Upload params.json with huggingface_hub
Browse files- params.json +34 -0
params.json
ADDED
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"dim": 3072,
|
3 |
+
"n_layers": 30,
|
4 |
+
"head_dim": 128,
|
5 |
+
"hidden_dim": 8192,
|
6 |
+
"n_heads": 32,
|
7 |
+
"n_kv_heads": 8,
|
8 |
+
"rope_theta": 100000000.0,
|
9 |
+
"norm_eps": 1e-05,
|
10 |
+
"vocab_size": 131072,
|
11 |
+
"max_position_embeddings": 32768,
|
12 |
+
"multimodal": {
|
13 |
+
"whisper_model_args": {
|
14 |
+
"encoder_args": {
|
15 |
+
"dim": 1280,
|
16 |
+
"n_layers": 32,
|
17 |
+
"head_dim": 64,
|
18 |
+
"hidden_dim": 5120,
|
19 |
+
"n_heads": 20,
|
20 |
+
"vocab_size": 51866,
|
21 |
+
"max_source_positions": 1500,
|
22 |
+
"audio_encoding_args": {
|
23 |
+
"sampling_rate": 16000,
|
24 |
+
"num_mel_bins": 128,
|
25 |
+
"hop_length": 160,
|
26 |
+
"window_size": 400
|
27 |
+
}
|
28 |
+
},
|
29 |
+
"downsample_args": {
|
30 |
+
"downsample_factor": 4
|
31 |
+
}
|
32 |
+
}
|
33 |
+
}
|
34 |
+
}
|