Update configs

.github/workflows/docker-push-vidvec.yml

@@ -30,7 +30,16 @@ jobs:
           workdir: src/api/
           dockerfile: Dockerfile.vid_vec_rep_resnet
           tags: ${{ steps.vars.outputs.sha_short }}
-          platforms: linux/amd64,linux/arm64
+          platforms: linux/amd64
+      - uses: elgohr/Publish-Docker-Github-Action@main
+        with:
+          username: tattletech
+          password: ${{ secrets.DOCKER_PASSWORD }}
+          name: tattletech/feluda-operator-vidvec
+          workdir: src/api/
+          dockerfile: Dockerfile.vid_vec_rep_resnet.graviton
+          tags: ${{ steps.vars.outputs.sha_short }}
+          platforms: linux/arm64
 
       # - name: deploy to cluster
       #   uses: steebchen/kubectl@v2.0.0

docker-compose.yml

@@ -26,20 +26,20 @@ services:
     cap_add:
       - IPC_LOCK
 
-  queue:
-    image: rabbitmq:3.8.7-management
-    container_name: rabbitmq
-    hostname: rabbit
-    volumes:
-      - ./.docker/rabbitmq/data:/var/lib/rabbitmq
-      - ./.docker/rabbitmq/logs:/var/log/rabbitmq
-    environment:
-      RABBITMQ_ERLANG_COOKIE: "secret-cookie"
-      RABBITMQ_DEFAULT_USER: "admin"
-      RABBITMQ_DEFAULT_PASS: "Admin123"
-    ports:
-      - 5672:5672
-      - 15672:15672
+#  queue:
+#    image: rabbitmq:3.8.7-management
+#    container_name: rabbitmq
+#    hostname: rabbit
+#    volumes:
+#      - ./.docker/rabbitmq/data:/var/lib/rabbitmq
+#      - ./.docker/rabbitmq/logs:/var/log/rabbitmq
+#    environment:
+#      RABBITMQ_ERLANG_COOKIE: "secret-cookie"
+#      RABBITMQ_DEFAULT_USER: "admin"
+#      RABBITMQ_DEFAULT_PASS: "Admin123"
+#    ports:
+#      - 5672:5672
+#      - 15672:15672
 
   api:
     container_name: feluda_api
@@ -57,8 +57,8 @@ services:
     depends_on:
       store:
         condition: service_started
-      queue:
-        condition: service_started
+#      queue:
+#        condition: service_started
 
   indexer:
     container_name: feluda_indexer
@@ -71,22 +71,22 @@ services:
     depends_on:
       store:
         condition: service_started
-      queue:
-        condition: service_started
+#      queue:
+#        condition: service_started
 
-  reporter:
-    container_name: feluda_reporter
-    build:
-      context: ./src/api
-      dockerfile: Dockerfile
-      target: debug
-    env_file: ./src/api/development.env
-    command: tail -f /dev/null
-    depends_on:
-      store:
-        condition: service_started
-      queue:
-        condition: service_started
+#  reporter:
+#    container_name: feluda_reporter
+#    build:
+#      context: ./src/api
+#      dockerfile: Dockerfile
+#      target: debug
+#    env_file: ./src/api/development.env
+#    command: tail -f /dev/null
+#    depends_on:
+#      store:
+#        condition: service_started
+#      queue:
+#        condition: service_started
 
 #networks:
 #  default:

src/api/Dockerfile

@@ -1,15 +1,15 @@
 # FROM jrottenberg/ffmpeg:4.0-scratch AS ffmpeg
 
-FROM python:3.11-slim@sha256:637774748f62b832dc11e7b286e48cd716727ed04b45a0322776c01bc526afc3 AS base
+FROM python:3.11-slim-bullseye AS base
 # COPY --from=ffmpeg / /
 RUN apt-get update \
     && apt-get -y upgrade \
-    && apt-get install -y \
-    --no-install-recommends gcc build-essential \
-    --no-install-recommends libgl1-mesa-glx libglib2.0-0 \
-    && apt-get purge -y --auto-remove \
-    gcc build-essential \
-    libgl1-mesa-glx libglib2.0-0 \
+#    && apt-get install -y \
+#    --no-install-recommends gcc build-essential \
+#    --no-install-recommends libgl1-mesa-glx libglib2.0-0 \
+#    && apt-get purge -y --auto-remove \
+#    gcc build-essential \
+#    libgl1-mesa-glx libglib2.0-0 \
     && rm -rf /var/lib/apt/lists/* 
 ENV PATH=/root/.local/bin:$PATH
 RUN pip install --no-cache-dir --upgrade pip

src/api/Dockerfile.vid_vec_rep_resnet

@@ -15,28 +15,6 @@ WORKDIR /app
 COPY ./core/operators/vid_vec_rep_resnet_requirements.txt /app/core/operators/vid_vec_rep_resnet_requirements.txt
 RUN pip install --no-cache-dir --user -r /app/core/operators/vid_vec_rep_resnet_requirements.txt
 
-### AWS Graviton Optimization ###
-# Graviton3(E) (e.g. c7g, c7gn and Hpc7g instances) supports BF16 format for ML acceleration. This can be enabled in oneDNN by setting the below environment variable
-ENV DNNL_DEFAULT_FPMATH_MODE=BF16
-
-# Enable primitive caching to avoid the redundant primitive allocation
-# latency overhead. Please note this caching feature increases the
-# memory footprint. Tune this cache capacity to a lower value to
-# reduce the additional memory requirement.
-ENV LRU_CACHE_CAPACITY=1024
-
-# Enable Transparent huge page allocations from PyTorch C10 allocator
-ENV THP_MEM_ALLOC_ENABLE=1
-
-# Make sure the openmp threads are distributed across all the processes for multi process applications to avoid over subscription for the vcpus. For example if there is a single application process, then num_processes should be set to '1' so that all the vcpus are assigned to it with one-to-one mapping to omp threads
-# RUN num_vcpus=8
-# RUN num_processes=1
-# RUN temp = $((1 > ($num_vcpus/$num_processes) ? 1 : ($num_vcpus/$num_processes)))
-ENV OMP_NUM_THREADS=1
-ENV OMP_PROC_BIND=false
-ENV OMP_PLACES=cores
-###
-
 COPY ./core/operators/vid_vec_rep_resnet.py /app/core/operators/vid_vec_rep_resnet.py
 
 COPY ./core/operators/sample_data/sample-cat-video.mp4 /app/core/operators/sample_data/sample-cat-video.mp4

src/api/Dockerfile.vid_vec_rep_resnet.graviton

@@ -0,0 +1,67 @@
+FROM python:3.11-slim-bullseye AS base
+RUN apt-get update \
+    && apt-get -y upgrade \
+    && apt-get install -y \
+    --no-install-recommends gcc build-essential \
+    --no-install-recommends libgl1-mesa-glx libglib2.0-0 \
+    && apt-get purge -y --auto-remove \
+    gcc build-essential \
+    libgl1-mesa-glx libglib2.0-0 \
+    && rm -rf /var/lib/apt/lists/* 
+ENV PATH=/root/.local/bin:$PATH
+
+RUN pip install --no-cache-dir --upgrade pip
+WORKDIR /app
+COPY ./core/operators/vid_vec_rep_resnet_requirements.txt /app/core/operators/vid_vec_rep_resnet_requirements.txt
+RUN pip install --no-cache-dir --user -r /app/core/operators/vid_vec_rep_resnet_requirements.txt
+
+### AWS Graviton Optimization ###
+
+# Graviton3(E) (e.g. c7g, c7gn and Hpc7g instances) supports BF16 format for ML acceleration. This can be enabled in oneDNN by setting the below environment variable
+ENV DNNL_DEFAULT_FPMATH_MODE=BF16
+
+# Enable primitive caching to avoid the redundant primitive allocation
+# latency overhead. Please note this caching feature increases the
+# memory footprint. Tune this cache capacity to a lower value to
+# reduce the additional memory requirement.
+ENV LRU_CACHE_CAPACITY=1024
+
+# Enable Transparent huge page allocations from PyTorch C10 allocator
+ENV THP_MEM_ALLOC_ENABLE=1
+
+# Make sure the openmp threads are distributed across all the processes for multi process applications to avoid over subscription for the vcpus. For example if there is a single application process, then num_processes should be set to '1' so that all the vcpus are assigned to it with one-to-one mapping to omp threads
+# RUN num_vcpus=8
+# RUN num_processes=1
+# RUN temp = $((1 > ($num_vcpus/$num_processes) ? 1 : ($num_vcpus/$num_processes)))
+ENV OMP_NUM_THREADS=1
+ENV OMP_PROC_BIND=false
+ENV OMP_PLACES=cores
+###
+
+COPY ./core/operators/vid_vec_rep_resnet.py /app/core/operators/vid_vec_rep_resnet.py
+
+COPY ./core/operators/sample_data/sample-cat-video.mp4 /app/core/operators/sample_data/sample-cat-video.mp4
+COPY ./core/operators/test_vid_vec_rep_resnet.py /app/core/operators/test_vid_vec_rep_resnet.py
+
+COPY ./video_vec_operator_cprofile.py /app/video_vec_operator_cprofile.py
+COPY ./video_vec_operator_cprofile.sh /app/video_vec_operator_cprofile.sh
+RUN chmod +x video_vec_operator_cprofile.sh
+
+COPY ./video_vec_operator_time.py /app/video_vec_operator_time.py
+COPY ./video_vec_operator_time.sh /app/video_vec_operator_time.sh
+RUN chmod +x video_vec_operator_time.sh
+
+COPY ./video_vec_operator_profile.py /app/video_vec_operator_profile.py
+COPY ./video_vec_operator_profile_memray.sh /app/video_vec_operator_profile_memray.sh
+COPY ./video_vec_operator_profile_pyinstrument.sh /app/video_vec_operator_profile_pyinstrument.sh
+RUN chmod +x video_vec_operator_profile_memray.sh
+RUN chmod +x video_vec_operator_profile_pyinstrument.sh
+
+COPY ./video_vec_operator_multicore.py /app/video_vec_operator_multicore.py
+COPY ./video_vec_operator_multicore.sh /app/video_vec_operator_multicore.sh
+RUN chmod +x video_vec_operator_multicore.sh
+
+COPY ./benchmark.sh /app/benchmark.sh
+RUN chmod +x benchmark.sh
+
+CMD tail -f /dev/null

src/api/core/feluda.py

@@ -5,7 +5,7 @@
 from core import config, store
 from core.server import Server
 from core.operators import Operator
-from core.queue import Queue
+# from core.queue import Queue
 from enum import Enum
 
 
@@ -23,9 +23,9 @@ def __init__(self, configPath):
             self.operators = Operator(self.config.operators)
         if self.config.store:
             self.store = store.get_store(self.config.store)
-        if self.config.queue:
-            # print("---> 1", self.config.queue)
-            self.queue = Queue.make(self.config.queue)
+        # if self.config.queue:
+        #     # print("---> 1", self.config.queue)
+        #     self.queue = Queue.make(self.config.queue)
         if self.config.server:
             self.server = Server(self.config.server)
 
@@ -43,8 +43,8 @@ def start(self):
             self.store.connect()
             self.store.optionally_create_index()
 
-        if self.queue:
-            self.queue.connect()
+        # if self.queue:
+        #     self.queue.connect()
 
         if self.server:
             self.server.start()
@@ -55,9 +55,9 @@ def start_component(self, component_type: ComponentType):
         elif component_type == ComponentType.STORE and self.store:
             self.store.connect()
             self.store.optionally_create_index()
-        elif component_type == ComponentType.QUEUE and self.queue:
-            self.queue.connect()
-            self.queue.initialize()
+        # elif component_type == ComponentType.QUEUE and self.queue:
+        #     self.queue.connect()
+        #     self.queue.initialize()
         else:
             raise Exception("Unsupported Component Type")
 

src/api/core/operators/vid_vec_rep_resnet_requirements.in

@@ -1,5 +1,5 @@
-torch==2.1.2+cpu
-torchvision==0.16.2+cpu
+torch==2.2.0+cpu
+torchvision==0.17.0+cpu
 numpy==1.26.3
 Pillow==10.2.0
 scipy==1.11.4

src/api/requirements.in

@@ -1,17 +1,15 @@
 flask==2.3.2
 flask_cors==3.0.9
-google-cloud==0.34.0
-google-cloud-vision==3.5.0
 Pillow==10.2.0
 elasticsearch==8.11.1
 wget==3.2
 pika==1.3.2
 python-dotenv==1.0.0
-boto3==1.34.34
 click==8.1.7
 pytest==7.4.4
 PyYAML==6.0.1
 dacite==1.8.1
 memray==1.11.0  # dev
 pyinstrument==4.6.2
-numpy==1.26.3
+numpy==1.26.3
+requests==2.31.0