Skip to content

Commit 39d9759

Browse files
grzegorz-roboflowgrzegorz-roboflow
authored
Reduce number of layers in lambda dockerfile (#1940)
* Reduce number of layers in docker/dockerfiles/Dockerfile.onnx.lambda * Remove build dependencies once build is done * E2E assertions 1e-3 -> 1e-2 (test_image_description_workflow, test_workflow_with_clip_as_classifier_replacing_predictions) * Remove rust & libjpeg-devel * Make Dockerfile.onnx.lambda.slim slim * Remove unsupported 'flush' param when logging
1 parent a4d114d commit 39d9759

File tree

5 files changed

+108
-92
lines changed

5 files changed

+108
-92
lines changed

‎docker/dockerfiles/Dockerfile.onnx.lambda‎

Lines changed: 52 additions & 44 deletions
Original file line numberDiff line numberDiff line change
@@ -5,20 +5,6 @@ WORKDIR /app
55
ARG DEBIAN_FRONTEND=noninteractive
66
ARG TARGETPLATFORM
77

8-
RUN dnf makecache -y && dnf install -y \
9-
libSM \
10-
libXext \
11-
python3-pip \
12-
git \
13-
zlib-devel \
14-
libjpeg-devel \
15-
gcc \
16-
mesa-libGL \
17-
pango \
18-
rustc \
19-
cargo \
20-
&& dnf clean all
21-
228
COPY requirements/requirements.clip.txt \
239
requirements/requirements.cpu.txt \
2410
requirements/requirements.http.txt \
@@ -32,11 +18,22 @@ COPY requirements/requirements.clip.txt \
3218
requirements/requirements.easyocr.txt \
3319
./
3420

35-
36-
RUN if [ "${TARGETPLATFORM}" == "linux/amd64" ]; then mv requirements.vino.txt requirements.cpu.txt; fi
37-
38-
RUN /var/lang/bin/python3.12 -m pip install --upgrade pip && rm -rf ~/.cache/pip
39-
RUN pip3 install \
21+
RUN dnf makecache -y && dnf install -y \
22+
libSM \
23+
libXext \
24+
python3-pip \
25+
git \
26+
zlib-devel \
27+
libjpeg-devel \
28+
gcc \
29+
mesa-libGL \
30+
pango \
31+
rustc \
32+
cargo \
33+
&& dnf clean all \
34+
&& if [ "${TARGETPLATFORM}" == "linux/amd64" ]; then mv requirements.vino.txt requirements.cpu.txt; fi \
35+
&& /var/lang/bin/python3.12 -m pip install --upgrade pip && rm -rf ~/.cache/pip \
36+
&& pip3 install \
4037
--extra-index-url https://download.pytorch.org/whl/cpu \
4138
certifi==2022.12.07 \
4239
-r _requirements.txt \
@@ -53,37 +50,48 @@ RUN pip3 install \
5350
"setuptools<=75.5.0" \
5451
--upgrade \
5552
--target "${LAMBDA_TASK_ROOT}" \
56-
&& rm -rf ~/.cache/pip
53+
&& rm -rf ~/.cache/pip \
54+
&& rpm -e --nodeps \
55+
gcc \
56+
cargo \
57+
rust \
58+
zlib-devel \
59+
git \
60+
libjpeg-turbo-devel \
61+
perl-Git \
62+
annobin-plugin-gcc \
63+
rust-std-static \
64+
&& dnf clean all \
65+
&& rm -rf /var/cache/dnf /build
5766

5867
COPY inference ${LAMBDA_TASK_ROOT}/inference
5968
COPY inference_sdk ${LAMBDA_TASK_ROOT}/inference_sdk
6069
COPY docker/config/lambda.py ${LAMBDA_TASK_ROOT}/lambda.py
6170

62-
ENV LAMBDA=True
63-
ENV CORE_MODEL_SAM_ENABLED=False
64-
ENV CORE_MODEL_SAM2_ENABLED=False
65-
ENV ALLOW_NUMPY_INPUT=False
66-
ENV INFERENCE_SERVER_ID=HostedInferenceLambda
67-
ENV DISABLE_VERSION_CHECK=true
68-
ENV DOCTR_MULTIPROCESSING_DISABLE=TRUE
69-
ENV REDIS_SSL=true
70-
ENV WORKFLOWS_STEP_EXECUTION_MODE=remote
71-
ENV WORKFLOWS_REMOTE_API_TARGET=hosted
72-
ENV API_LOGGING_ENABLED=True
73-
ENV MODEL_VALIDATION_DISABLED=True
74-
ENV ALLOW_NON_HTTPS_URL_INPUT=False
75-
ENV ALLOW_URL_INPUT_WITHOUT_FQDN=False
76-
ENV ALLOW_CUSTOM_PYTHON_EXECUTION_IN_WORKFLOWS=False
77-
ENV CORE_MODEL_TROCR_ENABLED=false
78-
ENV USE_FILE_CACHE_FOR_WORKFLOWS_DEFINITIONS=False
79-
ENV ALLOW_WORKFLOW_BLOCKS_ACCESSING_LOCAL_STORAGE=False
80-
ENV ALLOW_WORKFLOW_BLOCKS_ACCESSING_ENVIRONMENTAL_VARIABLES=False
81-
ENV ALLOW_LOADING_IMAGES_FROM_LOCAL_FILESYSTEM=False
82-
ENV DEPTH_ESTIMATION_ENABLED=False
83-
ENV CORE_MODEL_PE_ENABLED=false
84-
ENV CORE_MODEL_SAM3_ENABLED=False
71+
ENV LAMBDA=True \
72+
CORE_MODEL_SAM_ENABLED=False \
73+
CORE_MODEL_SAM2_ENABLED=False \
74+
ALLOW_NUMPY_INPUT=False \
75+
INFERENCE_SERVER_ID=HostedInferenceLambda \
76+
DISABLE_VERSION_CHECK=true \
77+
DOCTR_MULTIPROCESSING_DISABLE=TRUE \
78+
REDIS_SSL=true \
79+
WORKFLOWS_STEP_EXECUTION_MODE=remote \
80+
WORKFLOWS_REMOTE_API_TARGET=hosted \
81+
API_LOGGING_ENABLED=True \
82+
MODEL_VALIDATION_DISABLED=True \
83+
ALLOW_NON_HTTPS_URL_INPUT=False \
84+
ALLOW_URL_INPUT_WITHOUT_FQDN=False \
85+
ALLOW_CUSTOM_PYTHON_EXECUTION_IN_WORKFLOWS=False \
86+
CORE_MODEL_TROCR_ENABLED=false \
87+
USE_FILE_CACHE_FOR_WORKFLOWS_DEFINITIONS=False \
88+
ALLOW_WORKFLOW_BLOCKS_ACCESSING_LOCAL_STORAGE=False \
89+
ALLOW_WORKFLOW_BLOCKS_ACCESSING_ENVIRONMENTAL_VARIABLES=False \
90+
ALLOW_LOADING_IMAGES_FROM_LOCAL_FILESYSTEM=False \
91+
DEPTH_ESTIMATION_ENABLED=False \
92+
CORE_MODEL_PE_ENABLED=false \
93+
CORE_MODEL_SAM3_ENABLED=False
8594

8695
WORKDIR ${LAMBDA_TASK_ROOT}
87-
RUN rm -rf /build
8896

8997
CMD [ "lambda.handler" ]

‎docker/dockerfiles/Dockerfile.onnx.lambda.slim‎

Lines changed: 52 additions & 44 deletions
Original file line numberDiff line numberDiff line change
@@ -4,6 +4,14 @@ WORKDIR /app
44
ARG DEBIAN_FRONTEND=noninteractive
55
ARG TARGETPLATFORM
66

7+
COPY requirements/requirements.cpu.txt \
8+
requirements/requirements.http.txt \
9+
requirements/requirements.hosted.txt \
10+
requirements/_requirements.txt \
11+
requirements/requirements.sdk.http.txt \
12+
requirements/requirements.vino.txt \
13+
./
14+
715
RUN dnf makecache -y && dnf install -y \
816
libSM \
917
libXext \
@@ -16,20 +24,10 @@ RUN dnf makecache -y && dnf install -y \
1624
pango \
1725
rustc \
1826
cargo \
19-
&& dnf clean all
20-
21-
COPY requirements/requirements.cpu.txt \
22-
requirements/requirements.http.txt \
23-
requirements/requirements.hosted.txt \
24-
requirements/_requirements.txt \
25-
requirements/requirements.sdk.http.txt \
26-
requirements/requirements.vino.txt \
27-
./
28-
29-
RUN if [ "${TARGETPLATFORM}" == "linux/amd64" ]; then mv requirements.vino.txt requirements.cpu.txt; fi
30-
31-
RUN /var/lang/bin/python3.12 -m pip install --upgrade pip && rm -rf ~/.cache/pip
32-
RUN pip3 install \
27+
&& dnf clean all \
28+
&& if [ "${TARGETPLATFORM}" == "linux/amd64" ]; then mv requirements.vino.txt requirements.cpu.txt; fi \
29+
&& /var/lang/bin/python3.12 -m pip install --upgrade pip && rm -rf ~/.cache/pip \
30+
&& pip3 install \
3331
certifi==2022.12.07 \
3432
-r _requirements.txt \
3533
-r requirements.cpu.txt \
@@ -40,41 +38,51 @@ RUN pip3 install \
4038
"setuptools<=75.5.0" \
4139
--upgrade \
4240
--target "${LAMBDA_TASK_ROOT}" \
43-
&& rm -rf ~/.cache/pip
44-
45-
41+
&& rm -rf ~/.cache/pip \
42+
&& rpm -e --nodeps \
43+
gcc \
44+
cargo \
45+
rust \
46+
zlib-devel \
47+
git \
48+
libjpeg-turbo-devel \
49+
perl-Git \
50+
annobin-plugin-gcc \
51+
rust-std-static \
52+
&& dnf clean all \
53+
&& rm -rf /var/cache/dnf /build
4654

4755
COPY inference ${LAMBDA_TASK_ROOT}/inference
4856
COPY inference_sdk ${LAMBDA_TASK_ROOT}/inference_sdk
4957
COPY docker/config/lambda.py ${LAMBDA_TASK_ROOT}/lambda.py
5058

51-
ENV CORE_MODEL_CLIP_ENABLED=false
52-
ENV CORE_MODEL_PE_ENABLED=false
53-
ENV CORE_MODEL_SAM_ENABLED=false
54-
ENV CORE_MODEL_SAM2_ENABLED=false
55-
ENV CORE_MODEL_GAZE_ENABLED=false
56-
ENV CORE_MODEL_DOCTR_ENABLED=false
57-
ENV CORE_MODEL_YOLO_WORLD_ENABLED=false
58-
ENV CORE_MODELS_ENABLED=false
59-
ENV LAMBDA=True
60-
ENV ALLOW_NUMPY_INPUT=False
61-
ENV INFERENCE_SERVER_ID=HostedInferenceLambda
62-
ENV DISABLE_VERSION_CHECK=true
63-
ENV REDIS_SSL=true
64-
ENV WORKFLOWS_STEP_EXECUTION_MODE=remote
65-
ENV WORKFLOWS_REMOTE_API_TARGET=hosted
66-
ENV API_LOGGING_ENABLED=True
67-
ENV MODEL_VALIDATION_DISABLED=True
68-
ENV ALLOW_NON_HTTPS_URL_INPUT=False
69-
ENV ALLOW_URL_INPUT_WITHOUT_FQDN=False
70-
ENV ALLOW_CUSTOM_PYTHON_EXECUTION_IN_WORKFLOWS=False
71-
ENV CORE_MODEL_TROCR_ENABLED=false
72-
ENV USE_FILE_CACHE_FOR_WORKFLOWS_DEFINITIONS=False
73-
ENV ALLOW_WORKFLOW_BLOCKS_ACCESSING_LOCAL_STORAGE=False
74-
ENV ALLOW_WORKFLOW_BLOCKS_ACCESSING_ENVIRONMENTAL_VARIABLES=False
75-
ENV ALLOW_LOADING_IMAGES_FROM_LOCAL_FILESYSTEM=False
76-
ENV DEPTH_ESTIMATION_ENABLED=False
77-
ENV CORE_MODEL_SAM3_ENABLED=False
59+
ENV CORE_MODEL_CLIP_ENABLED=false \
60+
CORE_MODEL_PE_ENABLED=false \
61+
CORE_MODEL_SAM_ENABLED=false \
62+
CORE_MODEL_SAM2_ENABLED=false \
63+
CORE_MODEL_GAZE_ENABLED=false \
64+
CORE_MODEL_DOCTR_ENABLED=false \
65+
CORE_MODEL_YOLO_WORLD_ENABLED=false \
66+
CORE_MODELS_ENABLED=false \
67+
LAMBDA=True \
68+
ALLOW_NUMPY_INPUT=False \
69+
INFERENCE_SERVER_ID=HostedInferenceLambda \
70+
DISABLE_VERSION_CHECK=true \
71+
REDIS_SSL=true \
72+
WORKFLOWS_STEP_EXECUTION_MODE=remote \
73+
WORKFLOWS_REMOTE_API_TARGET=hosted \
74+
API_LOGGING_ENABLED=True \
75+
MODEL_VALIDATION_DISABLED=True \
76+
ALLOW_NON_HTTPS_URL_INPUT=False \
77+
ALLOW_URL_INPUT_WITHOUT_FQDN=False \
78+
ALLOW_CUSTOM_PYTHON_EXECUTION_IN_WORKFLOWS=False \
79+
CORE_MODEL_TROCR_ENABLED=false \
80+
USE_FILE_CACHE_FOR_WORKFLOWS_DEFINITIONS=False \
81+
ALLOW_WORKFLOW_BLOCKS_ACCESSING_LOCAL_STORAGE=False \
82+
ALLOW_WORKFLOW_BLOCKS_ACCESSING_ENVIRONMENTAL_VARIABLES=False \
83+
ALLOW_LOADING_IMAGES_FROM_LOCAL_FILESYSTEM=False \
84+
DEPTH_ESTIMATION_ENABLED=False \
85+
CORE_MODEL_SAM3_ENABLED=False
7886

7987
WORKDIR ${LAMBDA_TASK_ROOT}
8088

‎inference/core/interfaces/http/http_api.py‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1644,7 +1644,7 @@ def initialize_models(state: ModelInitState):
16441644
# Limit the number of concurrent tasks to prevent resource exhaustion
16451645

16461646
def load_model(model_id):
1647-
logger.debug(f"load_model({model_id}) - starting", flush=True)
1647+
logger.debug(f"load_model({model_id}) - starting")
16481648
try:
16491649
# TODO: how to add timeout here? Probably best to timeout model loading?
16501650
model_add(
@@ -1660,7 +1660,7 @@ def load_model(model_id):
16601660
logger.error(error_msg)
16611661
with state.lock:
16621662
state.initialization_errors.append((model_id, str(e)))
1663-
logger.debug(f"load_model({model_id}) - finished", flush=True)
1663+
logger.debug(f"load_model({model_id}) - finished")
16641664

16651665
if PRELOAD_MODELS:
16661666
# Create tasks for each model to be loaded

‎tests/inference/hosted_platform_tests/workflows_examples/test_workflow_with_clip.py‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -92,7 +92,7 @@ def test_workflow_with_clip_as_classifier_replacing_predictions(
9292
p["confidence"] for p in result[0]["original_predictions"]["predictions"]
9393
]
9494
assert np.allclose(
95-
detection_confidences, [0.856178879737854, 0.5191817283630371], atol=1e-3
95+
detection_confidences, [0.856178879737854, 0.5191817283630371], atol=1e-2
9696
), "Expected predictions to match what was observed while test creation"
9797
assert (
9898
len(result[0]["modified_predictions"]["predictions"]) == 2

‎tests/inference/hosted_platform_tests/workflows_examples/test_workflow_with_openai.py‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -92,7 +92,7 @@ def test_image_description_workflow(
9292
p["confidence"] for p in result[0]["detection_predictions"]["predictions"]
9393
]
9494
assert np.allclose(
95-
detection_confidences, [0.856178879737854, 0.5191817283630371], atol=1e-3
95+
detection_confidences, [0.856178879737854, 0.5191817283630371], atol=1e-2
9696
), "Expected predictions to match what was observed while test creation"
9797
assert len(result[0]["description"]) > 0, "Expected some description"
9898

0 commit comments

Comments
 (0)