Mindnlp打卡任务：mimi模型迁移

__init__.py

@@ -0,0 +1,25 @@
+# Copyright 2024 Your Company. All rights reserved.
+#
+# 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 mindnlp.utils import LazyModule
+from mindnlp.utils.import_utils import define_import_structure
+
+if TYPE_CHECKING:
+    from .configuration_mimi import *
+    from .modeling_mimi import *
+else:
+    import sys
+
+    _file = globals()["__file__"]
+    sys.modules[__name__] = LazyModule(__name__, _file, define_import_structure(_file), module_spec=__spec__)

configuration_mimi.py

@@ -0,0 +1,114 @@
+import math
+import numpy as np
+from mindnlp.configs import MindNLPConfig
+
+class MimiConfig(MindNLPConfig):
+    r"""
+    This is the configuration class to store the configuration of an [`MimiModel`].
+    """
+
+    model_type = "mimi"
+
+    def __init__(
+        self,
+        sampling_rate=24_000,
+        frame_rate=12.5,
+        audio_channels=1,
+        hidden_size=512,
+        num_filters=64,
+        num_residual_layers=1,
+        upsampling_ratios=None,
+        kernel_size=7,
+        last_kernel_size=3,
+        residual_kernel_size=3,
+        dilation_growth_rate=2,
+        use_causal_conv=True,
+        pad_mode="constant",
+        compress=2,
+        trim_right_ratio=1.0,
+        codebook_size=2048,
+        codebook_dim=256,
+        num_quantizers=32,
+        use_conv_shortcut=False,
+        vector_quantization_hidden_dimension=256,
+        num_semantic_quantizers=1,
+        upsample_groups=512,
+        num_hidden_layers=8,
+        intermediate_size=2048,
+        num_attention_heads=8,
+        num_key_value_heads=8,
+        head_dim=None,
+        hidden_act="gelu",
+        max_position_embeddings=8000,
+        initializer_range=0.02,
+        norm_eps=1e-5,
+        use_cache=False,
+        rope_theta=10000.0,
+        sliding_window=250,
+        attention_dropout=0.0,
+        layer_scale_initial_scale=0.01,
+        attention_bias=False,
+        **kwargs,
+    ):
+        # Parameters for audio processing
+        self.sampling_rate = sampling_rate
+        self.frame_rate = frame_rate
+        self.audio_channels = audio_channels
+
+        # Parameters for network architecture
+        self.hidden_size = hidden_size
+        self.num_filters = num_filters
+        self.num_residual_layers = num_residual_layers
+        self.upsampling_ratios = upsampling_ratios if upsampling_ratios else [8, 6, 5, 4]
+        self.kernel_size = kernel_size
+        self.last_kernel_size = last_kernel_size
+        self.residual_kernel_size = residual_kernel_size
+        self.dilation_growth_rate = dilation_growth_rate
+        self.use_causal_conv = use_causal_conv
+        self.pad_mode = pad_mode
+        self.compress = compress
+        self.trim_right_ratio = trim_right_ratio
+        self.codebook_size = codebook_size
+        self.codebook_dim = codebook_dim if codebook_dim is not None else hidden_size
+        self.num_quantizers = num_quantizers
+        self.use_conv_shortcut = use_conv_shortcut
+        self.vector_quantization_hidden_dimension = vector_quantization_hidden_dimension
+        self.upsample_groups = upsample_groups
+
+        # Parameters for transformer architecture
+        self.num_hidden_layers = num_hidden_layers
+        self.intermediate_size = intermediate_size
+        self.num_attention_heads = num_attention_heads
+        self.num_key_value_heads = num_key_value_heads
+        self.head_dim = head_dim or hidden_size // num_attention_heads
+        self.hidden_act = hidden_act
+        self.max_position_embeddings = max_position_embeddings
+        self.initializer_range = initializer_range
+        self.norm_eps = norm_eps
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.sliding_window = sliding_window
+        self.attention_dropout = attention_dropout
+        self.layer_scale_initial_scale = layer_scale_initial_scale
+        self.attention_bias = attention_bias
+
+        # Number of semantic quantizers validation
+        if num_semantic_quantizers >= self.num_quantizers:
+            raise ValueError(
+                f"The number of semantic quantizers should be lower than the total number of quantizers {self.num_quantizers}, but is currently {num_semantic_quantizers}."
+            )
+        self.num_semantic_quantizers = num_semantic_quantizers
+
+        super().__init__(**kwargs)
+
+    @property
+    def encodec_frame_rate(self) -> int:
+        hop_length = np.prod(self.upsampling_ratios)
+        return math.ceil(self.sampling_rate / hop_length)
+
+    @property
+    def num_codebooks(self) -> int:
+        # Alias to num_quantizers
+        return self.num_quantizers
+
+__all__ = ["MimiConfig"]