Skip to content

Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.

Dev gzf exp (#1664)

zhifu gao8fdc372c81ac
diff --git a/funasr/datasets/sense_voice_datasets/datasets.py b/funasr/datasets/sense_voice_datasets/datasets.py
index 4f14b35f9..6a79e7541 100644
--- a/funasr/datasets/sense_voice_datasets/datasets.py
+++ b/funasr/datasets/sense_voice_datasets/datasets.py
@@ -99,8 +99,9 @@ class SenseVoiceDataset(torch.utils.data.Dataset):
target_mask = (
[0] * (prompt_ids_len) + [1] * (target_ids_len) + [1]
) # [sos, task, lid, text, eos]: [0, 0, 1, 1, 1]
+ target_mask_lengths = len(target_mask)
target_mask = torch.tensor(target_mask, dtype=torch.float32)
-
+ target_mask_lengths = torch.tensor([target_mask_lengths], dtype=torch.int32)
return {
"speech": speech[0, :, :],
"speech_lengths": speech_lengths,
@@ -130,30 +131,26 @@ class SenseVoiceDataset(torch.utils.data.Dataset):
)
if self.batch_type != "example":
- b, t, _ = outputs["speech"].shape
- if b * t > self.batch_size:
- beg = torch.randint(0, 2, ()).item()
- logging.info(
- f"Warning, b * t: {b * t} > {self.batch_size}, drop half data 1st, beg:{beg}"
- )
- for key, data_list in outputs.items():
- outputs[key] = outputs[key][beg : beg + b : 2]
-
- b, t, _ = outputs["speech"].shape
- if b * t > self.batch_size:
- beg = torch.randint(0, 2, ()).item()
- logging.info(
- f"Warning, b * t: {b * t} > {self.batch_size}, drop half data 2nd, beg:{beg}"
- )
- for key, data_list in outputs.items():
- outputs[key] = outputs[key][beg : beg + b : 2]
-
- b, t, _ = outputs["speech"].shape
- if b * t > self.batch_size:
- beg = torch.randint(0, 2, ()).item()
- logging.info(
- f"Warning, b * t: {b * t} > {self.batch_size}, drop half data 3th, beg:{beg}"
- )
- for key, data_list in outputs.items():
- outputs[key] = outputs[key][beg : beg + b : 2]
+ for i in range(3):
+ outputs = self._filter_badcase(outputs)
+
+ return outputs
+
+ def _filter_badcase(self, outputs, i=0):
+ b, t, _ = outputs["speech"].shape
+ if b * t > self.batch_size:
+ beg = torch.randint(0, 2, ()).item()
+ logging.info(
+ f"Warning, b * t: {b * t} > {self.batch_size}, drop half data {i}th, beg:{beg}"
+ )
+ for key, data_list in outputs.items():
+ outputs[key] = outputs[key][beg : beg + b : 2]
+
+ speech_lengths_max = outputs["speech_lengths_max"].max().item()
+ outputs["speech"] = outputs["speech"][:, :speech_lengths_max, :]
+ text_lengths_max = outputs["text_lengths"].max().item()
+ outputs["text"] = outputs["text"][:, :text_lengths_max]
+ target_mask_lengths_max = outputs["target_mask_lengths_max"].max().item()
+ outputs["target_mask"] = outputs["target_mask"][:, :target_mask_lengths_max]
+
return outputs