| | |
|---|
| | | # fp16 |
|---|
| | | if kwargs.get("fp16", False): |
|---|
| | | model.to(torch.float16) |
|---|
| | | elif kwargs.get("bf16", False): |
|---|
| | | model.to(torch.bfloat16) |
|---|
| | | return model, kwargs |
|---|
| | | |
|---|
| | | def __call__(self, *args, **cfg): |
|---|
| | |
|---|
| | | ): |
|---|
| | | max_len_in_batch = max(max_len_in_batch, sample_length) |
|---|
| | | end_idx += 1 |
|---|
| | | results_sorted.append({'key': 'bad_data', 'text': '', 'timestamp': []}) |
|---|
| | | continue |
|---|
| | | |
|---|
| | | speech_j, speech_lengths_j = slice_padding_audio_samples( |
|---|
| | |
|---|
| | | end_idx += 1 |
|---|
| | | max_len_in_batch = sample_length |
|---|
| | | if len(results) < 1: |
|---|
| | | results.append({'key': 'bad_data', 'text': '', 'timestamp': []}) |
|---|
| | | continue |
|---|
| | | results_sorted.extend(results) |
|---|
| | | |
|---|
| | | # end_asr_total = time.time() |
|---|
| | |
|---|
| | | # f"time_speech_total_per_sample: {time_speech_total_per_sample: 0.3f}, " |
|---|
| | | # f"time_escape_total_per_sample: {time_escape_total_per_sample:0.3f}") |
|---|
| | | |
|---|
| | | if len(results_sorted) != n: |
|---|
| | | results_ret_list.append({"key": key, "text": "", "timestamp": []}) |
|---|
| | | logging.info("decoding, utt: {}, empty result".format(key)) |
|---|
| | | continue |
|---|
| | | restored_data = [0] * n |
|---|
| | | for j in range(n): |
|---|
| | | index = sorted_data[j][1] |
|---|
