Add Flash Attention 2 support to Musicgen and Musicgen Melody

docs/source/en/perf_infer_gpu_one.md

@@ -55,6 +55,8 @@ FlashAttention-2 is currently supported for the following architectures:
 * [MBart](https://huggingface.co/docs/transformers/model_doc/mbart#transformers.MBartModel)
 * [Mistral](https://huggingface.co/docs/transformers/model_doc/mistral#transformers.MistralModel)
 * [Mixtral](https://huggingface.co/docs/transformers/model_doc/mixtral#transformers.MixtralModel)
+* [Musicgen](https://huggingface.co/docs/transformers/model_doc/musicgen#transformers.MusicgenModel)
+* [MusicGen Melody](https://huggingface.co/docs/transformers/model_doc/musicgen_melody#transformers.MusicgenMelodyModel)
 * [OPT](https://huggingface.co/docs/transformers/model_doc/opt#transformers.OPTModel)
 * [Phi](https://huggingface.co/docs/transformers/model_doc/phi#transformers.PhiModel)
 * [StableLm](https://huggingface.co/docs/transformers/model_doc/stablelm#transformers.StableLmModel)

src/transformers/models/deprecated/_archive_maps.py

@@ -1470,6 +1470,12 @@ def __getitem__(self, item):
 
 MUSICGEN_PRETRAINED_MODEL_ARCHIVE_LIST = DeprecatedList(["facebook/musicgen-small"])
 
+MUSICGEN_MELODY_PRETRAINED_CONFIG_ARCHIVE_MAP = DeprecatedDict(
+    {"facebook/musicgen-melody": "https://huggingface.co/facebook/musicgen-melody/resolve/main/config.json"}
+)
+
+MUSICGEN_MELODY_PRETRAINED_MODEL_ARCHIVE_LIST = DeprecatedList(["facebook/musicgen-melody"])
+
 MVP_PRETRAINED_MODEL_ARCHIVE_LIST = DeprecatedList(
     [
         "RUCAIBox/mvp",

src/transformers/models/musicgen/configuration_musicgen.py

@@ -239,3 +239,20 @@ def from_sub_models_config(
     # This is a property because you might want to change the codec model on the fly
     def sampling_rate(self):
         return self.audio_encoder.sampling_rate
+
+    @property
+    def _attn_implementation(self):
+        # This property is made private for now (as it cannot be changed and a PreTrainedModel.use_attn_implementation method needs to be implemented.)
+        if hasattr(self, "_attn_implementation_internal"):
+            if self._attn_implementation_internal is None:
+                # `config.attn_implementation` should never be None, for backward compatibility.
+                return "eager"
+            else:
+                return self._attn_implementation_internal
+        else:
+            return "eager"
+
+    @_attn_implementation.setter
+    def _attn_implementation(self, value):
+        self._attn_implementation_internal = value
+        self.decoder._attn_implementation = value

src/transformers/models/musicgen_melody/configuration_musicgen_melody.py

@@ -21,9 +21,7 @@
 
 logger = logging.get_logger(__name__)
 
-MUSICGEN_MELODY_PRETRAINED_CONFIG_ARCHIVE_MAP = {
-    "facebook/musicgen-melody": "https://huggingface.co/facebook/musicgen-melody/resolve/main/config.json",
-}
+from ..deprecated._archive_maps import MUSICGEN_MELODY_PRETRAINED_CONFIG_ARCHIVE_MAP  # noqa: F401, E402
 
 
 class MusicgenMelodyDecoderConfig(PretrainedConfig):
@@ -254,3 +252,20 @@ def from_sub_models_config(
     # This is a property because you might want to change the codec model on the fly
     def sampling_rate(self):
         return self.audio_encoder.sampling_rate
+
+    @property
+    def _attn_implementation(self):
+        # This property is made private for now (as it cannot be changed and a PreTrainedModel.use_attn_implementation method needs to be implemented.)
+        if hasattr(self, "_attn_implementation_internal"):
+            if self._attn_implementation_internal is None:
+                # `config.attn_implementation` should never be None, for backward compatibility.
+                return "eager"
+            else:
+                return self._attn_implementation_internal
+        else:
+            return "eager"
+
+    @_attn_implementation.setter
+    def _attn_implementation(self, value):
+        self._attn_implementation_internal = value
+        self.decoder._attn_implementation = value