Merge pull request #4 from TeaPoly/main

zhifu gao

2022-12-02 419533aa3abedc26cd31c34276973d71b8b20f2e

Merge pull request #4 from TeaPoly/main

Fix some issue to make batch inference easy for predictor and decoder.

 funasr/bin/asr_inference_paraformer.py

@@ -189,7 +189,7 @@

        predictor_outs = self.asr_model.calc_predictor(enc, enc_len)
        pre_acoustic_embeds, pre_token_length = predictor_outs[0], predictor_outs[1]
        pre_token_length = torch.tensor([pre_acoustic_embeds.size(1)], device=pre_acoustic_embeds.device)
        pre_token_length = pre_token_length.long()
        decoder_outs = self.asr_model.cal_decoder_with_predictor(enc, enc_len, pre_acoustic_embeds, pre_token_length)
        decoder_out, ys_pad_lens = decoder_outs[0], decoder_outs[1]


 funasr/models/predictor/cif.py

@@ -147,7 +147,7 @@
        b, t, d = hidden.size()

        tail_threshold = self.tail_threshold

        tail_threshold = torch.tensor([tail_threshold], dtype=alphas.dtype).to(alphas.device)

        tail_threshold = torch.reshape(tail_threshold, (1, 1))

        tail_threshold = tail_threshold.unsqueeze(0).repeat(b, 1)

        alphas = torch.cat([alphas, tail_threshold], dim=1)

        zeros = torch.zeros((b, 1, d), dtype=hidden.dtype).to(hidden.device)

        hidden = torch.cat([hidden, zeros], dim=1)