Upload 2 files
Browse files
16TokensPenultimateLayer.ckpt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:25b3379e311d44fd2b0cc4957ef2526960ab805dfb44d5cd2eb4aefd19079ea3
|
| 3 |
+
size 12933175758
|
16TokensPenultimateLayer.json
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"audio_encoder": {
|
| 3 |
+
"embed_dim_audio": 1024,
|
| 4 |
+
"joint_dim": 512,
|
| 5 |
+
"audio_cfg": {
|
| 6 |
+
"audio_length": 1024,
|
| 7 |
+
"clip_samples": 480000,
|
| 8 |
+
"mel_bins": 64,
|
| 9 |
+
"sample_rate": 48000,
|
| 10 |
+
"window_size": 1024,
|
| 11 |
+
"hop_size": 480,
|
| 12 |
+
"fmin": 50,
|
| 13 |
+
"fmax": 14000,
|
| 14 |
+
"class_num": 527,
|
| 15 |
+
"model_type": "HTSAT",
|
| 16 |
+
"model_name": "base"
|
| 17 |
+
}
|
| 18 |
+
},
|
| 19 |
+
"LLM": {
|
| 20 |
+
"t5_model": "google/flan-t5-xl",
|
| 21 |
+
"repetition_penalty": 10.0
|
| 22 |
+
},
|
| 23 |
+
"num_query_tokens": 16,
|
| 24 |
+
"embed_dim": 256,
|
| 25 |
+
"max_txt_len": 120,
|
| 26 |
+
"prompt": "Caption this data:",
|
| 27 |
+
"apply_lemmatizer": false,
|
| 28 |
+
"atRandom": true
|
| 29 |
+
}
|