Upload model (#6)
Browse files- Upload model (449cc77fdbb4380d768678db8464fd2ffdb9a756)
- config.json +145 -0
- generation_config.json +290 -0
- pytorch_model.bin +3 -0
config.json
ADDED
@@ -0,0 +1,145 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"activation_dropout": 0.0,
|
3 |
+
"activation_function": "relu",
|
4 |
+
"adaptor_dropout": 0.1,
|
5 |
+
"adaptor_kernel_size": 8,
|
6 |
+
"adaptor_layer_norm": true,
|
7 |
+
"adaptor_stride": 8,
|
8 |
+
"add_adapter": true,
|
9 |
+
"architectures": [
|
10 |
+
"SeamlessM4TModel"
|
11 |
+
],
|
12 |
+
"attention_dropout": 0.1,
|
13 |
+
"bos_token_id": 2,
|
14 |
+
"conv_bias": false,
|
15 |
+
"conv_depthwise_kernel_size": 31,
|
16 |
+
"conv_dim": [
|
17 |
+
512,
|
18 |
+
512,
|
19 |
+
512,
|
20 |
+
512,
|
21 |
+
512,
|
22 |
+
512,
|
23 |
+
160
|
24 |
+
],
|
25 |
+
"conv_kernel": [
|
26 |
+
10,
|
27 |
+
3,
|
28 |
+
3,
|
29 |
+
3,
|
30 |
+
3,
|
31 |
+
2,
|
32 |
+
2
|
33 |
+
],
|
34 |
+
"conv_stride": [
|
35 |
+
5,
|
36 |
+
2,
|
37 |
+
2,
|
38 |
+
2,
|
39 |
+
2,
|
40 |
+
2,
|
41 |
+
2
|
42 |
+
],
|
43 |
+
"decoder_attention_heads": 16,
|
44 |
+
"decoder_ffn_dim": 4096,
|
45 |
+
"decoder_layerdrop": 0.05,
|
46 |
+
"decoder_layers": 12,
|
47 |
+
"decoder_start_token_id": 3,
|
48 |
+
"dropout": 0.1,
|
49 |
+
"encoder_attention_heads": 16,
|
50 |
+
"encoder_ffn_dim": 4096,
|
51 |
+
"encoder_layerdrop": 0.05,
|
52 |
+
"encoder_layers": 12,
|
53 |
+
"eos_token_id": 3,
|
54 |
+
"hidden_act": "gelu",
|
55 |
+
"hidden_size": 1024,
|
56 |
+
"init_std": 0.02,
|
57 |
+
"initializer_range": 0.02,
|
58 |
+
"is_encoder_decoder": true,
|
59 |
+
"lang_embed_dim": 256,
|
60 |
+
"layer_norm_eps": 1e-05,
|
61 |
+
"layerdrop": 0.1,
|
62 |
+
"leaky_relu_slope": 0.1,
|
63 |
+
"max_new_tokens": 256,
|
64 |
+
"max_position_embeddings": 4096,
|
65 |
+
"max_source_positions": 4096,
|
66 |
+
"model_in_dim": 1792,
|
67 |
+
"model_type": "seamless_m4t",
|
68 |
+
"num_adapter_layers": 1,
|
69 |
+
"num_attention_heads": 16,
|
70 |
+
"num_conv_pos_embedding_groups": 16,
|
71 |
+
"num_conv_pos_embeddings": 128,
|
72 |
+
"num_hidden_layers": 12,
|
73 |
+
"output_hidden_size": null,
|
74 |
+
"pad_token_id": 0,
|
75 |
+
"position_embeddings_type": "relative",
|
76 |
+
"resblock_dilation_sizes": [
|
77 |
+
[
|
78 |
+
1,
|
79 |
+
3,
|
80 |
+
5
|
81 |
+
],
|
82 |
+
[
|
83 |
+
1,
|
84 |
+
3,
|
85 |
+
5
|
86 |
+
],
|
87 |
+
[
|
88 |
+
1,
|
89 |
+
3,
|
90 |
+
5
|
91 |
+
]
|
92 |
+
],
|
93 |
+
"resblock_kernel_sizes": [
|
94 |
+
3,
|
95 |
+
7,
|
96 |
+
11
|
97 |
+
],
|
98 |
+
"rotary_embedding_base": 10000,
|
99 |
+
"sampling_rate": 16000,
|
100 |
+
"scale_embedding": true,
|
101 |
+
"speech_encoder_attention_heads": 16,
|
102 |
+
"speech_encoder_dropout": 0.0,
|
103 |
+
"speech_encoder_hidden_act": "swish",
|
104 |
+
"speech_encoder_intermediate_size": 4096,
|
105 |
+
"speech_encoder_layers": 12,
|
106 |
+
"spkr_embed_dim": 256,
|
107 |
+
"t2u_bos_token_id": 0,
|
108 |
+
"t2u_decoder_attention_heads": 16,
|
109 |
+
"t2u_decoder_ffn_dim": 8192,
|
110 |
+
"t2u_decoder_layers": 4,
|
111 |
+
"t2u_decoder_start_token_id": 2,
|
112 |
+
"t2u_encoder_attention_heads": 16,
|
113 |
+
"t2u_encoder_ffn_dim": 8192,
|
114 |
+
"t2u_encoder_layers": 4,
|
115 |
+
"t2u_eos_token_id": 2,
|
116 |
+
"t2u_max_new_tokens": 1024,
|
117 |
+
"t2u_num_langs": 38,
|
118 |
+
"t2u_pad_token_id": 1,
|
119 |
+
"torch_dtype": "float32",
|
120 |
+
"transformers_version": "4.33.0.dev0",
|
121 |
+
"unit_embed_dim": 1280,
|
122 |
+
"unit_hifi_gan_vocab_size": 10000,
|
123 |
+
"unit_vocab_size": 10082,
|
124 |
+
"upsample_initial_channel": 512,
|
125 |
+
"upsample_kernel_sizes": [
|
126 |
+
11,
|
127 |
+
8,
|
128 |
+
8,
|
129 |
+
4,
|
130 |
+
4
|
131 |
+
],
|
132 |
+
"upsample_rates": [
|
133 |
+
5,
|
134 |
+
4,
|
135 |
+
4,
|
136 |
+
2,
|
137 |
+
2
|
138 |
+
],
|
139 |
+
"use_cache": true,
|
140 |
+
"var_pred_dropout": 0.5,
|
141 |
+
"variance_predictor_kernel_size": 3,
|
142 |
+
"vocab_size": 256206,
|
143 |
+
"vocoder_num_langs": 36,
|
144 |
+
"vocoder_num_spkrs": 200
|
145 |
+
}
|
generation_config.json
ADDED
@@ -0,0 +1,290 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token_id": 2,
|
3 |
+
"decoder_start_token_id": 3,
|
4 |
+
"eos_token_id": 3,
|
5 |
+
"max_new_tokens": 256,
|
6 |
+
"pad_token_id": 0,
|
7 |
+
"t2u_lang_code_to_id": {
|
8 |
+
"arb": 0,
|
9 |
+
"ben": 1,
|
10 |
+
"cat": 2,
|
11 |
+
"ces": 3,
|
12 |
+
"cmn": 4,
|
13 |
+
"cym": 5,
|
14 |
+
"dan": 6,
|
15 |
+
"deu": 7,
|
16 |
+
"eng": 8,
|
17 |
+
"est": 9,
|
18 |
+
"fin": 10,
|
19 |
+
"fra": 11,
|
20 |
+
"hin": 12,
|
21 |
+
"ind": 13,
|
22 |
+
"ita": 14,
|
23 |
+
"jpn": 15,
|
24 |
+
"kan": 16,
|
25 |
+
"kor": 17,
|
26 |
+
"mlt": 18,
|
27 |
+
"nld": 19,
|
28 |
+
"pes": 20,
|
29 |
+
"pol": 21,
|
30 |
+
"por": 22,
|
31 |
+
"ron": 23,
|
32 |
+
"rus": 24,
|
33 |
+
"slk": 25,
|
34 |
+
"spa": 26,
|
35 |
+
"swe": 27,
|
36 |
+
"swh": 28,
|
37 |
+
"tam": 29,
|
38 |
+
"tel": 30,
|
39 |
+
"tgl": 31,
|
40 |
+
"tha": 32,
|
41 |
+
"tur": 33,
|
42 |
+
"ukr": 34,
|
43 |
+
"urd": 35,
|
44 |
+
"uzn": 36,
|
45 |
+
"vie": 37
|
46 |
+
},
|
47 |
+
"text_decoder_lang_to_code_id": {
|
48 |
+
"ace": 256001,
|
49 |
+
"ace_Latn": 256002,
|
50 |
+
"acm": 256003,
|
51 |
+
"acq": 256004,
|
52 |
+
"aeb": 256005,
|
53 |
+
"afr": 256006,
|
54 |
+
"ajp": 256007,
|
55 |
+
"aka": 256008,
|
56 |
+
"als": 256162,
|
57 |
+
"amh": 256009,
|
58 |
+
"apc": 256010,
|
59 |
+
"arb": 256011,
|
60 |
+
"ars": 256012,
|
61 |
+
"ary": 256013,
|
62 |
+
"arz": 256014,
|
63 |
+
"asm": 256015,
|
64 |
+
"ast": 256016,
|
65 |
+
"awa": 256017,
|
66 |
+
"ayr": 256018,
|
67 |
+
"azb": 256019,
|
68 |
+
"azj": 256020,
|
69 |
+
"bak": 256021,
|
70 |
+
"bam": 256022,
|
71 |
+
"ban": 256023,
|
72 |
+
"bel": 256024,
|
73 |
+
"bem": 256025,
|
74 |
+
"ben": 256026,
|
75 |
+
"bho": 256027,
|
76 |
+
"bjn": 256028,
|
77 |
+
"bjn_Latn": 256029,
|
78 |
+
"bod": 256030,
|
79 |
+
"bos": 256031,
|
80 |
+
"bug": 256032,
|
81 |
+
"bul": 256033,
|
82 |
+
"cat": 256034,
|
83 |
+
"ceb": 256035,
|
84 |
+
"ces": 256036,
|
85 |
+
"cjk": 256037,
|
86 |
+
"ckb": 256038,
|
87 |
+
"cmn": 256200,
|
88 |
+
"cmn_Hant": 256201,
|
89 |
+
"crh": 256039,
|
90 |
+
"cym": 256040,
|
91 |
+
"dan": 256041,
|
92 |
+
"deu": 256042,
|
93 |
+
"dik": 256043,
|
94 |
+
"dyu": 256044,
|
95 |
+
"dzo": 256045,
|
96 |
+
"ell": 256046,
|
97 |
+
"eng": 256047,
|
98 |
+
"epo": 256048,
|
99 |
+
"est": 256049,
|
100 |
+
"eus": 256050,
|
101 |
+
"ewe": 256051,
|
102 |
+
"fao": 256052,
|
103 |
+
"fij": 256054,
|
104 |
+
"fin": 256055,
|
105 |
+
"fon": 256056,
|
106 |
+
"fra": 256057,
|
107 |
+
"fur": 256058,
|
108 |
+
"fuv": 256059,
|
109 |
+
"gaz": 256135,
|
110 |
+
"gla": 256060,
|
111 |
+
"gle": 256061,
|
112 |
+
"glg": 256062,
|
113 |
+
"grn": 256063,
|
114 |
+
"guj": 256064,
|
115 |
+
"hat": 256065,
|
116 |
+
"hau": 256066,
|
117 |
+
"heb": 256067,
|
118 |
+
"hin": 256068,
|
119 |
+
"hne": 256069,
|
120 |
+
"hrv": 256070,
|
121 |
+
"hun": 256071,
|
122 |
+
"hye": 256072,
|
123 |
+
"ibo": 256073,
|
124 |
+
"ilo": 256074,
|
125 |
+
"ind": 256075,
|
126 |
+
"isl": 256076,
|
127 |
+
"ita": 256077,
|
128 |
+
"jav": 256078,
|
129 |
+
"jpn": 256079,
|
130 |
+
"kab": 256080,
|
131 |
+
"kac": 256081,
|
132 |
+
"kam": 256082,
|
133 |
+
"kan": 256083,
|
134 |
+
"kas": 256084,
|
135 |
+
"kas_Deva": 256085,
|
136 |
+
"kat": 256086,
|
137 |
+
"kaz": 256089,
|
138 |
+
"kbp": 256090,
|
139 |
+
"kea": 256091,
|
140 |
+
"khk": 256122,
|
141 |
+
"khm": 256092,
|
142 |
+
"kik": 256093,
|
143 |
+
"kin": 256094,
|
144 |
+
"kir": 256095,
|
145 |
+
"kmb": 256096,
|
146 |
+
"kmr": 256099,
|
147 |
+
"knc": 256087,
|
148 |
+
"knc_Latn": 256088,
|
149 |
+
"kon": 256097,
|
150 |
+
"kor": 256098,
|
151 |
+
"lao": 256100,
|
152 |
+
"lij": 256102,
|
153 |
+
"lim": 256103,
|
154 |
+
"lin": 256104,
|
155 |
+
"lit": 256105,
|
156 |
+
"lmo": 256106,
|
157 |
+
"ltg": 256107,
|
158 |
+
"ltz": 256108,
|
159 |
+
"lua": 256109,
|
160 |
+
"lug": 256110,
|
161 |
+
"luo": 256111,
|
162 |
+
"lus": 256112,
|
163 |
+
"lvs": 256101,
|
164 |
+
"mag": 256113,
|
165 |
+
"mai": 256114,
|
166 |
+
"mal": 256115,
|
167 |
+
"mar": 256116,
|
168 |
+
"min": 256117,
|
169 |
+
"mkd": 256118,
|
170 |
+
"mlt": 256120,
|
171 |
+
"mni": 256121,
|
172 |
+
"mos": 256123,
|
173 |
+
"mri": 256124,
|
174 |
+
"mya": 256126,
|
175 |
+
"nld": 256127,
|
176 |
+
"nno": 256128,
|
177 |
+
"nob": 256129,
|
178 |
+
"npi": 256130,
|
179 |
+
"nso": 256131,
|
180 |
+
"nus": 256132,
|
181 |
+
"nya": 256133,
|
182 |
+
"oci": 256134,
|
183 |
+
"ory": 256136,
|
184 |
+
"pag": 256137,
|
185 |
+
"pan": 256138,
|
186 |
+
"pap": 256139,
|
187 |
+
"pbt": 256143,
|
188 |
+
"pes": 256053,
|
189 |
+
"plt": 256119,
|
190 |
+
"pol": 256140,
|
191 |
+
"por": 256141,
|
192 |
+
"prs": 256142,
|
193 |
+
"quy": 256144,
|
194 |
+
"ron": 256145,
|
195 |
+
"run": 256146,
|
196 |
+
"rus": 256147,
|
197 |
+
"sag": 256148,
|
198 |
+
"san": 256149,
|
199 |
+
"sat": 256150,
|
200 |
+
"scn": 256151,
|
201 |
+
"shn": 256152,
|
202 |
+
"sin": 256153,
|
203 |
+
"slk": 256154,
|
204 |
+
"slv": 256155,
|
205 |
+
"smo": 256156,
|
206 |
+
"sna": 256157,
|
207 |
+
"snd": 256158,
|
208 |
+
"som": 256159,
|
209 |
+
"sot": 256160,
|
210 |
+
"spa": 256161,
|
211 |
+
"srd": 256163,
|
212 |
+
"srp": 256164,
|
213 |
+
"ssw": 256165,
|
214 |
+
"sun": 256166,
|
215 |
+
"swe": 256167,
|
216 |
+
"swh": 256168,
|
217 |
+
"szl": 256169,
|
218 |
+
"tam": 256170,
|
219 |
+
"taq": 256177,
|
220 |
+
"taq_Tfng": 256178,
|
221 |
+
"tat": 256171,
|
222 |
+
"tel": 256172,
|
223 |
+
"tgk": 256173,
|
224 |
+
"tgl": 256174,
|
225 |
+
"tha": 256175,
|
226 |
+
"tir": 256176,
|
227 |
+
"tpi": 256179,
|
228 |
+
"tsn": 256180,
|
229 |
+
"tso": 256181,
|
230 |
+
"tuk": 256182,
|
231 |
+
"tum": 256183,
|
232 |
+
"tur": 256184,
|
233 |
+
"twi": 256185,
|
234 |
+
"tzm": 256186,
|
235 |
+
"uig": 256187,
|
236 |
+
"ukr": 256188,
|
237 |
+
"umb": 256189,
|
238 |
+
"urd": 256190,
|
239 |
+
"uzn": 256191,
|
240 |
+
"vec": 256192,
|
241 |
+
"vie": 256193,
|
242 |
+
"war": 256194,
|
243 |
+
"wol": 256195,
|
244 |
+
"xho": 256196,
|
245 |
+
"ydd": 256197,
|
246 |
+
"yor": 256198,
|
247 |
+
"yue": 256199,
|
248 |
+
"zsm": 256125,
|
249 |
+
"zul": 256202
|
250 |
+
},
|
251 |
+
"transformers_version": "4.33.0.dev0",
|
252 |
+
"vocoder_lang_code_to_id": {
|
253 |
+
"arb": 0,
|
254 |
+
"ben": 1,
|
255 |
+
"cat": 2,
|
256 |
+
"ces": 3,
|
257 |
+
"cmn": 4,
|
258 |
+
"cym": 5,
|
259 |
+
"dan": 6,
|
260 |
+
"deu": 7,
|
261 |
+
"eng": 8,
|
262 |
+
"est": 9,
|
263 |
+
"fin": 10,
|
264 |
+
"fra": 11,
|
265 |
+
"hin": 12,
|
266 |
+
"ind": 13,
|
267 |
+
"ita": 14,
|
268 |
+
"jpn": 15,
|
269 |
+
"kor": 16,
|
270 |
+
"mlt": 17,
|
271 |
+
"nld": 18,
|
272 |
+
"pes": 19,
|
273 |
+
"pol": 20,
|
274 |
+
"por": 21,
|
275 |
+
"ron": 22,
|
276 |
+
"rus": 23,
|
277 |
+
"slk": 24,
|
278 |
+
"spa": 25,
|
279 |
+
"swe": 26,
|
280 |
+
"swh": 27,
|
281 |
+
"tel": 28,
|
282 |
+
"tgl": 29,
|
283 |
+
"tha": 30,
|
284 |
+
"tur": 31,
|
285 |
+
"ukr": 32,
|
286 |
+
"urd": 33,
|
287 |
+
"uzn": 34,
|
288 |
+
"vie": 35
|
289 |
+
}
|
290 |
+
}
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:395d43577e8eef756f27da4912d578b10f23a93a79184893cfb534e392ec7ad2
|
3 |
+
size 4838401068
|