Update tiny models for Nemotron3DiarizationModel
Browse files- config.json +55 -0
- model.safetensors +3 -0
- preprocessor_config.json +12 -0
config.json
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Nemotron3DiarizationModel"
|
| 4 |
+
],
|
| 5 |
+
"audio_config": {
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"dropout": 0.0,
|
| 8 |
+
"hidden_act": "gelu",
|
| 9 |
+
"hidden_size": 32,
|
| 10 |
+
"initializer_range": 0.02,
|
| 11 |
+
"intermediate_size": 32,
|
| 12 |
+
"max_position_embeddings": 5000,
|
| 13 |
+
"model_type": "nemotron3_diarization_audio",
|
| 14 |
+
"num_attention_heads": 2,
|
| 15 |
+
"num_hidden_layers": 2,
|
| 16 |
+
"num_key_value_heads": 2,
|
| 17 |
+
"num_mel_bins": 8,
|
| 18 |
+
"partial_rotary_factor": 1.0,
|
| 19 |
+
"rope_parameters": {
|
| 20 |
+
"partial_rotary_factor": 1.0,
|
| 21 |
+
"rope_theta": 10000.0,
|
| 22 |
+
"rope_type": "default"
|
| 23 |
+
},
|
| 24 |
+
"subsampling_factor": 8
|
| 25 |
+
},
|
| 26 |
+
"chunk_length": 64,
|
| 27 |
+
"chunk_right_context": 0,
|
| 28 |
+
"dtype": "float32",
|
| 29 |
+
"fifo_length": 4,
|
| 30 |
+
"head_config": {
|
| 31 |
+
"audio_hidden_size": 32,
|
| 32 |
+
"hidden_size": 8,
|
| 33 |
+
"model_type": "",
|
| 34 |
+
"num_speakers": 4,
|
| 35 |
+
"subsampling_factor": 8
|
| 36 |
+
},
|
| 37 |
+
"initializer_range": 0.02,
|
| 38 |
+
"model_type": "nemotron3_diarization",
|
| 39 |
+
"speaker_cache_update_period": 4,
|
| 40 |
+
"streaming_config": {
|
| 41 |
+
"fifo_length": 4,
|
| 42 |
+
"latest_frames_score_boost": 0.05,
|
| 43 |
+
"min_positive_scores_rate": 0.5,
|
| 44 |
+
"model_type": "",
|
| 45 |
+
"num_speakers": 4,
|
| 46 |
+
"prediction_score_threshold": 0.25,
|
| 47 |
+
"speaker_cache_length": 8,
|
| 48 |
+
"speaker_cache_silence_frames_per_speaker": 1,
|
| 49 |
+
"speaker_cache_update_period": 4,
|
| 50 |
+
"strong_boost_rate": 0.75,
|
| 51 |
+
"subsampling_factor": 8,
|
| 52 |
+
"weak_boost_rate": 1.5
|
| 53 |
+
},
|
| 54 |
+
"transformers_version": "5.19.0.dev0"
|
| 55 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8785cef4bbd5597262645a898b0116a769fb342c3fdb6709c87cd0c51ae2aadc
|
| 3 |
+
size 70536
|
preprocessor_config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"feature_extractor_type": "NemotronAsrStreamingFeatureExtractor",
|
| 3 |
+
"feature_size": 128,
|
| 4 |
+
"hop_length": 160,
|
| 5 |
+
"n_fft": 512,
|
| 6 |
+
"padding_side": "right",
|
| 7 |
+
"padding_value": 0.0,
|
| 8 |
+
"preemphasis": 0.97,
|
| 9 |
+
"return_attention_mask": true,
|
| 10 |
+
"sampling_rate": 16000,
|
| 11 |
+
"win_length": 400
|
| 12 |
+
}
|