funasr/models/pooling/statistic_pooling.py
@@ -82,10 +82,10 @@ tt = xs_pad.shape[2] num_chunk = int(math.ceil(tt / pooling_stride)) pad = pooling_size // 2 if xs_pad.shape == 4: features = F.pad(xs_pad, (0, 0, pad, pad), "reflect") if len(xs_pad.shape) == 4: features = F.pad(xs_pad, (0, 0, pad, pad), "replicate") else: features = F.pad(xs_pad, (pad, pad), "reflect") features = F.pad(xs_pad, (pad, pad), "replicate") stat_list = [] for i in range(num_chunk):
