@@ -3,8 +3,8 @@ FROM nvidia/cuda:12.2.0-devel-ubuntu22.04 AS base-builder
33ENV SCCACHE=0.10.0
44ENV RUSTC_WRAPPER=/usr/local/bin/sccache
55ENV PATH="/root/.cargo/bin:${PATH}"
6- # aligned with `cargo-chef` version in `lukemathwalker/cargo-chef:latest-rust-1.85 -bookworm`
7- ENV CARGO_CHEF=0.1.71
6+ # aligned with `cargo-chef` version in `lukemathwalker/cargo-chef:latest-rust-1.92 -bookworm`
7+ ENV CARGO_CHEF=0.1.73
88
99RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
1010 curl \
@@ -51,15 +51,15 @@ RUN --mount=type=secret,id=actions_results_url,env=ACTIONS_RESULTS_URL \
5151 --mount=type=secret,id=actions_runtime_token,env=ACTIONS_RUNTIME_TOKEN \
5252 if [ ${CUDA_COMPUTE_CAP} -ge 75 -a ${CUDA_COMPUTE_CAP} -lt 80 ]; \
5353 then \
54- nvprune --generate-code code=sm_${CUDA_COMPUTE_CAP} /usr/local/cuda/lib64/libcublas_static.a -o /usr/local/cuda/lib64/libcublas_static.a; \
54+ nvprune --generate-code code=sm_${CUDA_COMPUTE_CAP} /usr/local/cuda/lib64/libcublas_static.a -o /usr/local/cuda/lib64/libcublas_static.a; \
5555 elif [ ${CUDA_COMPUTE_CAP} -ge 80 -a ${CUDA_COMPUTE_CAP} -lt 90 ]; \
5656 then \
57- nvprune --generate-code code=sm_80 --generate-code code=sm_${CUDA_COMPUTE_CAP} /usr/local/cuda/lib64/libcublas_static.a -o /usr/local/cuda/lib64/libcublas_static.a; \
57+ nvprune --generate-code code=sm_80 --generate-code code=sm_${CUDA_COMPUTE_CAP} /usr/local/cuda/lib64/libcublas_static.a -o /usr/local/cuda/lib64/libcublas_static.a; \
5858 elif [ ${CUDA_COMPUTE_CAP} -eq 90 ]; \
5959 then \
60- nvprune --generate-code code=sm_90 /usr/local/cuda/lib64/libcublas_static.a -o /usr/local/cuda/lib64/libcublas_static.a; \
60+ nvprune --generate-code code=sm_90 /usr/local/cuda/lib64/libcublas_static.a -o /usr/local/cuda/lib64/libcublas_static.a; \
6161 else \
62- echo "cuda compute cap ${CUDA_COMPUTE_CAP} is not supported"; exit 1; \
62+ echo "cuda compute cap ${CUDA_COMPUTE_CAP} is not supported"; exit 1; \
6363 fi;
6464
6565COPY --from=planner /usr/src/recipe.json recipe.json
@@ -68,9 +68,9 @@ RUN --mount=type=secret,id=actions_results_url,env=ACTIONS_RESULTS_URL \
6868 --mount=type=secret,id=actions_runtime_token,env=ACTIONS_RUNTIME_TOKEN \
6969 if [ ${CUDA_COMPUTE_CAP} -ge 75 -a ${CUDA_COMPUTE_CAP} -lt 80 ]; \
7070 then \
71- cargo chef cook --release --features candle-cuda-turing --features static-linking --no-default-features --recipe-path recipe.json && sccache -s; \
71+ cargo chef cook --release --features candle-cuda-turing --features static-linking --no-default-features --recipe-path recipe.json && sccache -s; \
7272 else \
73- cargo chef cook --release --features candle-cuda --features static-linking --no-default-features --recipe-path recipe.json && sccache -s; \
73+ cargo chef cook --release --features candle-cuda --features static-linking --no-default-features --recipe-path recipe.json && sccache -s; \
7474 fi;
7575
7676COPY backends backends
@@ -85,9 +85,9 @@ RUN --mount=type=secret,id=actions_results_url,env=ACTIONS_RESULTS_URL \
8585 --mount=type=secret,id=actions_runtime_token,env=ACTIONS_RUNTIME_TOKEN \
8686 if [ ${CUDA_COMPUTE_CAP} -ge 75 -a ${CUDA_COMPUTE_CAP} -lt 80 ]; \
8787 then \
88- cargo build --release --bin text-embeddings-router -F candle-cuda-turing -F static-linking -F http --no-default-features && sccache -s; \
88+ cargo build --release --bin text-embeddings-router -F candle-cuda-turing -F static-linking -F http --no-default-features && sccache -s; \
8989 else \
90- cargo build --release --bin text-embeddings-router -F candle-cuda -F static-linking -F http --no-default-features && sccache -s; \
90+ cargo build --release --bin text-embeddings-router -F candle-cuda -F static-linking -F http --no-default-features && sccache -s; \
9191 fi;
9292
9393FROM builder AS grpc-builder
@@ -108,9 +108,9 @@ RUN --mount=type=secret,id=actions_results_url,env=ACTIONS_RESULTS_URL \
108108 --mount=type=secret,id=actions_runtime_token,env=ACTIONS_RUNTIME_TOKEN \
109109 if [ ${CUDA_COMPUTE_CAP} -ge 75 -a ${CUDA_COMPUTE_CAP} -lt 80 ]; \
110110 then \
111- cargo build --release --bin text-embeddings-router -F candle-cuda-turing -F static-linking -F grpc --no-default-features && sccache -s; \
111+ cargo build --release --bin text-embeddings-router -F candle-cuda-turing -F static-linking -F grpc --no-default-features && sccache -s; \
112112 else \
113- cargo build --release --bin text-embeddings-router -F candle-cuda -F static-linking -F grpc --no-default-features && sccache -s; \
113+ cargo build --release --bin text-embeddings-router -F candle-cuda -F static-linking -F grpc --no-default-features && sccache -s; \
114114 fi;
115115
116116FROM nvidia/cuda:12.2.0-base-ubuntu22.04 AS base
0 commit comments