| | |
|---|
| | | hotword_list_or_file = None |
|---|
| | | clas_scale = 1.0 |
|---|
| | | |
|---|
| | | if kwargs.get("device", None) == "cpu": |
|---|
| | | ngpu = 0 |
|---|
| | | if ngpu >= 1 and torch.cuda.is_available(): |
|---|
| | | device = "cuda" |
|---|
| | | else: |
|---|
| | |
|---|
| | | hotword_list_or_file = kwargs['hotword'] |
|---|
| | | |
|---|
| | | speech2vadsegment.vad_model.vad_opts.max_single_segment_time = kwargs.get("max_single_segment_time", 60000) |
|---|
| | | batch_size_token_threshold_s = kwargs.get("batch_size_token_threshold_s", int(speech2vadsegment.vad_model.vad_opts.max_single_segment_time*0.67/1000)) * 1000 |
|---|
| | | batch_size_token = kwargs.get("batch_size_token", 6000) |
|---|
| | | print("batch_size_token: ", batch_size_token) |
|---|
| | | |
|---|
| | |
|---|
| | | beg_idx = 0 |
|---|
| | | for j, _ in enumerate(range(0, n)): |
|---|
| | | batch_size_token_ms_cum += (sorted_data[j][0][1] - sorted_data[j][0][0]) |
|---|
| | | if j < n - 1 and (batch_size_token_ms_cum + sorted_data[j + 1][0][1] - sorted_data[j + 1][0][0]) < batch_size_token_ms and (sorted_data[j + 1][0][1] - sorted_data[j + 1][0][0]) < speech2vadsegment.vad_model.vad_opts.max_single_segment_time: |
|---|
| | | if j < n - 1 and (batch_size_token_ms_cum + sorted_data[j + 1][0][1] - sorted_data[j + 1][0][0]) < batch_size_token_ms and (sorted_data[j + 1][0][1] - sorted_data[j + 1][0][0]) < batch_size_token_threshold_s: |
|---|
| | | continue |
|---|
| | | batch_size_token_ms_cum = 0 |
|---|
| | | end_idx = j + 1 |
|---|
