Feat: new backend: transformers-musicgen (#1387)

Transformers-MusicGen
---------

Signed-off-by: Dave <dave@gray101.com>

.gitignore

@@ -3,6 +3,7 @@
 __pycache__/
 *.a
 get-sources
+prepare-sources
 /backend/cpp/llama/grpc-server
 /backend/cpp/llama/llama.cpp
 

Dockerfile

@@ -12,7 +12,9 @@ ARG TARGETARCH
 ARG TARGETVARIANT
 
 ENV BUILD_TYPE=${BUILD_TYPE}
-ENV EXTERNAL_GRPC_BACKENDS="huggingface-embeddings:/build/backend/python/sentencetransformers/run.sh,petals:/build/backend/python/petals/run.sh,transformers:/build/backend/python/transformers/run.sh,sentencetransformers:/build/backend/python/sentencetransformers/run.sh,autogptq:/build/backend/python/autogptq/run.sh,bark:/build/backend/python/bark/run.sh,diffusers:/build/backend/python/diffusers/run.sh,exllama:/build/backend/python/exllama/run.sh,vall-e-x:/build/backend/python/vall-e-x/run.sh,vllm:/build/backend/python/vllm/run.sh,exllama2:/build/backend/python/exllama2/run.sh"
+
+ENV EXTERNAL_GRPC_BACKENDS="huggingface-embeddings:/build/backend/python/sentencetransformers/run.sh,petals:/build/backend/python/petals/run.sh,transformers:/build/backend/python/transformers/run.sh,sentencetransformers:/build/backend/python/sentencetransformers/run.sh,autogptq:/build/backend/python/autogptq/run.sh,bark:/build/backend/python/bark/run.sh,diffusers:/build/backend/python/diffusers/run.sh,exllama:/build/backend/python/exllama/run.sh,vall-e-x:/build/backend/python/vall-e-x/run.sh,vllm:/build/backend/python/vllm/run.sh,exllama2:/build/backend/python/exllama2/run.sh,transformers-musicgen:/build/backend/python/transformers-musicgen/run.sh"
+
 ENV GALLERIES='[{"name":"model-gallery", "url":"github:go-skynet/model-gallery/index.yaml"}, {"url": "github:go-skynet/model-gallery/huggingface.yaml","name":"huggingface"}]'
 ARG GO_TAGS="stablediffusion tts"
 
@@ -187,6 +189,9 @@ RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
 RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
 	PATH=$PATH:/opt/conda/bin make -C backend/python/petals \
     ; fi
+RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
+	PATH=$PATH:/opt/conda/bin make -C backend/python/transformers-musicgen \
+    ; fi
 
 # Define the health check command
 HEALTHCHECK --interval=1m --timeout=10m --retries=10 \

Makefile

@@ -389,6 +389,7 @@ protogen-go:
 protogen-python:
 	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/sentencetransformers/ --grpc_python_out=backend/python/sentencetransformers/ backend/backend.proto
 	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/transformers/ --grpc_python_out=backend/python/transformers/ backend/backend.proto
+	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/transformers-musicgen/ --grpc_python_out=backend/python/transformers-musicgen/ backend/backend.proto
 	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/autogptq/ --grpc_python_out=backend/python/autogptq/ backend/backend.proto
 	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/exllama/ --grpc_python_out=backend/python/exllama/ backend/backend.proto
 	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/bark/ --grpc_python_out=backend/python/bark/ backend/backend.proto
@@ -407,6 +408,7 @@ prepare-extra-conda-environments:
 	$(MAKE) -C backend/python/vllm
 	$(MAKE) -C backend/python/sentencetransformers
 	$(MAKE) -C backend/python/transformers
+	$(MAKE) -C backend/python/transformers-musicgen
 	$(MAKE) -C backend/python/vall-e-x
 	$(MAKE) -C backend/python/exllama
 	$(MAKE) -C backend/python/petals

api/backend/tts.go

@@ -59,10 +59,14 @@ func ModelTTS(backend, text, modelFile string, loader *model.ModelLoader, o *opt
 	// If the model file is not empty, we pass it joined with the model path
 	modelPath := ""
 	if modelFile != "" {
+		if bb != model.TransformersMusicGen {
 			modelPath = filepath.Join(o.Loader.ModelPath, modelFile)
 			if err := utils.VerifyPath(modelPath, o.Loader.ModelPath); err != nil {
 				return "", nil, err
 			}
+		} else {
+			modelPath = modelFile
+		}
 	}
 
 	res, err := piperModel.TTS(context.Background(), &proto.TTSRequest{

backend/python/autogptq/backend_pb2.py

@@ -19,7 +19,6 @@ _globals = globals()
 _builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
 _builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
 if _descriptor._USE_C_DESCRIPTORS == False:
-
   DESCRIPTOR._options = None
   DESCRIPTOR._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
   _MEMORYUSAGEDATA_BREAKDOWNENTRY._options = None

backend/python/bark/backend_pb2.py

@@ -19,7 +19,6 @@ _globals = globals()
 _builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
 _builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
 if _descriptor._USE_C_DESCRIPTORS == False:
-
   DESCRIPTOR._options = None
   DESCRIPTOR._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
   _MEMORYUSAGEDATA_BREAKDOWNENTRY._options = None

backend/python/diffusers/backend_pb2.py

@@ -19,7 +19,6 @@ _globals = globals()
 _builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
 _builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
 if _descriptor._USE_C_DESCRIPTORS == False:
-
   DESCRIPTOR._options = None
   DESCRIPTOR._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
   _MEMORYUSAGEDATA_BREAKDOWNENTRY._options = None

backend/python/exllama/backend_pb2.py

@@ -19,7 +19,6 @@ _globals = globals()
 _builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
 _builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
 if _descriptor._USE_C_DESCRIPTORS == False:
-
   DESCRIPTOR._options = None
   DESCRIPTOR._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
   _MEMORYUSAGEDATA_BREAKDOWNENTRY._options = None

backend/python/petals/backend_pb2.py

@@ -19,7 +19,6 @@ _globals = globals()
 _builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
 _builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
 if _descriptor._USE_C_DESCRIPTORS == False:
-
   DESCRIPTOR._options = None
   DESCRIPTOR._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
   _MEMORYUSAGEDATA_BREAKDOWNENTRY._options = None

backend/python/sentencetransformers/backend_pb2.py

@@ -19,7 +19,6 @@ _globals = globals()
 _builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
 _builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
 if _descriptor._USE_C_DESCRIPTORS == False:
-
   DESCRIPTOR._options = None
   DESCRIPTOR._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
   _MEMORYUSAGEDATA_BREAKDOWNENTRY._options = None

backend/python/transformers-musicgen/Makefile

@@ -0,0 +1,25 @@
+
+TRANSFORMERS_MUSICGEN_CONDA_PATH = "transformers-musicgen.yml"
+
+ifeq ($(BUILD_TYPE), cublas)
+	TRANSFORMERS_MUSICGEN_CONDA_PATH = "transformers-musicgen-nvidia.yml"
+endif
+
+.PHONY: transformers-musicgen
+transformers-musicgen:
+	@echo "Creating virtual environment..."
+	@conda env create --name transformers-musicgen --file $(TRANSFORMERS_MUSICGEN_CONDA_PATH)
+	@echo "Virtual environment created."
+
+.PHONY: run
+run:
+	@echo "Running transformers..."
+	bash run.sh
+	@echo "transformers run."
+
+# It is not working well by using command line. It only6 works with IDE like VSCode.
+.PHONY: test
+test:
+	@echo "Testing transformers..."
+	bash test.sh
+	@echo "transformers tested."

backend/python/transformers-musicgen/README.md

@@ -0,0 +1,5 @@
+# Creating a separate environment for the transformers project
+
+```
+make transformers-musicgen
+```

backend/python/transformers-musicgen/backend_pb2_grpc.py

@@ -0,0 +1,363 @@
+# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
+"""Client and server classes corresponding to protobuf-defined services."""
+import grpc
+
+import backend_pb2 as backend__pb2
+
+
+class BackendStub(object):
+    """Missing associated documentation comment in .proto file."""
+
+    def __init__(self, channel):
+        """Constructor.
+
+        Args:
+            channel: A grpc.Channel.
+        """
+        self.Health = channel.unary_unary(
+                '/backend.Backend/Health',
+                request_serializer=backend__pb2.HealthMessage.SerializeToString,
+                response_deserializer=backend__pb2.Reply.FromString,
+                )
+        self.Predict = channel.unary_unary(
+                '/backend.Backend/Predict',
+                request_serializer=backend__pb2.PredictOptions.SerializeToString,
+                response_deserializer=backend__pb2.Reply.FromString,
+                )
+        self.LoadModel = channel.unary_unary(
+                '/backend.Backend/LoadModel',
+                request_serializer=backend__pb2.ModelOptions.SerializeToString,
+                response_deserializer=backend__pb2.Result.FromString,
+                )
+        self.PredictStream = channel.unary_stream(
+                '/backend.Backend/PredictStream',
+                request_serializer=backend__pb2.PredictOptions.SerializeToString,
+                response_deserializer=backend__pb2.Reply.FromString,
+                )
+        self.Embedding = channel.unary_unary(
+                '/backend.Backend/Embedding',
+                request_serializer=backend__pb2.PredictOptions.SerializeToString,
+                response_deserializer=backend__pb2.EmbeddingResult.FromString,
+                )
+        self.GenerateImage = channel.unary_unary(
+                '/backend.Backend/GenerateImage',
+                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
+                response_deserializer=backend__pb2.Result.FromString,
+                )
+        self.AudioTranscription = channel.unary_unary(
+                '/backend.Backend/AudioTranscription',
+                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
+                response_deserializer=backend__pb2.TranscriptResult.FromString,
+                )
+        self.TTS = channel.unary_unary(
+                '/backend.Backend/TTS',
+                request_serializer=backend__pb2.TTSRequest.SerializeToString,
+                response_deserializer=backend__pb2.Result.FromString,
+                )
+        self.TokenizeString = channel.unary_unary(
+                '/backend.Backend/TokenizeString',
+                request_serializer=backend__pb2.PredictOptions.SerializeToString,
+                response_deserializer=backend__pb2.TokenizationResponse.FromString,
+                )
+        self.Status = channel.unary_unary(
+                '/backend.Backend/Status',
+                request_serializer=backend__pb2.HealthMessage.SerializeToString,
+                response_deserializer=backend__pb2.StatusResponse.FromString,
+                )
+
+
+class BackendServicer(object):
+    """Missing associated documentation comment in .proto file."""
+
+    def Health(self, request, context):
+        """Missing associated documentation comment in .proto file."""
+        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
+        context.set_details('Method not implemented!')
+        raise NotImplementedError('Method not implemented!')
+
+    def Predict(self, request, context):
+        """Missing associated documentation comment in .proto file."""
+        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
+        context.set_details('Method not implemented!')
+        raise NotImplementedError('Method not implemented!')
+
+    def LoadModel(self, request, context):
+        """Missing associated documentation comment in .proto file."""
+        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
+        context.set_details('Method not implemented!')
+        raise NotImplementedError('Method not implemented!')
+
+    def PredictStream(self, request, context):
+        """Missing associated documentation comment in .proto file."""
+        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
+        context.set_details('Method not implemented!')
+        raise NotImplementedError('Method not implemented!')
+
+    def Embedding(self, request, context):
+        """Missing associated documentation comment in .proto file."""
+        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
+        context.set_details('Method not implemented!')
+        raise NotImplementedError('Method not implemented!')
+
+    def GenerateImage(self, request, context):
+        """Missing associated documentation comment in .proto file."""
+        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
+        context.set_details('Method not implemented!')
+        raise NotImplementedError('Method not implemented!')
+
+    def AudioTranscription(self, request, context):
+        """Missing associated documentation comment in .proto file."""
+        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
+        context.set_details('Method not implemented!')
+        raise NotImplementedError('Method not implemented!')
+
+    def TTS(self, request, context):
+        """Missing associated documentation comment in .proto file."""
+        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
+        context.set_details('Method not implemented!')
+        raise NotImplementedError('Method not implemented!')
+
+    def TokenizeString(self, request, context):
+        """Missing associated documentation comment in .proto file."""
+        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
+        context.set_details('Method not implemented!')
+        raise NotImplementedError('Method not implemented!')
+
+    def Status(self, request, context):
+        """Missing associated documentation comment in .proto file."""
+        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
+        context.set_details('Method not implemented!')
+        raise NotImplementedError('Method not implemented!')
+
+
+def add_BackendServicer_to_server(servicer, server):
+    rpc_method_handlers = {
+            'Health': grpc.unary_unary_rpc_method_handler(
+                    servicer.Health,
+                    request_deserializer=backend__pb2.HealthMessage.FromString,
+                    response_serializer=backend__pb2.Reply.SerializeToString,
+            ),
+            'Predict': grpc.unary_unary_rpc_method_handler(
+                    servicer.Predict,
+                    request_deserializer=backend__pb2.PredictOptions.FromString,
+                    response_serializer=backend__pb2.Reply.SerializeToString,
+            ),
+            'LoadModel': grpc.unary_unary_rpc_method_handler(
+                    servicer.LoadModel,
+                    request_deserializer=backend__pb2.ModelOptions.FromString,
+                    response_serializer=backend__pb2.Result.SerializeToString,
+            ),
+            'PredictStream': grpc.unary_stream_rpc_method_handler(
+                    servicer.PredictStream,
+                    request_deserializer=backend__pb2.PredictOptions.FromString,
+                    response_serializer=backend__pb2.Reply.SerializeToString,
+            ),
+            'Embedding': grpc.unary_unary_rpc_method_handler(
+                    servicer.Embedding,
+                    request_deserializer=backend__pb2.PredictOptions.FromString,
+                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
+            ),
+            'GenerateImage': grpc.unary_unary_rpc_method_handler(
+                    servicer.GenerateImage,
+                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
+                    response_serializer=backend__pb2.Result.SerializeToString,
+            ),
+            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
+                    servicer.AudioTranscription,
+                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
+                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
+            ),
+            'TTS': grpc.unary_unary_rpc_method_handler(
+                    servicer.TTS,
+                    request_deserializer=backend__pb2.TTSRequest.FromString,
+                    response_serializer=backend__pb2.Result.SerializeToString,
+            ),
+            'TokenizeString': grpc.unary_unary_rpc_method_handler(
+                    servicer.TokenizeString,
+                    request_deserializer=backend__pb2.PredictOptions.FromString,
+                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
+            ),
+            'Status': grpc.unary_unary_rpc_method_handler(
+                    servicer.Status,
+                    request_deserializer=backend__pb2.HealthMessage.FromString,
+                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
+            ),
+    }
+    generic_handler = grpc.method_handlers_generic_handler(
+            'backend.Backend', rpc_method_handlers)
+    server.add_generic_rpc_handlers((generic_handler,))
+
+
+ # This class is part of an EXPERIMENTAL API.
+class Backend(object):
+    """Missing associated documentation comment in .proto file."""
+
+    @staticmethod
+    def Health(request,
+            target,
+            options=(),
+            channel_credentials=None,
+            call_credentials=None,
+            insecure=False,
+            compression=None,
+            wait_for_ready=None,
+            timeout=None,
+            metadata=None):
+        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
+            backend__pb2.HealthMessage.SerializeToString,
+            backend__pb2.Reply.FromString,
+            options, channel_credentials,
+            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
+
+    @staticmethod
+    def Predict(request,
+            target,
+            options=(),
+            channel_credentials=None,
+            call_credentials=None,
+            insecure=False,
+            compression=None,
+            wait_for_ready=None,
+            timeout=None,
+            metadata=None):
+        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
+            backend__pb2.PredictOptions.SerializeToString,
+            backend__pb2.Reply.FromString,
+            options, channel_credentials,
+            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
+
+    @staticmethod
+    def LoadModel(request,
+            target,
+            options=(),
+            channel_credentials=None,
+            call_credentials=None,
+            insecure=False,
+            compression=None,
+            wait_for_ready=None,
+            timeout=None,
+            metadata=None):
+        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
+            backend__pb2.ModelOptions.SerializeToString,
+            backend__pb2.Result.FromString,
+            options, channel_credentials,
+            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
+
+    @staticmethod
+    def PredictStream(request,
+            target,
+            options=(),
+            channel_credentials=None,
+            call_credentials=None,
+            insecure=False,
+            compression=None,
+            wait_for_ready=None,
+            timeout=None,
+            metadata=None):
+        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
+            backend__pb2.PredictOptions.SerializeToString,
+            backend__pb2.Reply.FromString,
+            options, channel_credentials,
+            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
+
+    @staticmethod
+    def Embedding(request,
+            target,
+            options=(),
+            channel_credentials=None,
+            call_credentials=None,
+            insecure=False,
+            compression=None,
+            wait_for_ready=None,
+            timeout=None,
+            metadata=None):
+        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
+            backend__pb2.PredictOptions.SerializeToString,
+            backend__pb2.EmbeddingResult.FromString,
+            options, channel_credentials,
+            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
+
+    @staticmethod
+    def GenerateImage(request,
+            target,
+            options=(),
+            channel_credentials=None,
+            call_credentials=None,
+            insecure=False,
+            compression=None,
+            wait_for_ready=None,
+            timeout=None,
+            metadata=None):
+        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
+            backend__pb2.GenerateImageRequest.SerializeToString,
+            backend__pb2.Result.FromString,
+            options, channel_credentials,
+            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
+
+    @staticmethod
+    def AudioTranscription(request,
+            target,
+            options=(),
+            channel_credentials=None,
+            call_credentials=None,
+            insecure=False,
+            compression=None,
+            wait_for_ready=None,
+            timeout=None,
+            metadata=None):
+        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
+            backend__pb2.TranscriptRequest.SerializeToString,
+            backend__pb2.TranscriptResult.FromString,
+            options, channel_credentials,
+            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
+
+    @staticmethod
+    def TTS(request,
+            target,
+            options=(),
+            channel_credentials=None,
+            call_credentials=None,
+            insecure=False,
+            compression=None,
+            wait_for_ready=None,
+            timeout=None,
+            metadata=None):
+        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
+            backend__pb2.TTSRequest.SerializeToString,
+            backend__pb2.Result.FromString,
+            options, channel_credentials,
+            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
+
+    @staticmethod
+    def TokenizeString(request,
+            target,
+            options=(),
+            channel_credentials=None,
+            call_credentials=None,
+            insecure=False,
+            compression=None,
+            wait_for_ready=None,
+            timeout=None,
+            metadata=None):
+        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
+            backend__pb2.PredictOptions.SerializeToString,
+            backend__pb2.TokenizationResponse.FromString,
+            options, channel_credentials,
+            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
+
+    @staticmethod
+    def Status(request,
+            target,
+            options=(),
+            channel_credentials=None,
+            call_credentials=None,
+            insecure=False,
+            compression=None,
+            wait_for_ready=None,
+            timeout=None,
+            metadata=None):
+        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
+            backend__pb2.HealthMessage.SerializeToString,
+            backend__pb2.StatusResponse.FromString,
+            options, channel_credentials,
+            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)

backend/python/transformers-musicgen/run.sh

@@ -0,0 +1,16 @@
+#!/bin/bash
+
+##
+## A bash script wrapper that runs the transformers-musicgen server with conda
+
+echo "Launching gRPC server for transformers-musicgen"
+
+
+# Activate conda environment
+source $CONDA_PREFIX/etc/profile.d/conda.sh
+conda activate transformers-musicgen
+
+# get the directory where the bash script is located
+DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+python $DIR/transformers_server.py $@

backend/python/transformers-musicgen/test.sh

@@ -0,0 +1,11 @@
+#!/bin/bash
+##
+## A bash script wrapper that runs the transformers server with conda
+
+# Activate conda environment
+source conda activate transformers-musicgen
+
+# get the directory where the bash script is located
+DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+python -m unittest $DIR/test_transformers_server.py

backend/python/transformers-musicgen/test_transformers.py

@@ -0,0 +1,81 @@
+"""
+A test script to test the gRPC service
+"""
+import unittest
+import subprocess
+import time
+import backend_pb2
+import backend_pb2_grpc
+
+import grpc
+
+
+class TestBackendServicer(unittest.TestCase):
+    """
+    TestBackendServicer is the class that tests the gRPC service
+    """
+    def setUp(self):
+        """
+        This method sets up the gRPC service by starting the server
+        """
+        self.service = subprocess.Popen(["python3", "transformers_server.py", "--addr", "localhost:50051"])
+
+    def tearDown(self) -> None:
+        """
+        This method tears down the gRPC service by terminating the server
+        """
+        self.service.terminate()
+        self.service.wait()
+
+    def test_server_startup(self):
+        """
+        This method tests if the server starts up successfully
+        """
+        time.sleep(2)
+        try:
+            self.setUp()
+            with grpc.insecure_channel("localhost:50051") as channel:
+                stub = backend_pb2_grpc.BackendStub(channel)
+                response = stub.Health(backend_pb2.HealthMessage())
+                self.assertEqual(response.message, b'OK')
+        except Exception as err:
+            print(err)
+            self.fail("Server failed to start")
+        finally:
+            self.tearDown()
+
+    def test_load_model(self):
+        """
+        This method tests if the model is loaded successfully
+        """
+        try:
+            self.setUp()
+            with grpc.insecure_channel("localhost:50051") as channel:
+                stub = backend_pb2_grpc.BackendStub(channel)
+                response = stub.LoadModel(backend_pb2.ModelOptions(Model="facebook/musicgen-small"))
+                self.assertTrue(response.success)
+                self.assertEqual(response.message, "Model loaded successfully")
+        except Exception as err:
+            print(err)
+            self.fail("LoadModel service failed")
+        finally:
+            self.tearDown()
+
+    def test_tts(self):
+        """
+        This method tests if the embeddings are generated successfully
+        """
+        try:
+            self.setUp()
+            with grpc.insecure_channel("localhost:50051") as channel:
+                stub = backend_pb2_grpc.BackendStub(channel)
+                response = stub.LoadModel(backend_pb2.ModelOptions(Model="facebook/musicgen-small"))
+                self.assertTrue(response.success)
+                tts_request = backend_pb2.TTSRequest(Model="facebook/musicgen-small", Input="80s TV news production music hit for tonight's biggest story")
+                tts_response = stub.TTS(tts_request)
+                self.assertIsNotNone(tts_response)
+        except Exception as err:
+            print(err)
+            self.fail("TTS service failed")
+        finally:
+            self.tearDown()

backend/python/transformers-musicgen/transformers-musicgen-nvidia.yml

@@ -0,0 +1,71 @@
+name: transformers-musicgen
+channels:
+  - defaults
+dependencies:
+  - bzip2=1.0.8
+  - ca-certificates=2023.08.22
+  - libffi=3.4.4
+  - libuuid=1.41.5
+  - ncurses=6.4
+  - openssl=3.0.11
+  - pip=23.2.1
+  - python=3.11.5
+  - readline=8.2
+  - setuptools=68.0.0
+  - sqlite=3.41.2
+  - tk=8.6.12
+  - tzdata=2023c
+  - wheel=0.41.2
+  - xz=5.4.2
+  - zlib=1.2.13
+  - pip:
+      - certifi==2023.7.22
+      - charset-normalizer==3.3.0
+      - click==8.1.7
+      - filelock==3.12.4
+      - fsspec==2023.9.2
+      - grpcio==1.59.0
+      - huggingface-hub==0.17.3
+      - idna==3.4
+      - install==1.3.5
+      - jinja2==3.1.2
+      - joblib==1.3.2
+      - markupsafe==2.1.3
+      - mpmath==1.3.0
+      - networkx==3.1
+      - nltk==3.8.1
+      - numpy==1.26.0
+      - nvidia-cublas-cu12==12.1.3.1
+      - nvidia-cuda-cupti-cu12==12.1.105
+      - nvidia-cuda-nvrtc-cu12==12.1.105
+      - nvidia-cuda-runtime-cu12==12.1.105
+      - nvidia-cudnn-cu12==8.9.2.26
+      - nvidia-cufft-cu12==11.0.2.54
+      - nvidia-curand-cu12==10.3.2.106
+      - nvidia-cusolver-cu12==11.4.5.107
+      - nvidia-cusparse-cu12==12.1.0.106
+      - nvidia-nccl-cu12==2.18.1
+      - nvidia-nvjitlink-cu12==12.2.140
+      - nvidia-nvtx-cu12==12.1.105
+      - packaging==23.2
+      - pillow==10.0.1
+      - protobuf==4.24.4
+      - pyyaml==6.0.1
+      - regex==2023.10.3
+      - requests==2.31.0
+      - safetensors==0.4.0
+      - scikit-learn==1.3.1
+      - scipy==1.11.3
+      - sentence-transformers==2.2.2
+      - sentencepiece==0.1.99
+      - sympy==1.12
+      - threadpoolctl==3.2.0
+      - tokenizers==0.14.1
+      - torch==2.1.0
+      - torchvision==0.16.0
+      - tqdm==4.66.1
+      - transformers==4.34.0
+      - triton==2.1.0
+      - typing-extensions==4.8.0
+      - urllib3==2.0.6
+prefix: /opt/conda/envs/transformers-musicgen

backend/python/transformers-musicgen/transformers-musicgen.yml

@@ -0,0 +1,58 @@
+name: transformers-musicgen
+channels:
+  - defaults
+dependencies:
+  - bzip2=1.0.8
+  - ca-certificates=2023.08.22
+  - libffi=3.4.4
+  - libuuid=1.41.5
+  - ncurses=6.4
+  - openssl=3.0.11
+  - pip=23.2.1
+  - python=3.11.5
+  - readline=8.2
+  - setuptools=68.0.0
+  - sqlite=3.41.2
+  - tk=8.6.12
+  - tzdata=2023c
+  - wheel=0.41.2
+  - xz=5.4.2
+  - zlib=1.2.13
+  - pip:
+      - certifi==2023.7.22
+      - charset-normalizer==3.3.0
+      - click==8.1.7
+      - filelock==3.12.4
+      - fsspec==2023.9.2
+      - grpcio==1.59.0
+      - huggingface-hub==0.17.3
+      - idna==3.4
+      - install==1.3.5
+      - jinja2==3.1.2
+      - joblib==1.3.2
+      - markupsafe==2.1.3
+      - mpmath==1.3.0
+      - networkx==3.1
+      - nltk==3.8.1
+      - numpy==1.26.0
+      - packaging==23.2
+      - pillow==10.0.1
+      - protobuf==4.24.4
+      - pyyaml==6.0.1
+      - regex==2023.10.3
+      - requests==2.31.0
+      - safetensors==0.4.0
+      - scikit-learn==1.3.1
+      - scipy==1.11.3
+      - sentence-transformers==2.2.2
+      - sentencepiece==0.1.99
+      - sympy==1.12
+      - threadpoolctl==3.2.0
+      - tokenizers==0.14.1
+      - torch==2.1.0
+      - torchvision==0.16.0
+      - tqdm==4.66.1
+      - transformers==4.34.0
+      - typing-extensions==4.8.0
+      - urllib3==2.0.6
+prefix: /opt/conda/envs/transformers-musicgen

backend/python/transformers-musicgen/transformers_server.py

@@ -0,0 +1,122 @@
+#!/usr/bin/env python3
+"""
+Extra gRPC server for MusicgenForConditionalGeneration models.
+"""
+from concurrent import futures
+
+import argparse
+import signal
+import sys
+import os
+
+import time
+import backend_pb2
+import backend_pb2_grpc
+
+import grpc
+
+from scipy.io.wavfile import write as write_wav
+from transformers import AutoProcessor, MusicgenForConditionalGeneration
+
+_ONE_DAY_IN_SECONDS = 60 * 60 * 24
+
+# If MAX_WORKERS are specified in the environment use it, otherwise default to 1
+MAX_WORKERS = int(os.environ.get('PYTHON_GRPC_MAX_WORKERS', '1'))
+
+# Implement the BackendServicer class with the service methods
+class BackendServicer(backend_pb2_grpc.BackendServicer):
+    """
+    A gRPC servicer for the backend service.
+
+    This class implements the gRPC methods for the backend service, including Health, LoadModel, and Embedding.
+    """
+    def Health(self, request, context):
+        """
+        A gRPC method that returns the health status of the backend service.
+
+        Args:
+            request: A HealthRequest object that contains the request parameters.
+            context: A grpc.ServicerContext object that provides information about the RPC.
+
+        Returns:
+            A Reply object that contains the health status of the backend service.
+        """
+        return backend_pb2.Reply(message=bytes("OK", 'utf-8'))
+
+    def LoadModel(self, request, context):
+        """
+        A gRPC method that loads a model into memory.
+
+        Args:
+            request: A LoadModelRequest object that contains the request parameters.
+            context: A grpc.ServicerContext object that provides information about the RPC.
+
+        Returns:
+            A Result object that contains the result of the LoadModel operation.
+        """
+        model_name = request.Model
+        try:
+            self.processor = AutoProcessor.from_pretrained(model_name)
+            self.model = MusicgenForConditionalGeneration.from_pretrained(model_name)
+        except Exception as err:
+            return backend_pb2.Result(success=False, message=f"Unexpected {err=}, {type(err)=}")
+
+        return backend_pb2.Result(message="Model loaded successfully", success=True)
+
+    def TTS(self, request, context):
+        model_name = request.model
+        if model_name == "":
+            return backend_pb2.Result(success=False, message="request.model is required")
+        try:
+            self.processor = AutoProcessor.from_pretrained(model_name)
+            self.model = MusicgenForConditionalGeneration.from_pretrained(model_name)
+            inputs = self.processor(
+                text=[request.text],
+                padding=True,
+                return_tensors="pt",
+            )
+            tokens = 256
+            # TODO get tokens from request?
+            audio_values = self.model.generate(**inputs, max_new_tokens=tokens)
+            print("[transformers-musicgen] TTS generated!", file=sys.stderr)
+            sampling_rate = self.model.config.audio_encoder.sampling_rate
+            write_wav(request.dst, rate=sampling_rate, data=audio_values[0, 0].numpy())
+            print("[transformers-musicgen] TTS saved to", request.dst, file=sys.stderr)
+            print("[transformers-musicgen] TTS for", file=sys.stderr)
+            print(request, file=sys.stderr)
+        except Exception as err:
+            return backend_pb2.Result(success=False, message=f"Unexpected {err=}, {type(err)=}")
+        return backend_pb2.Result(success=True)
+
+
+def serve(address):
+    server = grpc.server(futures.ThreadPoolExecutor(max_workers=MAX_WORKERS))
+    backend_pb2_grpc.add_BackendServicer_to_server(BackendServicer(), server)
+    server.add_insecure_port(address)
+    server.start()
+    print("[transformers-musicgen] Server started. Listening on: " + address, file=sys.stderr)
+
+    # Define the signal handler function
+    def signal_handler(sig, frame):
+        print("[transformers-musicgen] Received termination signal. Shutting down...")
+        server.stop(0)
+        sys.exit(0)
+
+    # Set the signal handlers for SIGINT and SIGTERM
+    signal.signal(signal.SIGINT, signal_handler)
+    signal.signal(signal.SIGTERM, signal_handler)
+
+    try:
+        while True:
+            time.sleep(_ONE_DAY_IN_SECONDS)
+    except KeyboardInterrupt:
+        server.stop(0)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Run the gRPC server.")
+    parser.add_argument(
+        "--addr", default="localhost:50051", help="The address to bind the server to."
+    )
+    args = parser.parse_args()
+    print(f"[transformers-musicgen] startup: {args}", file=sys.stderr)
+    serve(args.addr)

backend/python/transformers/backend_pb2.py

@@ -19,7 +19,6 @@ _globals = globals()
 _builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
 _builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
 if _descriptor._USE_C_DESCRIPTORS == False:
-
   DESCRIPTOR._options = None
   DESCRIPTOR._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
   _MEMORYUSAGEDATA_BREAKDOWNENTRY._options = None

backend/python/vall-e-x/backend_pb2.py

@@ -19,7 +19,6 @@ _globals = globals()
 _builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
 _builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
 if _descriptor._USE_C_DESCRIPTORS == False:
-
   DESCRIPTOR._options = None
   DESCRIPTOR._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
   _MEMORYUSAGEDATA_BREAKDOWNENTRY._options = None

backend/python/vllm/backend_pb2.py

@@ -19,7 +19,6 @@ _globals = globals()
 _builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
 _builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
 if _descriptor._USE_C_DESCRIPTORS == False:
-
   DESCRIPTOR._options = None
   DESCRIPTOR._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
   _MEMORYUSAGEDATA_BREAKDOWNENTRY._options = None

docs/content/features/text-to-audio.md

@@ -52,6 +52,20 @@ Note:
 - The model name is case sensitive.
 - LocalAI must be compiled with the `GO_TAGS=tts` flag.
 
+LocalAI also has experimental support for `transformers-musicgen` for the generation of short musical compositions. Currently, this is implemented via the same requests used for text to speech:
+
+```
+curl --request POST \
+  --url http://localhost:8080/tts \
+  --header 'Content-Type: application/json' \
+  --data '{
+    "backend": "transformers-musicgen",
+    "model": "facebook/musicgen-medium",
+    "input": "Cello Rave"
+}' | aplay```
+
+Future versions of LocalAI will expose additional control over audio generation beyond the text prompt.
+
 #### Configuration
 
 Audio models can be configured via `YAML` files. This allows to configure specific setting for each backend. For instance, backends might be specifying a voice or supports voice cloning which must be specified in the configuration file.

examples/bruno/LocalAI Test Requests/tts/musicgen.bru

@@ -0,0 +1,23 @@
+meta {
+  name: musicgen
+  type: http
+  seq: 2
+}
+
+post {
+  url: {{PROTOCOL}}{{HOST}}:{{PORT}}/tts
+  body: json
+  auth: none
+}
+
+headers {
+  Content-Type: application/json
+}
+
+body:json {
+  {
+      "backend": "transformers-musicgen",
+      "model": "facebook/musicgen-small",
+      "input": "80s Synths playing Jazz"
+  }
+}

pkg/model/initializers.go

@@ -37,6 +37,9 @@ const (
 	StableDiffusionBackend = "stablediffusion"
 	PiperBackend           = "piper"
 	LCHuggingFaceBackend   = "langchain-huggingface"
+
+	// External Backends that need special handling within LocalAI:
+	TransformersMusicGen = "transformers-musicgen"
 )
 
 var AutoLoadBackends []string = []string{