From f6e309dd4df068270946ccca7e1db1ad4f1c9c07 Mon Sep 17 00:00:00 2001 From: Ettore Di Giacinto Date: Fri, 17 Jan 2025 16:29:34 +0100 Subject: [PATCH] Some adaptations for the MusicgenForConditionalGeneration type Signed-off-by: Ettore Di Giacinto --- backend/python/transformers/backend.py | 23 +++++++++++++---------- backend/python/transformers/test.py | 4 +--- 2 files changed, 14 insertions(+), 13 deletions(-) diff --git a/backend/python/transformers/backend.py b/backend/python/transformers/backend.py index 05f6d7fe..3f6838ad 100644 --- a/backend/python/transformers/backend.py +++ b/backend/python/transformers/backend.py @@ -206,19 +206,22 @@ class BackendServicer(backend_pb2_grpc.BackendServicer): torch_dtype=compute) if request.ContextSize > 0: self.max_tokens = request.ContextSize - else: + elif request.Type != "MusicgenForConditionalGeneration": self.max_tokens = self.model.config.max_position_embeddings + else: + self.max_tokens = 512 - self.tokenizer = AutoTokenizer.from_pretrained(model_name, use_safetensors=True) - self.XPU = False + if request.Type != "MusicgenForConditionalGeneration": + self.tokenizer = AutoTokenizer.from_pretrained(model_name, use_safetensors=True) + self.XPU = False - if XPU and self.OV == False: - self.XPU = True - try: - print("Optimizing model", model_name, "to XPU.", file=sys.stderr) - self.model = ipex.optimize_transformers(self.model, inplace=True, dtype=torch.float16, device="xpu") - except Exception as err: - print("Not using XPU:", err, file=sys.stderr) + if XPU and self.OV == False: + self.XPU = True + try: + print("Optimizing model", model_name, "to XPU.", file=sys.stderr) + self.model = ipex.optimize_transformers(self.model, inplace=True, dtype=torch.float16, device="xpu") + except Exception as err: + print("Not using XPU:", err, file=sys.stderr) except Exception as err: print("Error:", err, file=sys.stderr) diff --git a/backend/python/transformers/test.py b/backend/python/transformers/test.py index c5f31932..305b0a93 100644 --- a/backend/python/transformers/test.py +++ b/backend/python/transformers/test.py @@ -19,6 +19,7 @@ class TestBackendServicer(unittest.TestCase): This method sets up the gRPC service by starting the server """ self.service = subprocess.Popen(["python3", "backend.py", "--addr", "localhost:50051"]) + time.sleep(10) def tearDown(self) -> None: """ @@ -31,7 +32,6 @@ class TestBackendServicer(unittest.TestCase): """ This method tests if the server starts up successfully """ - time.sleep(10) try: self.setUp() with grpc.insecure_channel("localhost:50051") as channel: @@ -48,7 +48,6 @@ class TestBackendServicer(unittest.TestCase): """ This method tests if the model is loaded successfully """ - time.sleep(10) try: self.setUp() with grpc.insecure_channel("localhost:50051") as channel: @@ -66,7 +65,6 @@ class TestBackendServicer(unittest.TestCase): """ This method tests if the embeddings are generated successfully """ - time.sleep(10) try: self.setUp() with grpc.insecure_channel("localhost:50051") as channel: