fix(docker): remove unused dependencies (#205)

Dockerfile

@@ -154,7 +154,7 @@ COPY server/Makefile server/Makefile
 RUN cd server && \
     make gen-server && \
     pip install -r requirements.txt && \
-    pip install ".[bnb]" --no-cache-dir
+    pip install ".[bnb, accelerate]" --no-cache-dir
 
 # Install router
 COPY --from=builder /usr/src/target/release/text-generation-router /usr/local/bin/text-generation-router

server/Makefile

@@ -17,7 +17,10 @@ install-torch:
 install: gen-server install-torch install-transformers
 	pip install pip --upgrade
 	pip install -r requirements.txt
-	pip install -e . --no-cache-dir
+	pip install -e ".[bnb, accelerate]"
 
 run-dev:
-	SAFETENSORS_FAST_GPU=1 python -m torch.distributed.run --nproc_per_node=2 text_generation_server/cli.py serve bigscience/bloom-560m --sharded
\ No newline at end of file
+	SAFETENSORS_FAST_GPU=1 python -m torch.distributed.run --nproc_per_node=2 text_generation_server/cli.py serve bigscience/bloom-560m --sharded
+
+export-requirements:
+	poetry export -o requirements.txt -E bnb --without-hashes
\ No newline at end of file

server/pyproject.toml

@@ -15,8 +15,8 @@ grpcio-status = "^1.51.1"
 grpcio-reflection = "^1.51.1"
 grpc-interceptor = "^0.15.0"
 typer = "^0.6.1"
-accelerate = "^0.15.0"
-bitsandbytes = "^0.38.1"
+accelerate = { version = "^0.15.0", optional = true }
+bitsandbytes = { version = "^0.38.1", optional = true }
 safetensors = "^0.2.4"
 loguru = "^0.6.0"
 opentelemetry-api = "^1.15.0"
@@ -28,6 +28,7 @@ tokenizers = "0.13.3"
 huggingface-hub = {git = "https://github.com/huggingface/huggingface_hub.git", rev = "4f27b44ee536cd654e171c7f37478eaf1996cc3f"}
 
 [tool.poetry.extras]
+accelerate = ["accelerate"]
 bnb = ["bitsandbytes"]
 
 [tool.poetry.group.dev.dependencies]