Add MOSS music audio captioner

toolkit/audio/moss_music/LICENSE

@@ -0,0 +1,21 @@
+The code in this directory is adapted from MOSS-Music by the OpenMOSS Team
+(https://github.com/OpenMOSS/MOSS-Music), licensed under the Apache License,
+Version 2.0. Modifications for ai-toolkit: transformers 5.x compatibility
+(WhisperEncoderLayer tensor return, generate input preparation), removal of the
+chat-template loader, and direct Qwen2Tokenizer loading.
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   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.

toolkit/audio/moss_music/__init__.py

@@ -0,0 +1,35 @@
+import torch
+
+from .configuration import MossMusicConfig, MossMusicEncoderConfig
+from .modeling import MossMusicModel
+from .processing import MelConfig, MossMusicProcessor
+
+HF_REPO = "OpenMOSS-Team/MOSS-Music-8B-Instruct"
+
+
+def load_moss_music(
+    name_or_path: str = HF_REPO,
+    dtype: torch.dtype = torch.bfloat16,
+    device="cpu",
+    enable_time_marker: bool = True,
+):
+    """Model + processor from a hub id or local checkpoint dir. Time markers
+    (elapsed seconds interleaved into the audio tokens) are what the model was
+    trained with and keep long transcriptions from looping; leave them on."""
+    model = MossMusicModel.from_pretrained(name_or_path, dtype=dtype, device_map=device)
+    model.eval()
+    processor = MossMusicProcessor.from_pretrained(
+        name_or_path, enable_time_marker=enable_time_marker
+    )
+    return model, processor
+
+
+__all__ = [
+    "HF_REPO",
+    "MelConfig",
+    "MossMusicConfig",
+    "MossMusicEncoderConfig",
+    "MossMusicModel",
+    "MossMusicProcessor",
+    "load_moss_music",
+]

toolkit/audio/moss_music/configuration.py

@@ -0,0 +1,135 @@
+from typing import List, Optional
+from dataclasses import dataclass, field
+
+from transformers import PretrainedConfig, Qwen3Config
+
+
+@dataclass
+class MossMusicEncoderConfig:
+    d_model: int = 1280
+    output_dim: int = 1280
+    num_mel_bins: int = 128
+    encoder_layers: int = 32
+    encoder_attention_heads: int = 20
+    encoder_ffn_dim: int = 5120
+    downsample_rate: int = 8
+    downsample_hidden_size: int = 480
+    encoder_attention_window_size: int = 100
+    max_source_positions: int = 1500
+    dropout: float = 0.1
+    attention_dropout: float = 0.1
+    activation_dropout: float = 0.0
+    activation_function: str = "gelu"
+    layer_norm_eps: float = 1e-5
+    _attn_implementation: str = "eager"
+    pretrained_path: str = ""
+
+    n_window: int = 200
+    conv_chunksize: int = 64
+
+    deepstack_encoder_layer_indexes: List[int] = field(
+        default_factory=lambda: [8, 16, 24]
+    )
+
+    @classmethod
+    def from_dict(cls, config_dict):
+        if config_dict is None:
+            return cls()
+        allowed_keys = set(cls.__dataclass_fields__.keys())
+        filtered = {k: v for k, v in config_dict.items() if k in allowed_keys}
+        return cls(**filtered)
+
+    def to_dict(self):
+        return {
+            "d_model": self.d_model,
+            "output_dim": self.output_dim,
+            "num_mel_bins": self.num_mel_bins,
+            "encoder_layers": self.encoder_layers,
+            "encoder_attention_heads": self.encoder_attention_heads,
+            "encoder_ffn_dim": self.encoder_ffn_dim,
+            "downsample_rate": self.downsample_rate,
+            "downsample_hidden_size": self.downsample_hidden_size,
+            "encoder_attention_window_size": self.encoder_attention_window_size,
+            "max_source_positions": self.max_source_positions,
+            "dropout": self.dropout,
+            "attention_dropout": self.attention_dropout,
+            "activation_dropout": self.activation_dropout,
+            "activation_function": self.activation_function,
+            "layer_norm_eps": self.layer_norm_eps,
+            "_attn_implementation": self._attn_implementation,
+            "pretrained_path": self.pretrained_path,
+            "n_window": self.n_window,
+            "conv_chunksize": self.conv_chunksize,
+            "deepstack_encoder_layer_indexes": list(
+                self.deepstack_encoder_layer_indexes or []
+            ),
+        }
+
+
+class MossMusicConfig(PretrainedConfig):
+    model_type = "moss_music"
+    is_composition = True
+
+    def __init__(
+        self,
+        audio_config=None,
+        language_config=None,
+        adapter_hidden_size=8192,
+        ignore_index=-100,
+        deepstack_num_inject_layers: Optional[int] = None,
+        **kwargs,
+    ):
+        if isinstance(audio_config, dict):
+            audio_config = MossMusicEncoderConfig.from_dict(audio_config)
+        elif audio_config is None:
+            audio_config = MossMusicEncoderConfig()
+
+        if isinstance(language_config, dict):
+            language_config = Qwen3Config(**language_config)
+        elif language_config is None:
+            language_config = Qwen3Config()
+
+        self.audio_config = audio_config
+        self.language_config = language_config
+        self.adapter_hidden_size = adapter_hidden_size
+        self.ignore_index = ignore_index
+        self.deepstack_num_inject_layers = deepstack_num_inject_layers
+
+        _propagate_keys = {
+            "num_hidden_layers",
+            "eos_token_id",
+            "bos_token_id",
+            "vocab_size",
+            "tie_word_embeddings",
+        }
+        for key in ("num_hidden_layers", "eos_token_id", "bos_token_id", "vocab_size"):
+            kwargs.setdefault(key, getattr(language_config, key, None))
+        kwargs.setdefault("tie_word_embeddings", False)
+
+        if hasattr(language_config, "to_dict"):
+            _lang_keys = set(language_config.to_dict().keys())
+            for key in list(kwargs.keys()):
+                if key in _lang_keys and key not in _propagate_keys:
+                    kwargs.pop(key)
+
+        super().__init__(**kwargs)
+
+    def to_dict(self):
+        output = super().to_dict()
+        output["audio_config"] = (
+            self.audio_config.to_dict()
+            if hasattr(self.audio_config, "to_dict")
+            else self.audio_config
+        )
+        output["language_config"] = (
+            self.language_config.to_dict()
+            if hasattr(self.language_config, "to_dict")
+            else self.language_config
+        )
+        output["adapter_hidden_size"] = self.adapter_hidden_size
+        output["ignore_index"] = self.ignore_index
+        output["deepstack_num_inject_layers"] = self.deepstack_num_inject_layers
+        return output
+
+
+__all__ = ["MossMusicEncoderConfig", "MossMusicConfig"]

toolkit/audio/moss_music/modeling.py

@@ -0,0 +1,568 @@
+from typing import Any, List, Optional, Tuple
+import math
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from transformers.generation.utils import GenerationMixin
+from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast
+from transformers.modeling_utils import PreTrainedModel
+from transformers.models.qwen3.modeling_qwen3 import Qwen3DecoderLayer, Qwen3Model
+from transformers.models.whisper.modeling_whisper import WhisperEncoderLayer
+
+from .configuration import MossMusicConfig, MossMusicEncoderConfig
+
+
+class SinusoidsPositionEmbedding(nn.Module):
+    def __init__(self, num_positions: int, embedding_dim: int):
+        super().__init__()
+        max_timescale = 10000.0
+        log_timescale_increment = math.log(max_timescale) / (embedding_dim // 2 - 1)
+        inv_timescales = torch.exp(
+            -log_timescale_increment * torch.arange(embedding_dim // 2).float()
+        )
+        self.register_buffer("inv_timescales", inv_timescales, persistent=False)
+
+    def forward(self, seq_len: int, device: torch.device):
+        scaled_time = torch.arange(
+            seq_len, device=device, dtype=self.inv_timescales.dtype
+        ).unsqueeze(1) * self.inv_timescales.unsqueeze(0)
+        sin_emb = torch.sin(scaled_time)
+        cos_emb = torch.cos(scaled_time)
+        pos_emb = torch.cat([sin_emb, cos_emb], dim=1)
+        return pos_emb.unsqueeze(0)
+
+
+class MossMusicEncoder(nn.Module):
+    """Audio encoder with conv-stem downsampling and Whisper transformer layers."""
+
+    def __init__(self, config: MossMusicEncoderConfig):
+        super().__init__()
+        self.config = config
+        self.gelu = nn.GELU()
+
+        self.conv1 = nn.Conv2d(
+            1,
+            config.downsample_hidden_size,
+            kernel_size=(3, 3),
+            stride=(2, 2),
+            padding=(1, 1),
+        )
+        self.conv2 = nn.Conv2d(
+            config.downsample_hidden_size,
+            config.downsample_hidden_size,
+            kernel_size=(3, 3),
+            stride=(2, 2),
+            padding=(1, 1),
+        )
+        self.conv3 = nn.Conv2d(
+            config.downsample_hidden_size,
+            config.downsample_hidden_size,
+            kernel_size=(3, 3),
+            stride=(2, 2),
+            padding=(1, 1),
+        )
+
+        # 128 mel bins / 8 = 16 after 3 convs with stride=2
+        self.stem_proj = nn.Linear(config.downsample_hidden_size * 16, config.d_model)
+        self.embed_positions = SinusoidsPositionEmbedding(
+            config.max_source_positions, config.d_model
+        )
+        self.layers = nn.ModuleList(
+            [WhisperEncoderLayer(config) for _ in range(config.encoder_layers)]
+        )
+        self.layer_norm = nn.LayerNorm(config.d_model, eps=config.layer_norm_eps)
+        self.out_proj = (
+            nn.Linear(config.d_model, config.output_dim, bias=False)
+            if config.output_dim != config.d_model
+            else nn.Identity()
+        )
+
+        self.deepstack_encoder_layer_indexes = list(
+            config.deepstack_encoder_layer_indexes or []
+        )
+        self._deepstack_capture_map = {
+            layer_idx: capture_idx
+            for capture_idx, layer_idx in enumerate(self.deepstack_encoder_layer_indexes)
+        }
+
+        self.n_window = int(config.n_window)
+        self.chunk_frames = int(self.n_window * 2)
+        self.conv_chunksize = int(config.conv_chunksize)
+
+    @property
+    def dtype(self) -> torch.dtype:
+        return self.conv1.weight.dtype
+
+    @staticmethod
+    def _compute_downsampled_length(lengths: torch.Tensor) -> torch.Tensor:
+        def conv_out_len(L):
+            return (L - 1) // 2 + 1
+
+        return conv_out_len(conv_out_len(conv_out_len(lengths)))
+
+    def _encode_chunk_batch(
+        self,
+        input_features: torch.Tensor,
+        seq_lengths: torch.Tensor,
+    ) -> Tuple[torch.Tensor, List[torch.Tensor]]:
+        """Encode a batch of (already padded) chunks through the conv stem and
+        transformer layers. Returns (last_hidden, ordered_deepstack_hidden_states).
+        """
+        if input_features.dim() == 2:
+            input_features = input_features.unsqueeze(0)
+
+        downsampled_lengths = self._compute_downsampled_length(seq_lengths)
+
+        # [B, n_mels, T] -> [B, 1, n_mels, T]
+        x = input_features.unsqueeze(1)
+        x = self.gelu(self.conv1(x))
+        x = self.gelu(self.conv2(x))
+        x = self.gelu(self.conv3(x))
+
+        # [B, C, F, T] -> [B, T, C*F]
+        x = x.permute(0, 3, 1, 2).contiguous().flatten(2)
+        x = self.stem_proj(x)
+
+        max_len = int(downsampled_lengths.max().item())
+        if x.size(1) > max_len:
+            x = x[:, :max_len, :]
+
+        positions = self.embed_positions(x.shape[1], x.device)
+        x = x + positions.to(x.dtype)
+
+        padding_mask = (
+            torch.arange(x.size(1), device=x.device)[None, :]
+            >= downsampled_lengths[:, None]
+        )
+        attention_mask = (1.0 - (~padding_mask).to(dtype=x.dtype)) * torch.finfo(
+            x.dtype
+        ).min
+        attention_mask = attention_mask.unsqueeze(1).unsqueeze(1)
+
+        deepstack_hidden_states: List[Optional[torch.Tensor]] = [None] * len(
+            self.deepstack_encoder_layer_indexes
+        )
+        for layer_idx, layer in enumerate(self.layers):
+            x = layer(x, attention_mask)
+            capture_idx = self._deepstack_capture_map.get(layer_idx)
+            if capture_idx is not None:
+                deepstack_hidden_states[capture_idx] = x
+
+        x = self.layer_norm(x)
+        x = self.out_proj(x)
+
+        ordered_deepstack_hidden_states = [
+            h for h in deepstack_hidden_states if h is not None
+        ]
+        if not isinstance(self.out_proj, nn.Identity):
+            ordered_deepstack_hidden_states = [
+                self.out_proj(h) for h in ordered_deepstack_hidden_states
+            ]
+        return x, ordered_deepstack_hidden_states
+
+    def forward(
+        self,
+        input_features: torch.Tensor,
+        feature_lens: Optional[torch.Tensor] = None,
+        output_deepstack_hidden_states: bool = True,
+    ) -> BaseModelOutputWithPast:
+        if input_features.dim() == 3:
+            if feature_lens is None:
+                feature_lens = torch.full(
+                    (input_features.size(0),),
+                    input_features.size(-1),
+                    dtype=torch.long,
+                    device=input_features.device,
+                )
+            else:
+                feature_lens = feature_lens.to(
+                    device=input_features.device, dtype=torch.long
+                )
+            valid_chunks = [
+                input_features[i, :, : int(feature_lens[i].item())]
+                for i in range(int(input_features.shape[0]))
+            ]
+            input_features = torch.cat(valid_chunks, dim=1)
+        elif input_features.dim() != 2:
+            raise ValueError(
+                f"Expected [n_mels, T] or [B, n_mels, T], got {tuple(input_features.shape)}."
+            )
+
+        if feature_lens is None:
+            feature_lens = torch.tensor(
+                [int(input_features.shape[1])],
+                device=input_features.device,
+                dtype=torch.long,
+            )
+        else:
+            feature_lens = feature_lens.to(
+                device=input_features.device, dtype=torch.long
+            )
+
+        chunk_frames = int(self.chunk_frames)
+        chunk_num = torch.ceil(
+            feature_lens.to(torch.float32) / float(chunk_frames)
+        ).long()
+        chunk_lengths = torch.full(
+            (int(chunk_num.sum().item()),),
+            chunk_frames,
+            dtype=torch.long,
+            device=feature_lens.device,
+        )
+        tail_chunk_index = F.pad(chunk_num, (1, 0), value=-1).cumsum(0)[1:]
+        chunk_lengths[tail_chunk_index] = feature_lens % chunk_frames
+        chunk_lengths[chunk_lengths == 0] = chunk_frames
+
+        chunk_list = input_features.T.split(chunk_lengths.tolist(), dim=0)
+        padded_feature = nn.utils.rnn.pad_sequence(
+            chunk_list, batch_first=True
+        ).transpose(1, 2)
+
+        feature_lens_after_cnn = self._compute_downsampled_length(chunk_lengths)
+        t_down_max = (
+            int(feature_lens_after_cnn.max().item())
+            if feature_lens_after_cnn.numel() > 0
+            else 0
+        )
+        padded_mask_after_cnn = nn.utils.rnn.pad_sequence(
+            [
+                torch.ones(int(L.item()), dtype=torch.bool, device=padded_feature.device)
+                for L in feature_lens_after_cnn
+            ],
+            batch_first=True,
+        )
+        if padded_mask_after_cnn.shape[1] < t_down_max:
+            padded_mask_after_cnn = F.pad(
+                padded_mask_after_cnn,
+                (0, t_down_max - padded_mask_after_cnn.shape[1]),
+                value=False,
+            )
+
+        num_deepstack = len(self.deepstack_encoder_layer_indexes)
+        padded_embeds: List[torch.Tensor] = []
+        deepstack_padded_embeds: List[List[torch.Tensor]] = [
+            [] for _ in range(num_deepstack)
+        ]
+        for feat_chunk, len_chunk in zip(
+            padded_feature.split(self.conv_chunksize, dim=0),
+            chunk_lengths.split(self.conv_chunksize, dim=0),
+        ):
+            out, deepstack_outs = self._encode_chunk_batch(feat_chunk, len_chunk)
+            if out.shape[1] < t_down_max:
+                out = F.pad(out, (0, 0, 0, t_down_max - out.shape[1]))
+            padded_embeds.append(out)
+            if output_deepstack_hidden_states and num_deepstack > 0:
+                if len(deepstack_outs) != num_deepstack:
+                    raise RuntimeError(
+                        "Deepstack output count does not match configured layer indexes."
+                    )
+                for capture_idx, ds in enumerate(deepstack_outs):
+                    if ds.shape[1] < t_down_max:
+                        ds = F.pad(ds, (0, 0, 0, t_down_max - ds.shape[1]))
+                    deepstack_padded_embeds[capture_idx].append(ds)
+
+        if padded_embeds:
+            padded_embed = torch.cat(padded_embeds, dim=0)
+        else:
+            padded_embed = torch.empty(
+                (0, t_down_max, self.config.output_dim),
+                device=padded_feature.device,
+            )
+
+        valid_tokens = padded_embed[padded_mask_after_cnn]  # [N_valid, D]
+        last_hidden_state = valid_tokens.unsqueeze(0)  # [1, N_valid, D]
+
+        deepstack_states: Optional[Tuple[torch.Tensor, ...]] = None
+        if output_deepstack_hidden_states and num_deepstack > 0:
+            collected: List[torch.Tensor] = []
+            for chunks_list in deepstack_padded_embeds:
+                if chunks_list:
+                    ds = torch.cat(chunks_list, dim=0)
+                    collected.append(ds[padded_mask_after_cnn].unsqueeze(0))
+                else:
+                    collected.append(
+                        torch.empty(
+                            (1, 0, self.config.output_dim),
+                            device=padded_feature.device,
+                            dtype=padded_embed.dtype,
+                        )
+                    )
+            deepstack_states = tuple(collected)
+
+        return BaseModelOutputWithPast(
+            last_hidden_state=last_hidden_state,
+            hidden_states=deepstack_states,
+        )
+
+
+class GatedMLP(nn.Module):
+    def __init__(self, input_size, hidden_size, output_size):
+        super().__init__()
+        self.gate_proj = nn.Linear(input_size, hidden_size, bias=False)
+        self.up_proj = nn.Linear(input_size, hidden_size, bias=False)
+        self.down_proj = nn.Linear(hidden_size, output_size, bias=False)
+        self.act_fn = nn.SiLU()
+
+    def forward(self, x):
+        return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
+
+
+class MossMusicPreTrainedModel(PreTrainedModel):
+    config_class = MossMusicConfig
+    config: MossMusicConfig
+    base_model_prefix = ""
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["Qwen3DecoderLayer"]
+    _skip_keys_device_placement = ["past_key_values"]
+    _supports_flash_attn = True
+    _supports_sdpa = True
+    _supports_flex_attn = True
+
+    _can_compile_fullgraph = False
+    _supports_attention_backend = True
+    _can_record_outputs = {"hidden_states": Qwen3DecoderLayer}
+
+
+class MossMusicModel(MossMusicPreTrainedModel, GenerationMixin):
+    config_class = MossMusicConfig
+    _tied_weights_keys: List[str] = []
+
+    def __init__(self, config: MossMusicConfig):
+        super().__init__(config)
+
+        self.audio_encoder = MossMusicEncoder(config.audio_config)
+        self.language_model = Qwen3Model(config.language_config)
+
+        self.audio_adapter = GatedMLP(
+            input_size=config.audio_config.output_dim,
+            hidden_size=config.adapter_hidden_size,
+            output_size=config.language_config.hidden_size,
+        )
+
+        deepstack_k = len(
+            getattr(config.audio_config, "deepstack_encoder_layer_indexes", []) or []
+        )
+        if config.deepstack_num_inject_layers is not None:
+            deepstack_k = min(deepstack_k, int(config.deepstack_num_inject_layers))
+        self.deepstack_audio_merger_list = nn.ModuleList(
+            [
+                GatedMLP(
+                    input_size=config.audio_config.output_dim,
+                    hidden_size=config.adapter_hidden_size,
+                    output_size=config.language_config.hidden_size,
+                )
+                for _ in range(deepstack_k)
+            ]
+        )
+
+        self.vocab_size = config.language_config.vocab_size
+        self.lm_head = nn.Linear(
+            config.language_config.hidden_size, self.vocab_size, bias=False
+        )
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.language_model.get_input_embeddings()
+
+    def set_input_embeddings(self, value):
+        self.language_model.set_input_embeddings(value)
+
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head = new_embeddings
+
+    def get_audio_features(self, input_features, feature_lens):
+        audio_outputs = self.audio_encoder(
+            input_features=input_features,
+            feature_lens=feature_lens,
+            output_deepstack_hidden_states=True,
+        )
+        deepstack = (
+            list(audio_outputs.hidden_states)
+            if audio_outputs.hidden_states is not None
+            else None
+        )
+        return audio_outputs.last_hidden_state, deepstack
+
+    def _apply_deepstack_to_hidden_states(
+        self,
+        hidden_states: torch.Tensor,
+        audio_input_mask: torch.Tensor,
+        deepstack_embeds: torch.Tensor,
+    ) -> torch.Tensor:
+        audio_input_mask = audio_input_mask.to(hidden_states.device)
+        deepstack_embeds = deepstack_embeds.to(hidden_states.device, hidden_states.dtype)
+        flat = deepstack_embeds.reshape(-1, deepstack_embeds.shape[-1])
+        hs = hidden_states.clone()
+        hs[audio_input_mask] = hs[audio_input_mask] + flat
+        return hs
+
+    def _register_llm_deepstack_hooks(
+        self,
+        audio_input_mask: torch.Tensor,
+        deepstack_audio_embeds: List[torch.Tensor],
+    ):
+        if deepstack_audio_embeds is None or len(deepstack_audio_embeds) == 0:
+            return []
+
+        layers = getattr(self.language_model, "layers", None)
+        if layers is None:
+            raise RuntimeError(
+                "Qwen3Model does not expose `.layers`; cannot register DeepStack hooks."
+            )
+
+        num_inject = len(deepstack_audio_embeds)
+        handles = []
+
+        for layer_idx, layer in enumerate(layers):
+            if layer_idx >= num_inject:
+                break
+
+            def _make_llm_hook(k: int):
+                def _hook(_module, _inputs, _output):
+                    if isinstance(_output, (tuple, list)):
+                        hs = _output[0]
+                        new_hs = self._apply_deepstack_to_hidden_states(
+                            hs, audio_input_mask, deepstack_audio_embeds[k]
+                        )
+                        return (new_hs,) + tuple(_output[1:])
+                    else:
+                        return self._apply_deepstack_to_hidden_states(
+                            _output, audio_input_mask, deepstack_audio_embeds[k]
+                        )
+
+                return _hook
+
+            handles.append(layer.register_forward_hook(_make_llm_hook(layer_idx)))
+
+        return handles
+
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        audio_data: Optional[torch.FloatTensor] = None,
+        audio_data_seqlens: Optional[torch.Tensor] = None,
+        audio_input_mask: Optional[torch.Tensor] = None,
+        **kwargs: Any,
+    ) -> CausalLMOutputWithPast:
+        # generate() still passes return_dict; transformers 5 models ignore it
+        kwargs.pop("return_dict", None)
+        if inputs_embeds is None:
+            inputs_embeds = self.get_input_embeddings()(input_ids)
+
+        hook_handles = []
+        if audio_data is not None:
+            if audio_input_mask is None:
+                raise ValueError(
+                    "audio_input_mask is required when audio_data is provided."
+                )
+
+            audio_embeds, deepstack = self.get_audio_features(audio_data, audio_data_seqlens)
+            audio_embeds = self.audio_adapter(audio_embeds)
+
+            audio_token_count = int(audio_input_mask.to(torch.int32).sum().item())
+            if audio_token_count != int(audio_embeds.shape[1]):
+                raise ValueError(
+                    f"Audio token count mismatch: audio_input_mask has {audio_token_count} audio tokens, "
+                    f"but audio_embeds has length {int(audio_embeds.shape[1])}."
+                )
+
+            mask_expanded = audio_input_mask.unsqueeze(-1).expand_as(inputs_embeds)
+            inputs_embeds = inputs_embeds.clone()
+            inputs_embeds.masked_scatter_(mask_expanded, audio_embeds)
+
+            if deepstack is not None and len(self.deepstack_audio_merger_list) > 0:
+                deepstack_audio_embeds = []
+                for i, x in enumerate(deepstack[: len(self.deepstack_audio_merger_list)]):
+                    ds = self.deepstack_audio_merger_list[i](x)
+                    if int(ds.shape[1]) != audio_token_count:
+                        raise ValueError(
+                            f"DeepStack audio seq_len mismatch at index {i}: "
+                            f"expected {audio_token_count}, got {int(ds.shape[1])}."
+                        )
+                    deepstack_audio_embeds.append(ds)
+
+                try:
+                    hook_handles = self._register_llm_deepstack_hooks(
+                        audio_input_mask, deepstack_audio_embeds
+                    )
+                except Exception:
+                    for h in hook_handles:
+                        h.remove()
+                    raise
+
+        try:
+            outputs = self.language_model(
+                input_ids=None,
+                attention_mask=attention_mask,
+                position_ids=position_ids,
+                past_key_values=past_key_values,
+                inputs_embeds=inputs_embeds,
+                use_cache=use_cache,
+                **kwargs,
+            )
+        finally:
+            for h in hook_handles:
+                h.remove()
+
+        hidden_states = outputs[0]
+        logits = self.lm_head(hidden_states)
+
+        loss = None
+        if labels is not None:
+            shift_logits = logits[..., :-1, :].contiguous()
+            shift_labels = labels[..., 1:].contiguous()
+            loss_fct = nn.CrossEntropyLoss(ignore_index=self.config.ignore_index)
+            shift_logits = shift_logits.view(-1, self.config.language_config.vocab_size)
+            shift_labels = shift_labels.view(-1)
+            shift_labels = shift_labels.to(shift_logits.device)
+            loss = loss_fct(shift_logits, shift_labels)
+
+        return CausalLMOutputWithPast(
+            loss=loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+    def prepare_inputs_for_generation(
+        self,
+        input_ids,
+        next_sequence_length=None,
+        past_key_values=None,
+        attention_mask=None,
+        inputs_embeds=None,
+        is_first_iteration=False,
+        **kwargs,
+    ):
+        # audio is encoded once into the prompt embeds; re-sending it on decode
+        # steps would re-run the encoder against a stale audio_input_mask
+        if not is_first_iteration:
+            for key in ("audio_data", "audio_data_seqlens", "audio_input_mask"):
+                kwargs.pop(key, None)
+        return super().prepare_inputs_for_generation(
+            input_ids,
+            next_sequence_length=next_sequence_length,
+            past_key_values=past_key_values,
+            attention_mask=attention_mask,
+            inputs_embeds=inputs_embeds,
+            is_first_iteration=is_first_iteration,
+            **kwargs,
+        )
+
+
+__all__ = [
+    "MossMusicEncoderConfig",
+    "MossMusicConfig",
+    "MossMusicModel",
+]

toolkit/audio/moss_music/processing.py

@@ -0,0 +1,250 @@
+import re
+from dataclasses import dataclass
+from typing import List, Optional, Sequence, Union
+
+import numpy as np
+import torch
+from transformers import BatchEncoding, Qwen2Tokenizer
+
+
+@dataclass
+class MelConfig:
+    mel_sr: int = 16000
+    mel_dim: int = 128
+    mel_n_fft: int = 400
+    mel_hop_length: int = 160
+    mel_dtype: torch.dtype = torch.bfloat16
+
+
+class MossMusicProcessor:
+    """Builds MOSS-Music inputs: Whisper log-mel features plus a Qwen chat
+    prompt whose audio span is one placeholder token per 80ms of audio, with
+    the elapsed seconds written in as digit tokens every 2s (time markers)."""
+
+    _AUDIO_SPAN_RE = re.compile(r"<\|audio_bos\|>(?:<\|AUDIO\|>)+<\|audio_eos\|>")
+
+    def __init__(
+        self,
+        tokenizer,
+        *,
+        mel_config: Optional[MelConfig] = None,
+        enable_time_marker: bool = True,
+        audio_token_id: int = 151654,
+        audio_start_id: int = 151669,
+        audio_end_id: int = 151670,
+    ):
+        self.tokenizer = tokenizer
+        self.audio_token_id = int(audio_token_id)
+        self.audio_start_id = int(audio_start_id)
+        self.audio_end_id = int(audio_end_id)
+        self.enable_time_marker = bool(enable_time_marker)
+        self.config = mel_config or MelConfig()
+        self._whisper_feature_extractor = None
+
+        self._digit_token_ids = {str(d): 15 + d for d in range(10)}
+        self.audio_tokens_per_second = 12.5
+        self.time_marker_every_seconds = 2
+        self.time_marker_every_audio_tokens = int(
+            self.audio_tokens_per_second * self.time_marker_every_seconds
+        )
+        self.model_input_names = [
+            "input_ids",
+            "attention_mask",
+            "audio_data",
+            "audio_data_seqlens",
+        ]
+
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path, **kwargs):
+        tokenizer_kwargs = {}
+        for key in ["cache_dir", "revision", "token", "local_files_only"]:
+            if key in kwargs:
+                tokenizer_kwargs[key] = kwargs[key]
+        # the checkpoint's tokenizer_config names Qwen2Tokenizer; loading it
+        # directly skips AutoTokenizer's remote-code prompt for the auto_map
+        tokenizer = Qwen2Tokenizer.from_pretrained(
+            pretrained_model_name_or_path, **tokenizer_kwargs
+        )
+        return cls(
+            tokenizer,
+            mel_config=kwargs.pop("mel_config", None),
+            enable_time_marker=kwargs.pop("enable_time_marker", True),
+            audio_token_id=kwargs.pop("audio_token_id", 151654),
+            audio_start_id=kwargs.pop("audio_start_id", 151669),
+            audio_end_id=kwargs.pop("audio_end_id", 151670),
+        )
+
+    @staticmethod
+    def _conv3_downsample_len(raw_mel_len: int) -> int:
+        def conv_out_len(length: int) -> int:
+            return (length - 1) // 2 + 1
+
+        return conv_out_len(conv_out_len(conv_out_len(int(raw_mel_len))))
+
+    def _get_whisper_feature_extractor(self):
+        if self._whisper_feature_extractor is not None:
+            return self._whisper_feature_extractor
+
+        from transformers.models.whisper.feature_extraction_whisper import (
+            WhisperFeatureExtractor,
+        )
+
+        self._whisper_feature_extractor = WhisperFeatureExtractor(
+            feature_size=int(self.config.mel_dim),
+            sampling_rate=int(self.config.mel_sr),
+            hop_length=int(self.config.mel_hop_length),
+            n_fft=int(self.config.mel_n_fft),
+        )
+        return self._whisper_feature_extractor
+
+    def _extract_mel(self, audio: Union[np.ndarray, torch.Tensor]) -> torch.Tensor:
+        if isinstance(audio, torch.Tensor):
+            audio = audio.detach().cpu().numpy()
+        wav_np = np.ascontiguousarray(audio, dtype=np.float32)
+        if wav_np.ndim == 2:
+            wav_np = wav_np[0]
+        # the raw fbank pass has no 30s chunking, so full songs come through whole
+        fe = self._get_whisper_feature_extractor()
+        feats = fe._np_extract_fbank_features(wav_np[None, ...], device="cpu")
+        return torch.from_numpy(feats[0]).to(dtype=self.config.mel_dtype)
+
+    def _get_time_marker_token_ids(self, second: int) -> List[int]:
+        return [self._digit_token_ids[digit] for digit in str(second)]
+
+    def _build_audio_tokens_with_time_markers(self, audio_seq_len: int) -> List[int]:
+        total_duration_seconds = audio_seq_len / self.audio_tokens_per_second
+        num_full_seconds = int(total_duration_seconds)
+
+        token_ids: List[int] = []
+        audio_tokens_consumed = 0
+        for second in range(
+            self.time_marker_every_seconds,
+            num_full_seconds + 1,
+            self.time_marker_every_seconds,
+        ):
+            marker_pos = (
+                second // self.time_marker_every_seconds
+            ) * self.time_marker_every_audio_tokens
+            audio_segment_len = marker_pos - audio_tokens_consumed
+            if audio_segment_len > 0:
+                token_ids.extend([self.audio_token_id] * audio_segment_len)
+                audio_tokens_consumed += audio_segment_len
+            token_ids.extend(self._get_time_marker_token_ids(second))
+
+        remaining = audio_seq_len - audio_tokens_consumed
+        if remaining > 0:
+            token_ids.extend([self.audio_token_id] * remaining)
+        return token_ids
+
+    def _build_audio_placeholder_ids(self, num_audio_tokens: int) -> List[int]:
+        if self.enable_time_marker:
+            return self._build_audio_tokens_with_time_markers(num_audio_tokens)
+        return [self.audio_token_id] * num_audio_tokens
+
+    def _build_default_prompt(self, text: str, has_audio: bool) -> str:
+        if has_audio:
+            return (
+                "<|im_start|>system\n"
+                "You are a helpful assistant.<|im_end|>\n"
+                "<|im_start|>user\n"
+                "<|audio_bos|><|AUDIO|><|audio_eos|>\n"
+                f"{text}<|im_end|>\n"
+                "<|im_start|>assistant\n"
+            )
+        return (
+            "<|im_start|>system\n"
+            "You are a helpful assistant.<|im_end|>\n"
+            "<|im_start|>user\n"
+            f"{text}<|im_end|>\n"
+            "<|im_start|>assistant\n"
+        )
+
+    def _build_input_from_prompt(self, prompt: str, token_lens: List[int]) -> List[int]:
+        spans = list(self._AUDIO_SPAN_RE.finditer(prompt))
+        if len(spans) != len(token_lens):
+            raise ValueError(
+                f"Audio placeholder count mismatch: found {len(spans)} spans in text, "
+                f"but got {len(token_lens)} audio inputs."
+            )
+
+        input_ids: List[int] = []
+        cursor = 0
+        for index, match in enumerate(spans):
+            prefix = prompt[cursor : match.start()]
+            if prefix:
+                input_ids.extend(self.tokenizer.encode(prefix, add_special_tokens=False))
+
+            input_ids.append(self.audio_start_id)
+            input_ids.extend(self._build_audio_placeholder_ids(int(token_lens[index])))
+            input_ids.append(self.audio_end_id)
+            cursor = match.end()
+
+        suffix = prompt[cursor:]
+        if suffix:
+            input_ids.extend(self.tokenizer.encode(suffix, add_special_tokens=False))
+        return input_ids
+
+    def __call__(
+        self,
+        *,
+        text: Union[str, Sequence[str]],
+        audios: Optional[Sequence[Union[np.ndarray, torch.Tensor]]] = None,
+        audio: Optional[Sequence[Union[np.ndarray, torch.Tensor]]] = None,
+        return_tensors: str = "pt",
+        **kwargs,
+    ):
+        if isinstance(text, (list, tuple)):
+            if len(text) != 1:
+                raise ValueError(f"Expected text batch size 1, got {len(text)}")
+            prompt_text = text[0]
+        else:
+            prompt_text = text
+
+        audio_list = audios if audios is not None else audio
+        audio_list = [] if audio_list is None else list(audio_list)
+
+        mels: List[torch.Tensor] = []
+        raw_lengths: List[int] = []
+        token_lens: List[int] = []
+        for one_audio in audio_list:
+            mel = self._extract_mel(one_audio)
+            raw_len = int(mel.shape[-1])
+            mels.append(mel)
+            raw_lengths.append(raw_len)
+            token_lens.append(self._conv3_downsample_len(raw_len))
+
+        if mels:
+            max_length = max(raw_lengths)
+            audio_batch = torch.zeros(
+                (len(mels), self.config.mel_dim, max_length),
+                dtype=self.config.mel_dtype,
+            )
+            for index, mel in enumerate(mels):
+                audio_batch[index, :, : mel.shape[-1]] = mel
+            seqlens_tensor = torch.tensor(raw_lengths, dtype=torch.long)
+        else:
+            audio_batch = None
+            seqlens_tensor = None
+
+        if self._AUDIO_SPAN_RE.search(prompt_text) is None:
+            prompt_text = self._build_default_prompt(prompt_text, has_audio=bool(audio_list))
+        input_ids_list = self._build_input_from_prompt(prompt_text, token_lens)
+
+        input_ids_tensor = torch.tensor([input_ids_list], dtype=torch.long)
+        data = {
+            "input_ids": input_ids_tensor,
+            "attention_mask": torch.ones_like(input_ids_tensor),
+        }
+        if audio_batch is not None:
+            data["audio_data"] = audio_batch
+            data["audio_data_seqlens"] = seqlens_tensor
+        return BatchEncoding(data=data, tensor_type=return_tensors)
+
+    def batch_decode(self, *args, **kwargs):
+        return self.tokenizer.batch_decode(*args, **kwargs)
+
+    def decode(self, *args, **kwargs):
+        return self.tokenizer.decode(*args, **kwargs)
+
+
+__all__ = ["MelConfig", "MossMusicProcessor"]