kamilakesbi commited on
Commit
34759c9
1 Parent(s): c12d49f

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +46 -1
README.md CHANGED
@@ -27,7 +27,52 @@ It achieves the following results on the evaluation set:
27
 
28
  ## Model description
29
 
30
- More information needed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
31
 
32
  ## Intended uses & limitations
33
 
 
27
 
28
  ## Model description
29
 
30
+ This segmentation model has been trained on Chinese data (Callhome) using [diarizers](https://github.com/huggingface/diarizers/tree/main).
31
+ It can be loaded with two lines of code:
32
+
33
+ ```python
34
+ from diarizers import SegmentationModel
35
+
36
+ segmentation_model = SegmentationModel().from_pretrained('diarizers-community/speaker-segmentation-fine-tuned-callhome-zho')
37
+ ```
38
+
39
+ To use it within a pyannote speaker diarization pipeline, load the [pyannote/speaker-diarization-3.1](https://huggingface.co/pyannote/speaker-diarization-3.1) pipeline, and convert the model to a pyannote compatible format:
40
+
41
+ ```python
42
+
43
+ from pyannote.audio import Pipeline
44
+ import torch
45
+
46
+ device = torch.device("cuda:0") if torch.cuda.is_available() else torch.device("cpu")
47
+
48
+
49
+ # load the pre-trained pyannote pipeline
50
+ pipeline = Pipeline.from_pretrained("pyannote/speaker-diarization-3.1")
51
+ pipeline.to(device)
52
+
53
+ # replace the segmentation model with your fine-tuned one
54
+ segmentation_model = segmentation_model.to_pyannote_model()
55
+ pipeline._segmentation.model = model.to(device)
56
+ ```
57
+
58
+ You can now use the pipeline on audio examples:
59
+
60
+ ```python
61
+ # load dataset example
62
+ dataset = load_dataset("diarizers-community/callhome", "jpn", split="data")
63
+ sample = dataset[0]["audio"]
64
+
65
+ # pre-process inputs
66
+ sample["waveform"] = torch.from_numpy(sample.pop("array")[None, :]).to(device, dtype=model.dtype)
67
+ sample["sample_rate"] = sample.pop("sampling_rate")
68
+
69
+ # perform inference
70
+ diarization = pipeline(sample)
71
+
72
+ # dump the diarization output to disk using RTTM format
73
+ with open("audio.rttm", "w") as rttm:
74
+ diarization.write_rttm(rttm)
75
+ ```
76
 
77
  ## Intended uses & limitations
78