mirror of
https://github.com/pyannote/pyannote-audio.git
synced 2026-11-20 13:07:51 +09:00
* related bugs: #1119 #1128 #1130 * related discussions: #1123 #1103 #1126 #1121
191 KiB
191 KiB
In [1]:
# preparing notebook for visualization purposes
# (only show outputs between t=0s and t=30s)
from pyannote.core import notebook, Segment
notebook.crop = Segment(0, 30)In [2]:
# clone pyannote-audio Github repository and update ROOT_DIR accordingly
ROOT_DIR = "/Users/hbredin/Development/pyannote/pyannote-audio"
AUDIO_FILE = f"{ROOT_DIR}/tutorials/assets/sample.wav"In [3]:
from pyannote.database.util import load_rttm
REFERENCE = f"{ROOT_DIR}/tutorials/assets/sample.rttm"
reference = load_rttm(REFERENCE)["sample"]
referenceOut [3]:
In [16]:
from huggingface_hub import HfApi
available_models = [m.modelId for m in HfApi().list_models(filter="pyannote-audio-model")]
list(filter(lambda p: p.startswith("pyannote/"), available_models))Out [16]:
['pyannote/Segmentation-PyanNet-DIHARD', 'pyannote/TestModelForContinuousIntegration', 'pyannote/embedding', 'pyannote/segmentation', 'pyannote/brouhaha']
In [5]:
from huggingface_hub import notebook_login
notebook_login()VBox(children=(HTML(value='<center> <img\nsrc=https://huggingface.co/front/assets/huggingface_logo-noborder.sv…
In [6]:
from pyannote.audio import Model
model = Model.from_pretrained("pyannote/segmentation", use_auth_token=True)In [7]:
model.summarize()Out [7]:
| Name | Type | Params | In sizes | Out sizes -------------------------------------------------------------------------------------------------------- 0 | sincnet | SincNet | 42.6 K | [3, 1, 32000] | [3, 60, 115] 1 | lstm | LSTM | 1.4 M | [3, 115, 60] | [[3, 115, 256], [[8, 3, 128], [8, 3, 128]]] 2 | linear | ModuleList | 49.4 K | ? | ? 3 | classifier | Linear | 387 | [3, 115, 128] | [3, 115, 3] 4 | activation | Sigmoid | 0 | [3, 115, 3] | [3, 115, 3] -------------------------------------------------------------------------------------------------------- 1.5 M Trainable params 0 Non-trainable params 1.5 M Total params 5.891 Total estimated model params size (MB)
In [8]:
specs = model.specifications
specsOut [8]:
Specifications(problem=<Problem.MULTI_LABEL_CLASSIFICATION: 2>, resolution=<Resolution.FRAME: 1>, duration=5.0, warm_up=(0.0, 0.0), classes=['speaker#1', 'speaker#2', 'speaker#3'], permutation_invariant=True)
In [9]:
from pyannote.audio import Inference
inference = Inference(model, step=2.5)
output = inference(AUDIO_FILE)
outputOut [9]:
In [10]:
output.data.shapeOut [10]:
(11, 293, 3)
In [11]:
import torchaudio
waveform, sample_rate = torchaudio.load(AUDIO_FILE)
print(f"{type(waveform)=}")
print(f"{waveform.shape=}")
print(f"{waveform.dtype=}")
audio_in_memory = {"waveform": waveform, "sample_rate": sample_rate}type(waveform)=<class 'torch.Tensor'> waveform.shape=torch.Size([1, 480000]) waveform.dtype=torch.float32
In [12]:
output = inference(audio_in_memory)
outputOut [12]:
In [13]:
from pyannote.core import Segment
output = inference.crop(AUDIO_FILE, Segment(10, 20))
outputOut [13]:
In [14]:
# look ma: no hands!
offline_model = Model.from_pretrained("pytorch_model.bin")In [15]:
# just checking weights are the same...
import torch
for weights, offline_weights in zip(model.parameters(), offline_model.parameters()):
assert torch.equal(weights, offline_weights)