feat: Support DPO

cosyvoice/llm/llm_dpo.py

@@ -0,0 +1,556 @@
+# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu, Zhihao Du)
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from typing import Dict, Optional, Callable, List, Generator
+import torch
+from torch import nn
+import torch.nn.functional as F
+from transformers import Qwen2ForCausalLM
+from torch.nn.utils.rnn import pad_sequence, unpad_sequence
+from cosyvoice.utils.common import IGNORE_ID
+from cosyvoice.transformer.label_smoothing_loss import LabelSmoothingLoss
+from cosyvoice.utils.common import th_accuracy
+from cosyvoice.utils.file_utils import logging
+from cosyvoice.utils.mask import make_pad_mask
+
+
+class TransformerLM(torch.nn.Module):
+    def __init__(
+            self,
+            text_encoder_input_size: int,
+            llm_input_size: int,
+            llm_output_size: int,
+            text_token_size: int,
+            speech_token_size: int,
+            text_encoder: torch.nn.Module,
+            llm: torch.nn.Module,
+            sampling: Callable,
+            length_normalized_loss: bool = True,
+            lsm_weight: float = 0.0,
+            spk_embed_dim: int = 192,
+    ):
+        super().__init__()
+        self.llm_input_size = llm_input_size
+        self.speech_token_size = speech_token_size
+        # 1. build text token inputs related modules
+        self.text_embedding = torch.nn.Embedding(text_token_size, text_encoder_input_size)
+        self.text_encoder = text_encoder
+        self.text_encoder_affine_layer = nn.Linear(
+            self.text_encoder.output_size(),
+            llm_input_size
+        )
+
+        # 2. build speech token language model related modules
+        self.sos_eos = 0
+        self.task_id = 1
+        self.llm_embedding = torch.nn.Embedding(2, llm_input_size)
+        self.llm = llm
+        self.llm_decoder = nn.Linear(llm_output_size, speech_token_size + 1)
+        self.criterion_ce = LabelSmoothingLoss(
+            size=speech_token_size + 1,
+            padding_idx=IGNORE_ID,
+            smoothing=lsm_weight,
+            normalize_length=length_normalized_loss,
+        )
+
+        # 3. [Optional] build speech token related modules
+        self.speech_embedding = torch.nn.Embedding(speech_token_size, llm_input_size)
+        self.spk_embed_affine_layer = torch.nn.Linear(spk_embed_dim, llm_input_size)
+
+        # 4. sampling method
+        self.sampling = sampling
+
+    def encode(
+            self,
+            text: torch.Tensor,
+            text_lengths: torch.Tensor,
+    ):
+        encoder_out, encoder_mask = self.text_encoder(text, text_lengths, decoding_chunk_size=1, num_decoding_left_chunks=-1)
+        encoder_out_lens = encoder_mask.squeeze(1).sum(1)
+        encoder_out = self.text_encoder_affine_layer(encoder_out)
+        return encoder_out, encoder_out_lens
+
+    def pad_unpad_sequence(self, sos_eos_emb, embedding, text_token, text_token_len, task_id_emb, speech_token, speech_token_len):
+        text_token = unpad_sequence(text_token, text_token_len.cpu(), batch_first=True)
+        speech_token = unpad_sequence(speech_token, speech_token_len.cpu(), batch_first=True)
+        lm_input = [torch.concat([sos_eos_emb.squeeze(dim=0), embedding[i], text_token[i], task_id_emb.squeeze(dim=0), speech_token[i]], dim=0)
+                    for i in range(len(text_token))]
+        lm_input_len = torch.tensor([i.size(0) for i in lm_input], dtype=torch.int32)
+        lm_input = pad_sequence(lm_input, batch_first=True, padding_value=IGNORE_ID)
+        return lm_input, lm_input_len
+
+    def forward(
+            self,
+            batch: dict,
+            device: torch.device,
+    ) -> Dict[str, Optional[torch.Tensor]]:
+        """
+        Args:
+            text: (B, L, D)
+            text_lengths: (B,)
+            audio: (B, T, N) or (B, T)
+            audio_lengths: (B,)
+        """
+        text_token = batch['text_token'].to(device)
+        text_token_len = batch['text_token_len'].to(device)
+        speech_token = batch['speech_token'].to(device)
+        speech_token_len = batch['speech_token_len'].to(device)
+        embedding = batch['embedding'].to(device)
+
+        # 1. prepare llm_target
+        lm_target = [torch.tensor([IGNORE_ID] * (2 + text_token_len[i]) + speech_token[i, :speech_token_len[i]].tolist() +
+                                  [self.speech_token_size]) for i in range(text_token.size(0))]
+        lm_target = pad_sequence(lm_target, batch_first=True, padding_value=IGNORE_ID).to(device)
+
+        # 1. encode text_token
+        text_token = self.text_embedding(text_token)
+        text_token, text_token_len = self.encode(text_token, text_token_len)
+
+        # 2. embedding projection
+        embedding = F.normalize(embedding, dim=1)
+        embedding = self.spk_embed_affine_layer(embedding)
+        embedding = embedding.unsqueeze(1)
+
+        # 3. eos and task_id
+        sos_eos_emb = self.llm_embedding.weight[self.sos_eos].reshape(1, 1, -1)
+        task_id_emb = self.llm_embedding.weight[self.task_id].reshape(1, 1, -1)
+
+        # 4. encode speech_token
+        speech_token = self.speech_embedding(speech_token)
+
+        # 5. unpad and pad
+        lm_input, lm_input_len = self.pad_unpad_sequence(sos_eos_emb, embedding, text_token, text_token_len,
+                                                         task_id_emb, speech_token, speech_token_len)
+
+        # 6. run lm forward
+        lm_output, lm_output_mask = self.llm(lm_input, lm_input_len.to(device))
+        logits = self.llm_decoder(lm_output)
+        loss = self.criterion_ce(logits, lm_target)
+        acc = th_accuracy(logits.view(-1, self.speech_token_size + 1), lm_target, ignore_label=IGNORE_ID)
+        return {'loss': loss, 'acc': acc}
+
+    def sampling_ids(
+            self,
+            weighted_scores: torch.Tensor,
+            decoded_tokens: List,
+            sampling: int,
+            ignore_eos: bool = True,
+    ):
+        num_trials, max_trials = 0, 100
+        while True:
+            top_ids = self.sampling(weighted_scores, decoded_tokens, sampling)
+            if (not ignore_eos) or (self.speech_token_size not in top_ids):
+                break
+            num_trials += 1
+            if num_trials > max_trials:
+                raise RuntimeError('sampling reaches max_trials {} and still get eos when ignore_eos is True, check your input!'.format(max_trials))
+        return top_ids
+
+    @torch.inference_mode()
+    def inference(
+            self,
+            text: torch.Tensor,
+            text_len: torch.Tensor,
+            prompt_text: torch.Tensor,
+            prompt_text_len: torch.Tensor,
+            prompt_speech_token: torch.Tensor,
+            prompt_speech_token_len: torch.Tensor,
+            embedding: torch.Tensor,
+            sampling: int = 25,
+            max_token_text_ratio: float = 20,
+            min_token_text_ratio: float = 2,
+    ) -> Generator[torch.Tensor, None, None]:
+        if self.fp16 is True:
+            embedding = embedding.half()
+
+        device = text.device
+        text = torch.concat([prompt_text, text], dim=1)
+        text_len += prompt_text_len
+        text = self.text_embedding(text)
+
+        # 1. encode text
+        text, text_len = self.encode(text, text_len)
+
+        # 2. encode embedding
+        if embedding.shape[0] != 0:
+            embedding = F.normalize(embedding, dim=1)
+            embedding = self.spk_embed_affine_layer(embedding)
+            embedding = embedding.unsqueeze(dim=1)
+        else:
+            embedding = torch.zeros(1, 0, self.llm_input_size, dtype=text.dtype).to(device).to(text.dtype)
+
+        # 3. concat llm_input
+        sos_eos_emb = self.llm_embedding.weight[self.sos_eos].reshape(1, 1, -1)
+        task_id_emb = self.llm_embedding.weight[self.task_id].reshape(1, 1, -1)
+        if prompt_speech_token_len != 0:
+            prompt_speech_token_emb = self.speech_embedding(prompt_speech_token)
+        else:
+            prompt_speech_token_emb = torch.zeros(1, 0, self.llm_input_size, dtype=text.dtype).to(device)
+        lm_input = torch.concat([sos_eos_emb, embedding, text, task_id_emb, prompt_speech_token_emb], dim=1)
+
+        # 4. cal min/max_length
+        min_len = int((text_len - prompt_text_len) * min_token_text_ratio)
+        max_len = int((text_len - prompt_text_len) * max_token_text_ratio)
+
+        # 5. step by step decode
+        out_tokens = []
+        offset = 0
+        att_cache, cnn_cache = torch.zeros((0, 0, 0, 0), device=lm_input.device), torch.zeros((0, 0, 0, 0), device=lm_input.device)
+        for i in range(max_len):
+            y_pred, att_cache, cnn_cache = self.llm.forward_chunk(lm_input, offset=offset, required_cache_size=-1,
+                                                                  att_cache=att_cache, cnn_cache=cnn_cache,
+                                                                  att_mask=torch.tril(torch.ones((1, lm_input.shape[1], lm_input.shape[1]),
+                                                                                                 device=lm_input.device)).to(torch.bool))
+            logp = self.llm_decoder(y_pred[:, -1]).log_softmax(dim=-1)
+            # force continue decode first token
+            if i == 0:
+                logp[:, self.speech_token_size] = -float('inf')
+            top_ids = self.sampling_ids(logp.squeeze(dim=0), out_tokens, sampling, ignore_eos=True if i < min_len else False).item()
+            if top_ids == self.speech_token_size:
+                break
+            # in stream mode, yield token one by one
+            yield top_ids
+            out_tokens.append(top_ids)
+            offset += lm_input.size(1)
+            lm_input = self.speech_embedding.weight[top_ids].reshape(1, 1, -1)
+
+
+class Qwen2Encoder(torch.nn.Module):
+    def __init__(self, pretrain_path):
+        super().__init__()
+        self.model = Qwen2ForCausalLM.from_pretrained(pretrain_path)
+
+    def forward_one_step(self, xs, masks, cache=None):
+        input_masks = masks[:, -1, :]
+        outs = self.model(
+            inputs_embeds=xs,
+            attention_mask=input_masks,
+            output_hidden_states=True,
+            return_dict=True,
+            use_cache=True,
+            past_key_values=cache,
+        )
+        xs = outs.hidden_states[-1]
+        new_cache = outs.past_key_values
+        return xs, new_cache
+
+
+class Qwen2LM(TransformerLM):
+    def __init__(
+            self,
+            llm_input_size: int,
+            llm_output_size: int,
+            speech_token_size: int,
+            llm: torch.nn.Module,
+            sampling: Callable,
+            length_normalized_loss: bool = True,
+            lsm_weight: float = 0.0,
+            mix_ratio: List[int] = [5, 15],
+            dpo: bool = False,
+    ):
+        torch.nn.Module.__init__(self)
+        self.llm_input_size = llm_input_size
+        self.llm_output_size = llm_output_size
+        self.speech_token_size = speech_token_size
+
+        # 2. build speech token language model related modules
+        self.sos_eos = 0
+        self.task_id = 1
+        self.fill_token = 2
+
+        self.llm_embedding = torch.nn.Embedding(2, llm_input_size)
+        self.llm = llm
+        self.llm_decoder = nn.Linear(llm_output_size, speech_token_size + 3)
+        self.criterion_ce = LabelSmoothingLoss(
+            size=speech_token_size + 3,
+            padding_idx=IGNORE_ID,
+            smoothing=lsm_weight,
+            normalize_length=length_normalized_loss,
+        )
+
+        # 3. [Optional] build speech token related modules
+        self.speech_embedding = torch.nn.Embedding(speech_token_size + 3, llm_input_size)
+
+        # 4. sampling method
+        self.sampling = sampling
+        self.mix_ratio = mix_ratio
+
+        # 5. [Optional] set dpo
+        self.dpo = dpo
+
+
+    def forward(
+            self,
+            batch: dict,
+            device: torch.device,
+        ) -> Dict[str, Optional[torch.Tensor]]:
+        text_token = batch['text_token'].to(device)
+        text_token_len = batch['text_token_len'].to(device)
+        speech_token = batch['speech_token'].to(device)
+        speech_token_len = batch['speech_token_len'].to(device)
+        if self.dpo:
+            reject_speech_token = batch['reject_speech_token'].to(device)
+            reject_speech_token_len = batch['reject_speech_token_len'].to(device)
+        # 1. prepare llm_target
+        sos_eos_emb = self.llm_embedding.weight[self.sos_eos].reshape(1, 1, -1)
+        task_id_emb = self.llm_embedding.weight[self.task_id].reshape(1, 1, -1)
+        target_ids = [torch.tensor([IGNORE_ID] * (1 + text_token_len[i]) + speech_token[i, :speech_token_len[i]].tolist() +
+                                        [self.speech_token_size]) for i in range(text_token.size(0))]
+        if self.dpo:
+            reject_target_ids = [torch.tensor([IGNORE_ID] * (1 + text_token_len[i]) + reject_speech_token[i, :reject_speech_token_len[i]].tolist() +
+                                            [self.speech_token_size]) for i in range(text_token.size(0))]
+            target_ids.extend(reject_target_ids)
+        target_ids = pad_sequence(target_ids, batch_first=True, padding_value=IGNORE_ID).to(device)
+
+        # 2. speech token projection
+        speech_emb = self.speech_embedding(speech_token)
+        if self.dpo:
+            reject_speech_emb = self.speech_embedding(reject_speech_token)
+
+        # 3. text token projection
+        text_token_lst = unpad_sequence(text_token, text_token_len, batch_first=True)
+        text_emb = [self.llm.model.model.embed_tokens(y) for y in text_token_lst]
+
+        # 4. prepare llm_input
+        speech_emb = unpad_sequence(speech_emb, speech_token_len.cpu(), batch_first=True)
+        input_emb = [torch.concat([sos_eos_emb.squeeze(dim=0), text_emb[i], task_id_emb.squeeze(dim=0), speech_emb[i]], dim=0)
+                     for i in range(len(text_emb))]
+        if self.dpo:
+            reject_speech_emb = unpad_sequence(reject_speech_emb, reject_speech_token_len.cpu(), batch_first=True)
+            reject_input_emb = [torch.concat([sos_eos_emb.squeeze(dim=0), text_emb[i], task_id_emb.squeeze(dim=0), reject_speech_emb[i]], dim=0)
+                                for i in range(len(text_emb))]
+            input_emb.extend(reject_input_emb)
+        input_emb_lengths = torch.tensor([i.size(0) for i in input_emb], dtype=torch.int32).to(device)
+        input_emb = pad_sequence(input_emb, batch_first=True, padding_value=IGNORE_ID).to(device)
+
+        attention_mask = ~make_pad_mask(input_emb_lengths)
+
+        result = self.llm.model(
+            inputs_embeds=input_emb,
+            attention_mask=attention_mask,
+            return_dict=True
+        )
+        hidden_states = result.hidden_states
+        logits = self.llm_decoder(hidden_states[-1])
+        loss = self.criterion_ce(logits[: speech_token.shape[0]], target_ids[: speech_token.shape[0]])
+        acc = th_accuracy(
+            logits[: speech_token.shape[0]].view(-1, self.speech_token_size + 3),
+            target_ids[: speech_token.shape[0]],
+            ignore_label=IGNORE_ID,
+        )
+        if not self.dpo:
+            return {
+                "loss": loss,
+                "acc": acc,
+            }
+        else:
+            all_logps_sum, all_logps_mean = self.get_batch_logps(
+                logits, target_ids, attention_mask, text_token_len, average_log_prob=False, ignore_id=IGNORE_ID
+            )
+            chosen_logps = all_logps_sum[: speech_token.shape[0]]
+            rejected_logps = all_logps_sum[speech_token.shape[0]:]
+            return {
+                "loss": loss,
+                "acc": acc,
+                "chosen_logps": chosen_logps,
+                "rejected_logps": rejected_logps
+            }
+
+
+    def get_batch_logps(
+        self,
+        logits: torch.FloatTensor,
+        labels: torch.LongTensor,
+        attention_mask,
+        prompt_token_lens,
+        average_log_prob: bool = False,
+        ignore_id: int = -1,
+    ) -> torch.FloatTensor:
+        """Compute the log probabilities of the given labels under the given logits.
+
+        Args:
+            logits: Logits of the model (unnormalized). Shape: (batch_size, sequence_length, vocab_size)
+            labels: Labels for which to compute the log probabilities. Label tokens with a value of -100 are ignored. Shape: (batch_size, sequence_length)
+            average_log_prob: If True, return the average log probability per (non-masked) token. Otherwise, return the sum of the log probabilities of the (non-masked) tokens.
+
+        Returns:
+            A tensor of shape (batch_size,) containing the average/sum log probabilities of the given labels under the given logits.
+        """
+        assert average_log_prob == False
+        assert logits.shape[:-1] == labels.shape
+        labels = labels[:, 1:].clone()
+        logits = logits[:, :-1, :]
+        loss_masks = attention_mask.clone().bool()
+        # mask prompts
+        for mask, text_token_len in zip(loss_masks, prompt_token_lens):
+            mask[:text_token_len + 1] = False
+        loss_masks = loss_masks[:, 1:]
+        labels[loss_masks == False] = 0
+        # dummy token; we'll ignore the losses on these tokens later
+        ignore = labels == ignore_id
+        labels = labels.masked_fill(ignore, 0)  # avoid -1 index
+        per_token_logps = torch.gather(logits.log_softmax(-1), dim=2, index=labels.unsqueeze(2)).squeeze(2)   # (bs, time,)
+        logprobs_sums = (per_token_logps * loss_masks).sum(-1)
+        logprobs_means = (per_token_logps * loss_masks).sum(-1) / loss_masks.sum(-1)
+        return logprobs_sums, logprobs_means
+
+
+    @torch.inference_mode()
+    def inference(
+            self,
+            text: torch.Tensor,
+            text_len: torch.Tensor,
+            prompt_text: torch.Tensor,
+            prompt_text_len: torch.Tensor,
+            prompt_speech_token: torch.Tensor,
+            prompt_speech_token_len: torch.Tensor,
+            embedding: torch.Tensor,
+            sampling: int = 25,
+            max_token_text_ratio: float = 20,
+            min_token_text_ratio: float = 2,
+    ) -> Generator[torch.Tensor, None, None]:
+        device = text.device
+        text = torch.concat([prompt_text, text], dim=1)
+        text_len += prompt_text_len
+        text = self.llm.model.model.embed_tokens(text)
+
+        # 3. concat llm_input
+        sos_eos_emb = self.llm_embedding.weight[self.sos_eos].reshape(1, 1, -1)
+        task_id_emb = self.llm_embedding.weight[self.task_id].reshape(1, 1, -1)
+        if prompt_speech_token_len != 0:
+            prompt_speech_token_emb = self.speech_embedding(prompt_speech_token)
+        else:
+            prompt_speech_token_emb = torch.zeros(1, 0, self.llm_input_size, dtype=text.dtype).to(device)
+        lm_input = torch.concat([sos_eos_emb, text, task_id_emb, prompt_speech_token_emb], dim=1)
+
+        # 4. cal min/max_length
+        min_len = int((text_len - prompt_text_len) * min_token_text_ratio)
+        max_len = int((text_len - prompt_text_len) * max_token_text_ratio)
+
+        # 5. step by step decode
+        out_tokens = []
+        cache = None
+        for i in range(max_len):
+            y_pred, cache = self.llm.forward_one_step(lm_input,
+                                                      masks=torch.tril(torch.ones((1, lm_input.shape[1], lm_input.shape[1]), device=lm_input.device)).to(torch.bool),
+                                                      cache=cache)
+            logp = self.llm_decoder(y_pred[:, -1]).log_softmax(dim=-1)
+            top_ids = self.sampling_ids(logp.squeeze(dim=0), out_tokens, sampling, ignore_eos=True if i < min_len else False).item()
+            if top_ids == self.speech_token_size:
+                break
+            if top_ids > self.speech_token_size:
+                continue
+            # in stream mode, yield token one by one
+            yield top_ids
+            out_tokens.append(top_ids)
+            lm_input = self.speech_embedding.weight[top_ids].reshape(1, 1, -1)
+
+    @torch.inference_mode()
+    def inference_bistream(
+            self,
+            text: Generator,
+            prompt_text: torch.Tensor,
+            prompt_text_len: torch.Tensor,
+            prompt_speech_token: torch.Tensor,
+            prompt_speech_token_len: torch.Tensor,
+            embedding: torch.Tensor,
+            sampling: int = 25,
+            max_token_text_ratio: float = 20,
+            min_token_text_ratio: float = 2,
+    ) -> Generator[torch.Tensor, None, None]:
+
+        device = prompt_text.device
+        # 1. prepare input
+        sos_eos_emb = self.llm_embedding.weight[self.sos_eos].reshape(1, 1, -1)
+        task_id_emb = self.llm_embedding.weight[self.task_id].reshape(1, 1, -1)
+        if prompt_speech_token_len != 0:
+            prompt_speech_token_emb = self.speech_embedding(prompt_speech_token)
+        else:
+            prompt_speech_token_emb = torch.zeros(1, 0, self.llm_input_size, dtype=prompt_text.dtype).to(device)
+        lm_input = torch.concat([sos_eos_emb], dim=1)
+
+        # 2. iterate text
+        out_tokens = []
+        cache = None
+        # NOTE init prompt_text as text_cache as it is basically impossible prompt_speech_token/prompt_text < 15/5
+        text_cache = self.llm.model.model.embed_tokens(prompt_text)
+        next_fill_index = -1
+        for this_text in text:
+            text_cache = torch.concat([text_cache, self.llm.model.model.embed_tokens(this_text)], dim=1)
+            # prompt_speech_token_emb not empty, try append to lm_input
+            while prompt_speech_token_emb.size(1) != 0:
+                if text_cache.size(1) >= self.mix_ratio[0]:
+                    lm_input_text, lm_input_speech = text_cache[:, :self.mix_ratio[0]], prompt_speech_token_emb[:, :self.mix_ratio[1]]
+                    logging.info('append {} text token {} speech token'.format(lm_input_text.size(1), lm_input_speech.size(1)))
+                    lm_input = torch.concat([lm_input, lm_input_text, lm_input_speech], dim=1)
+                    text_cache, prompt_speech_token_emb = text_cache[:, self.mix_ratio[0]:], prompt_speech_token_emb[:, self.mix_ratio[1]:]
+                else:
+                    logging.info('not enough text token to decode, wait for more')
+                    break
+            # no prompt_speech_token_emb remain, can decode some speech token
+            if prompt_speech_token_emb.size(1) == 0:
+                if (len(out_tokens) != 0 and out_tokens[-1] == self.speech_token_size + 2) or (len(out_tokens) == 0 and lm_input.size(1) == 1):
+                    logging.info('get fill token, need to append more text token')
+                    if text_cache.size(1) >= self.mix_ratio[0]:
+                        lm_input_text = text_cache[:, :self.mix_ratio[0]]
+                        logging.info('append {} text token'.format(lm_input_text.size(1)))
+                        if len(out_tokens) != 0 and out_tokens[-1] == self.speech_token_size + 2:
+                            lm_input = lm_input_text
+                        else:
+                            lm_input = torch.concat([lm_input, lm_input_text], dim=1)
+                        text_cache = text_cache[:, self.mix_ratio[0]:]
+                    else:
+                        logging.info('not enough text token to decode, wait for more')
+                        continue
+                while True:
+                    seq_len = lm_input.shape[1] if cache is None else lm_input.shape[1] + cache[0][0].size(2)
+                    y_pred, cache = self.llm.forward_one_step(lm_input,
+                                                masks=torch.tril(torch.ones((1, seq_len, seq_len), device=lm_input.device)).to(torch.bool),
+                                                cache=cache)
+                    logp = self.llm_decoder(y_pred[:, -1]).log_softmax(dim=-1)
+                    if next_fill_index != -1 and len(out_tokens) == next_fill_index:
+                        top_ids = self.speech_token_size + 2
+                        next_fill_index += (self.mix_ratio[1] + 1)
+                    else:
+                        top_ids = self.sampling_ids(logp.squeeze(dim=0), out_tokens, sampling, ignore_eos=True).item()
+                    if top_ids == self.speech_token_size + 2:
+                        next_fill_index = len(out_tokens) + self.mix_ratio[1] + 1
+                        logging.info('fill_token index {} next fill_token index {}'.format(len(out_tokens), next_fill_index))
+                    out_tokens.append(top_ids)
+                    if top_ids >= self.speech_token_size:
+                        if top_ids == self.speech_token_size + 2:
+                            break
+                        else:
+                            raise ValueError('should not get token {}'.format(top_ids))
+                    yield top_ids
+                    lm_input = self.speech_embedding.weight[top_ids].reshape(1, 1, -1)
+
+        # 3. final decode
+        lm_input = torch.concat([lm_input, text_cache, task_id_emb], dim=1)
+        logging.info('no more text token, decode until met eos')
+        while True:
+            seq_len = lm_input.shape[1] if cache is None else lm_input.shape[1] + cache[0][0].size(2)
+            y_pred, cache = self.llm.forward_one_step(lm_input,
+                                                      masks=torch.tril(torch.ones((1, seq_len, seq_len), device=lm_input.device)).to(torch.bool),
+                                                      cache=cache)
+            logp = self.llm_decoder(y_pred[:, -1]).log_softmax(dim=-1)
+            top_ids = self.sampling_ids(logp.squeeze(dim=0), out_tokens, sampling, ignore_eos=False).item()
+            out_tokens.append(top_ids)
+            if top_ids >= self.speech_token_size:
+                if top_ids == self.speech_token_size:
+                    break
+                else:
+                    raise ValueError('should not get token {}'.format(top_ids))
+            # in stream mode, yield token one by one
+            yield top_ids
+            lm_input = self.speech_embedding.weight[top_ids].reshape(1, 1, -1)