wanchichen
commited on
Commit
•
f1f16d2
1
Parent(s):
1db6e2f
Update README.md
Browse files
README.md
CHANGED
@@ -4091,10 +4091,10 @@ xeus_model, xeus_train_args = SSLTask.build_model_from_file(
|
|
4091 |
|
4092 |
wavs, sampling_rate = sf.read('/path/to/audio.wav') # sampling rate should be 16000
|
4093 |
wav_lengths = torch.LongTensor([len(wav) for wav in [wavs]]).to(device)
|
4094 |
-
wavs = pad_sequence([wavs], batch_first=True).to(device)
|
4095 |
|
4096 |
# we recommend use_mask=True during fine-tuning
|
4097 |
-
feats = xeus_model.encode(wavs, wav_lengths, use_mask=False, use_final_output=False)[0][-1] # take the output of the last layer
|
4098 |
```
|
4099 |
|
4100 |
With Flash Attention:
|
|
|
4091 |
|
4092 |
wavs, sampling_rate = sf.read('/path/to/audio.wav') # sampling rate should be 16000
|
4093 |
wav_lengths = torch.LongTensor([len(wav) for wav in [wavs]]).to(device)
|
4094 |
+
wavs = pad_sequence(torch.Tensor([wavs]), batch_first=True).to(device)
|
4095 |
|
4096 |
# we recommend use_mask=True during fine-tuning
|
4097 |
+
feats = xeus_model.encode(wavs, wav_lengths, use_mask=False, use_final_output=False)[0][-1] # take the output of the last layer -> batch_size x seq_len x hdim
|
4098 |
```
|
4099 |
|
4100 |
With Flash Attention:
|