Skip to content

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

fix(paraformer): place cif_v1_export empty-fire guard after fire_idxs is defined

LauraGPT57f97776165c
diff --git a/funasr/models/paraformer/cif_predictor.py b/funasr/models/paraformer/cif_predictor.py
index 32f73a4fa..93645a8ff 100644
--- a/funasr/models/paraformer/cif_predictor.py
+++ b/funasr/models/paraformer/cif_predictor.py
@@ -616,15 +616,6 @@ def cif_v1_export(hidden, alphas, threshold: float):
device = hidden.device
dtype = hidden.dtype
batch_size, len_time, hidden_size = hidden.size()
- if fire_idxs.sum() == 0:
- # No integration boundaries fired (e.g. a very short / silent
- # segment): return zero frames instead of indexing into the
- # resulting empty tensors, which raised an IndexError.
- max_label_len = torch.round(alphas.sum(-1)).int().max()
- return (
- torch.zeros(batch_size, max_label_len, hidden_size, dtype=dtype, device=device),
- fires,
- )
threshold = torch.tensor([threshold], dtype=alphas.dtype).to(alphas.device)
frames = torch.zeros(batch_size, len_time, hidden_size, dtype=dtype, device=device)
@@ -645,6 +636,16 @@ def cif_v1_export(hidden, alphas, threshold: float):
fires[fire_idxs] = 1
fires = fires + prefix_sum - prefix_sum_floor
+ if fire_idxs.sum() == 0:
+ # No integration boundaries fired (e.g. a very short / silent segment):
+ # return zero frames instead of indexing into the resulting empty
+ # tensors, which raised an IndexError at `shift_frames[shift_batch_idxs]`.
+ # Placed after fire_idxs/fires are computed (TorchScript requires every
+ # referenced name to be defined first).
+ max_label_len = torch.floor(alphas.sum(dim=-1)).max().to(dtype=torch.int64)
+ frame_fires = torch.zeros(batch_size, max_label_len, hidden_size, dtype=dtype, device=device)
+ return frame_fires, fires
+
# prefix_sum_hidden = torch.cumsum(alphas.unsqueeze(-1).tile((1, 1, hidden_size)) * hidden, dim=1)
prefix_sum_hidden = torch.cumsum(alphas.unsqueeze(-1).repeat((1, 1, hidden_size)) * hidden, dim=1)
frames = prefix_sum_hidden[fire_idxs]