 3,
|
223 |
"vq_vae_lmu": 0.99,
|
224 |
"vq_vae_m_conv": 1,
|
|
|
233 |
2,
|
234 |
2
|
235 |
],
|
236 |
+
"vq_vae_width": 32,
|
237 |
"width": [
|
238 |
+
2048,
|
239 |
1920,
|
240 |
+
1920
|
241 |
],
|
242 |
"y_bins": [
|
243 |
+
[
|
244 |
+
604,
|
245 |
+
7898
|
246 |
+
],
|
247 |
+
[
|
248 |
+
120,
|
249 |
+
4111
|
250 |
+
],
|
251 |
+
[
|
252 |
+
120,
|
253 |
+
4111
|
254 |
+
]
|
255 |
],
|
256 |
"zero_out": false
|
257 |
}
|