v1.2.2

Author: OlivierDehaene

Cargo.lock

@@ -11,7 +11,7 @@ members = [
 resolver = "2"
 
 [workspace.package]
-version = "1.2.1"
+version = "1.2.2"
 edition = "2021"
 authors = ["Olivier Dehaene"]
 homepage = "https://github.com/huggingface/text-embeddings-inference"

Cargo.toml

@@ -102,17 +102,18 @@ COPY --from=grpc-builder /usr/src/target/release/text-embeddings-router /usr/loc
 ENTRYPOINT ["text-embeddings-router"]
 CMD ["--json-output"]
 
-FROM base
+FROM base AS http
 
 COPY --from=http-builder /usr/src/target/release/text-embeddings-router /usr/local/bin/text-embeddings-router
 
-ENTRYPOINT ["text-embeddings-router"]
-CMD ["--json-output"]
-
 # Amazon SageMaker compatible image
-FROM base AS sagemaker
-
-COPY --from=http-builder /usr/src/target/release/text-embeddings-router /usr/local/bin/text-embeddings-router
+FROM http as sagemaker
 COPY --chmod=775 sagemaker-entrypoint.sh entrypoint.sh
 
 ENTRYPOINT ["./entrypoint.sh"]
+
+# Default image
+FROM http
+
+ENTRYPOINT ["text-embeddings-router"]
+CMD ["--json-output"]

Dockerfile

@@ -40,6 +40,9 @@ ARG ACTIONS_CACHE_URL
 ARG ACTIONS_RUNTIME_TOKEN
 ARG SCCACHE_GHA_ENABLED
 
+# limit the number of kernels built at the same time
+ARG RAYON_NUM_THREADS=2
+
 WORKDIR /usr/src
 
 COPY --from=planner /usr/src/recipe.json recipe.json
@@ -122,15 +125,17 @@ COPY --from=builder-75 /usr/src/target/release/text-embeddings-router /usr/local
 COPY --from=builder-80 /usr/src/target/release/text-embeddings-router /usr/local/bin/text-embeddings-router-80
 COPY --from=builder-90 /usr/src/target/release/text-embeddings-router /usr/local/bin/text-embeddings-router-90
 
-COPY cuda-all-entrypoint.sh entrypoint.sh
-RUN chmod +x entrypoint.sh
-
-ENTRYPOINT ["./entrypoint.sh"]
-CMD ["--json-output"]
-
 # Amazon SageMaker compatible image
 FROM base AS sagemaker
 
 COPY --chmod=775 sagemaker-entrypoint-cuda-all.sh entrypoint.sh
 
-ENTRYPOINT ["./entrypoint.sh"]
+ENTRYPOINT ["./entrypoint.sh"]
+
+# Default image
+FROM base
+
+COPY --chmod=775 cuda-all-entrypoint.sh entrypoint.sh
+
+ENTRYPOINT ["./entrypoint.sh"]
+CMD ["--json-output"]

Dockerfile-cuda-all

@@ -10,7 +10,7 @@
       "name": "Apache 2.0",
       "url": "https://www.apache.org/licenses/LICENSE-2.0"
     },
-    "version": "1.2.1"
+    "version": "1.2."
   },
   "paths": {
     "/decode": {