fix dockerfile and install

Author: fxmarty

Dockerfile-rocm

@@ -111,6 +111,11 @@ ARG DEFAULT_USE_FLASH_ATTENTION=True
 COPY backends/python/Makefile-flash-att-v2 Makefile-flash-att-v2
 RUN make -f Makefile-flash-att-v2 install-flash-attention-v2-rocm
 
+# Install python backend
+COPY backends/python/server /tei_backends/python/server
+COPY backends/proto tei_backends/proto
+RUN make -C /tei_backends/python/server install
+
 ENV HUGGINGFACE_HUB_CACHE=/data \
     PORT=80 \
     USE_FLASH_ATTENTION=$DEFAULT_USE_FLASH_ATTENTION

backends/python/server/pyproject.toml

@@ -15,12 +15,13 @@ grpcio-status = "^1.51.1"
 grpcio-reflection = "^1.51.1"
 grpc-interceptor = "^0.15.0"
 typer = "^0.6.1"
-safetensors = "^0.3.2"
+safetensors = "^0.4.0"
 loguru = "^0.6.0"
 opentelemetry-api = "^1.15.0"
 opentelemetry-exporter-otlp = "^1.15.0"
 opentelemetry-instrumentation-grpc = "^0.36b0"
 torch = { version = "==2.3.1" }
+transformers = { version = "^4.39.0"}
 
 [tool.poetry.extras]