diff --git a/.github/workflows/build-container.yaml b/.github/workflows/build-container.yaml
index 8b1cbf69..031207c0 100644
--- a/.github/workflows/build-container.yaml
+++ b/.github/workflows/build-container.yaml
@@ -18,40 +18,36 @@ jobs:
   starlette-pytorch-cpu:
     uses: ./.github/workflows/docker-build-action.yaml
     with:
-      context: ./
-      repository: huggingface
-      image: hf-endpoints-inference-pytorch-cpu
-      dockerfile: dockerfiles/starlette/pytorch/Dockerfile.cpu
+      image: inference-pytorch-cpu
+      dockerfile: dockerfiles/pytorch/cpu/Dockerfile
     secrets:
-      DOCKER_USERNAME: ${{ secrets.DOCKER_USERNAME }}
-      DOCKER_PASSWORD: ${{ secrets.DOCKER_PASSWORD }}
+      TAILSCALE_AUTHKEY: ${{ secrets.TAILSCALE_AUTHKEY }}
+      REGISTRY_USERNAME: ${{ secrets.REGISTRY_USERNAME }}
+      REGISTRY_PASSWORD: ${{ secrets.REGISTRY_PASSWORD }}
   starlette-pytorch-gpu:
     uses: ./.github/workflows/docker-build-action.yaml
     with:
-      context: ./
-      repository: huggingface
-      image: hf-endpoints-inference-pytorch-gpu
-      dockerfile: dockerfiles/starlette/pytorch/Dockerfile.gpu
+      image: inference-pytorch-gpu
+      dockerfile: dockerfiles/pytorch/gpu/Dockerfile
     secrets:
-      DOCKER_USERNAME: ${{ secrets.DOCKER_USERNAME }}
-      DOCKER_PASSWORD: ${{ secrets.DOCKER_PASSWORD }}
+      TAILSCALE_AUTHKEY: ${{ secrets.TAILSCALE_AUTHKEY }}
+      REGISTRY_USERNAME: ${{ secrets.REGISTRY_USERNAME }}
+      REGISTRY_PASSWORD: ${{ secrets.REGISTRY_PASSWORD }}
   starlette-tensorflow-cpu:
     uses: ./.github/workflows/docker-build-action.yaml
     with:
-      context: ./
-      repository: huggingface
-      image: hf-endpoints-inference-tensorflow-cpu
-      dockerfile: dockerfiles/starlette/tensorflow/Dockerfile.cpu
+      image: inference-tensorflow-cpu
+      dockerfile: dockerfiles/tensorflow/cpu/Dockerfile
     secrets:
-      DOCKER_USERNAME: ${{ secrets.DOCKER_USERNAME }}
-      DOCKER_PASSWORD: ${{ secrets.DOCKER_PASSWORD }}
+      TAILSCALE_AUTHKEY: ${{ secrets.TAILSCALE_AUTHKEY }}
+      REGISTRY_USERNAME: ${{ secrets.REGISTRY_USERNAME }}
+      REGISTRY_PASSWORD: ${{ secrets.REGISTRY_PASSWORD }}
   starlette-tensorflow-gpu:
     uses: ./.github/workflows/docker-build-action.yaml
     with:
-      context: ./
-      repository: huggingface
-      image: hf-endpoints-inference-tensorflow-gpu
-      dockerfile: dockerfiles/starlette/tensorflow/Dockerfile.gpu
+      image: inference-tensorflow-gpu
+      dockerfile: dockerfiles/tensorflow/gpu/Dockerfile
     secrets:
-      DOCKER_USERNAME: ${{ secrets.DOCKER_USERNAME }}
-      DOCKER_PASSWORD: ${{ secrets.DOCKER_PASSWORD }}
+      TAILSCALE_AUTHKEY: ${{ secrets.TAILSCALE_AUTHKEY }}
+      REGISTRY_USERNAME: ${{ secrets.REGISTRY_USERNAME }}
+      REGISTRY_PASSWORD: ${{ secrets.REGISTRY_PASSWORD }}
diff --git a/.github/workflows/docker-build-action.yaml b/.github/workflows/docker-build-action.yaml
index 9c3af40a..eae64905 100644
--- a/.github/workflows/docker-build-action.yaml
+++ b/.github/workflows/docker-build-action.yaml
@@ -3,17 +3,15 @@ on:
     inputs:
       context:
         type: string
-        required: true
+        required: false
+        default: "./"
       repository:
         type: string
-        required: true
+        required: false
+        default: "registry.internal.huggingface.tech/hf-endpoints"
       image:
         type: string
         required: true
-      platforms:
-        type: string
-        required: false
-        default: linux/amd64
       build_args:
         type: string
         required: false
@@ -23,9 +21,11 @@ on:
         required: false
         default: "Dockerfile"
     secrets:
-      DOCKER_USERNAME:
+      REGISTRY_USERNAME:
+        required: true
+      REGISTRY_PASSWORD:
         required: true
-      DOCKER_PASSWORD:
+      TAILSCALE_AUTHKEY:
         required: true
 
 jobs:
@@ -35,17 +35,23 @@ jobs:
       - name: Check out
         uses: actions/checkout@v3
 
-      - name: Set up QEMU
-        uses: docker/setup-qemu-action@v2
+      - name: Tailscale
+        uses: tailscale/github-action@v1
+        with:
+          authkey: ${{ secrets.TAILSCALE_AUTHKEY }}
+          version: 1.38.2
 
       - name: Set up Docker Buildx
         uses: docker/setup-buildx-action@v2.0.0
+        with:
+          install: true
 
       - name: Login to container registry
         uses: docker/login-action@v2.0.0
         with:
-          username: ${{ secrets.DOCKER_USERNAME }}
-          password: ${{ secrets.DOCKER_PASSWORD }}
+          registry: ${{ inputs.repository }}
+          username: ${{ secrets.REGISTRY_USERNAME }}
+          password: ${{ secrets.REGISTRY_PASSWORD }}
 
       - name: Inject slug/short variables
         uses: rlespinasse/github-slug-action@v4
@@ -57,4 +63,4 @@ jobs:
           context: ${{ inputs.context }}
           build-args: ${{ inputs.build_args }}
           file:  ${{ inputs.context }}/${{ inputs.dockerfile }}
-          tags: ${{ inputs.repository }}/${{ inputs.image }}:${{ env.GITHUB_SHA_SHORT }},${{ inputs.repository }}/${{ inputs.image }}:latest,
+          tags: ${{ inputs.repository }}/${{ inputs.image }}:${{ env.GITHUB_SHA_SHORT }},${{ inputs.repository }}/${{ inputs.image }}:latest
\ No newline at end of file
diff --git a/.github/workflows/gpu-integ-test.yaml b/.github/workflows/gpu-integ-test.yaml
index 42c9e169..ede153ea 100644
--- a/.github/workflows/gpu-integ-test.yaml
+++ b/.github/workflows/gpu-integ-test.yaml
@@ -64,7 +64,7 @@ jobs:
       - name: Install Python dependencies
         run: pip install -e .[test,dev,torch]
       - name: Build Docker
-        run: docker build -t starlette-transformers:gpu -f dockerfiles/starlette/pytorch/Dockerfile.gpu .
+        run: docker build -t starlette-transformers:gpu -f dockerfiles/pytorch/gpu/Dockerfile .
       - name: Run Integration Tests
         run: RUN_SLOW=True make integ-test
   tensorflow-integration-test:
@@ -86,7 +86,7 @@ jobs:
       - name: Install Python dependencies
         run: pip install -e .[test,dev,tensorflow]
       - name: Build Docker
-        run: docker build -t starlette-transformers:gpu -f dockerfiles/starlette/tensorflow/Dockerfile.gpu .
+        run: docker build -t starlette-transformers:gpu -f dockerfiles/tensorflow/gpu/Dockerfile .
       - name: Run Integration Tests
         run: RUN_SLOW=True make integ-test
 
diff --git a/.github/workflows/integ-test.yaml b/.github/workflows/integ-test.yaml
index e77d4a32..f6f6bba0 100644
--- a/.github/workflows/integ-test.yaml
+++ b/.github/workflows/integ-test.yaml
@@ -24,7 +24,7 @@ jobs:
     - name: Install Python dependencies
       run: pip install -e .[test,dev,torch]
     - name: Build Docker
-      run: docker build -t starlette-transformers:cpu -f dockerfiles/starlette/pytorch/Dockerfile.cpu .
+      run: docker build -t starlette-transformers:cpu -f dockerfiles/pytorch/cpu/Dockerfile .
     - name: Run Integration Tests
       env:
         AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
@@ -42,7 +42,7 @@ jobs:
     - name: Install Python dependencies
       run: pip install -e .[test,dev,tensorflow]
     - name: Build Docker
-      run: docker build -t starlette-transformers:cpu -f dockerfiles/starlette/tensorflow/Dockerfile.cpu .
+      run: docker build -t starlette-transformers:cpu -f dockerfiles/tensorflow/cpu/Dockerfile .
     - name: Run Integration Tests
       env:
         AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
diff --git a/README.md b/README.md
index 63c2b568..52b664a8 100644
--- a/README.md
+++ b/README.md
@@ -24,14 +24,14 @@ HF_MODEL_ID=hf-internal-testing/tiny-random-distilbert HF_MODEL_DIR=tmp2 HF_TASK
 
 _cpu images_
 ```bash
-docker build -t starlette-transformers:cpu -f dockerfiles/starlette/pytorch/Dockerfile.cpu .
-docker build -t starlette-transformers:cpu -f dockerfiles/starlette/tensorflow/Dockerfile.cpu .
+docker build -t starlette-transformers:cpu -f dockerfiles/pytorch/cpu/Dockerfile .
+docker build -t starlette-transformers:cpu -f dockerfiles/tensorflow/cpu/Dockerfile .
 ```
 
 _gpu images_
 ```bash
-docker build -t starlette-transformers:gpu -f dockerfiles/starlette/pytorch/Dockerfile.gpu .
-docker build -t starlette-transformers:gpu -f dockerfiles/starlette/tensorflow/Dockerfile.gpu .
+docker build -t starlette-transformers:gpu -f dockerfiles/pytorch/gpu/Dockerfile .
+docker build -t starlette-transformers:gpu -f dockerfiles/tensorflow/gpu/Dockerfile .
 ```
 
 2. Run the container and provide either environment variables to the HUB model you want to use or mount a volume to the container, where your model is stored.
@@ -39,6 +39,7 @@ docker build -t starlette-transformers:gpu -f dockerfiles/starlette/tensorflow/D
 
 ```bash
 docker run -ti -p 5000:5000 -e HF_MODEL_ID=distilbert-base-uncased-distilled-squad -e HF_TASK=question-answering starlette-transformers:cpu
+docker run -ti -p 5000:5000 -e HF_MODEL_DIR=/repository -v $(pwd)/distilbert-base-uncased-emotion:/repository starlette-transformers:cpu
 ```
 
 
@@ -56,6 +57,7 @@ curl --request POST \
 }'
 ```
 
+
 ---
 
 ## 🛠️ Environment variables
diff --git a/dockerfiles/pytorch/cpu/Dockerfile b/dockerfiles/pytorch/cpu/Dockerfile
new file mode 100644
index 00000000..551bfa9a
--- /dev/null
+++ b/dockerfiles/pytorch/cpu/Dockerfile
@@ -0,0 +1,52 @@
+FROM ubuntu:22.04
+
+LABEL maintainer="Hugging Face"
+
+ENV DEBIAN_FRONTEND=noninteractive
+
+RUN apt-get update \
+    && apt-get -y upgrade --only-upgrade systemd openssl cryptsetup \
+    && apt-get install -y \
+    bzip2 \
+    curl \
+    git \
+    git-lfs \
+    tar \
+    gcc \
+    g++ \
+    # audio
+    libsndfile1-dev \
+    ffmpeg \
+    && apt-get clean autoremove --yes \
+    && rm -rf /var/lib/{apt,dpkg,cache,log}
+
+# install micromamba
+ENV MAMBA_ROOT_PREFIX=/opt/conda
+ENV PATH=/opt/conda/bin:$PATH
+RUN curl -L https://micromamba.snakepit.net/api/micromamba/linux-64/latest | tar -xj "bin/micromamba" \
+    && touch /root/.bashrc \
+    && ./bin/micromamba shell init -s bash -p /opt/conda  \
+    && grep -v '[ -z "\$PS1" ] && return' /root/.bashrc  > /opt/conda/bashrc
+
+WORKDIR /app
+
+# install base python dependencies
+COPY dockerfiles/pytorch/cpu/environment.yaml /app/environment.yaml
+RUN micromamba install -y -n base -f environment.yaml \
+    && rm environment.yaml \
+    && micromamba clean --all --yes
+
+# install huggingface inference toolkit
+COPY requirements.txt /tmp/requirements.txt
+RUN pip install  --no-cache-dir -r /tmp/requirements.txt && rm /tmp/requirements.txt
+
+# copy application
+COPY src/huggingface_inference_toolkit huggingface_inference_toolkit
+COPY src/huggingface_inference_toolkit/webservice_starlette.py webservice_starlette.py
+
+# copy entrypoint and change permissions
+COPY scripts/entrypoint.sh entrypoint.sh
+RUN chmod +x entrypoint.sh
+
+# run app
+ENTRYPOINT ["/bin/bash", "entrypoint.sh"]
diff --git a/dockerfiles/pytorch/cpu/environment.yaml b/dockerfiles/pytorch/cpu/environment.yaml
new file mode 100644
index 00000000..4c3efe47
--- /dev/null
+++ b/dockerfiles/pytorch/cpu/environment.yaml
@@ -0,0 +1,12 @@
+name: base
+channels:
+- conda-forge
+dependencies:
+- python=3.9.13
+- pytorch::pytorch=1.13.1=py3.9_cpu_0
+- pip:
+  - transformers[sklearn,sentencepiece,audio,vision]==4.27.2
+  - sentence_transformers==2.2.2
+  - torchvision==0.14.1
+  - diffusers==0.14.0
+  - accelerate==0.17.1
\ No newline at end of file
diff --git a/dockerfiles/pytorch/gpu/Dockerfile b/dockerfiles/pytorch/gpu/Dockerfile
new file mode 100644
index 00000000..3db0d4b0
--- /dev/null
+++ b/dockerfiles/pytorch/gpu/Dockerfile
@@ -0,0 +1,54 @@
+FROM nvidia/cuda:11.7.0-devel-ubuntu22.04
+
+LABEL maintainer="Hugging Face"
+
+ENV DEBIAN_FRONTEND=noninteractive
+
+RUN apt-get update \
+    && apt-get -y upgrade --only-upgrade systemd openssl cryptsetup \
+    && apt-get install -y \
+    bzip2 \
+    curl \
+    git \
+    git-lfs \
+    tar \
+    gcc \
+    g++ \
+    # audio
+    libsndfile1-dev \
+    ffmpeg \
+    && apt-get clean autoremove --yes \
+    && rm -rf /var/lib/{apt,dpkg,cache,log}
+
+# install micromamba
+ENV MAMBA_ROOT_PREFIX=/opt/conda
+ENV PATH=/opt/conda/bin:$PATH
+ENV LD_LIBRARY_PATH="/opt/conda/lib:${LD_LIBRARY_PATH}"
+
+RUN curl -L https://micromamba.snakepit.net/api/micromamba/linux-64/latest | tar -xj "bin/micromamba" \
+    && touch /root/.bashrc \
+    && ./bin/micromamba shell init -s bash -p /opt/conda  \
+    && grep -v '[ -z "\$PS1" ] && return' /root/.bashrc  > /opt/conda/bashrc
+
+WORKDIR /app
+
+# install base python dependencies
+COPY dockerfiles/pytorch/gpu/environment.yaml /app/environment.yaml
+RUN micromamba install -y -n base -f environment.yaml \
+    && rm environment.yaml \
+    && micromamba clean --all --yes
+
+# install huggingface inference toolkit
+COPY requirements.txt /tmp/requirements.txt
+RUN pip install  --no-cache-dir -r /tmp/requirements.txt && rm /tmp/requirements.txt
+
+# copy application
+COPY src/huggingface_inference_toolkit huggingface_inference_toolkit
+COPY src/huggingface_inference_toolkit/webservice_starlette.py webservice_starlette.py
+
+# copy entrypoint and change permissions
+COPY scripts/entrypoint.sh entrypoint.sh
+RUN chmod +x entrypoint.sh
+
+# run app
+ENTRYPOINT ["/bin/bash", "entrypoint.sh"]
diff --git a/dockerfiles/pytorch/gpu/environment.yaml b/dockerfiles/pytorch/gpu/environment.yaml
new file mode 100644
index 00000000..e65a10bb
--- /dev/null
+++ b/dockerfiles/pytorch/gpu/environment.yaml
@@ -0,0 +1,13 @@
+name: base
+channels:
+- conda-forge
+dependencies:
+- python=3.9.13
+- nvidia::cudatoolkit=11.7
+- pytorch::pytorch=1.13.1=py3.9_cuda11.7*
+- pip:
+  - transformers[sklearn,sentencepiece,audio,vision]==4.27.2
+  - sentence_transformers==2.2.2
+  - torchvision==0.14.1
+  - diffusers==0.14.0
+  - accelerate==0.17.1
\ No newline at end of file
diff --git a/dockerfiles/robyn/Dockerfile.cpu b/dockerfiles/robyn/Dockerfile.cpu
deleted file mode 100644
index f7a884ce..00000000
--- a/dockerfiles/robyn/Dockerfile.cpu
+++ /dev/null
@@ -1,12 +0,0 @@
-FROM public.ecr.aws/t6m7g5n4/transformers-inference:4.20.1-pt1.11-cpu
-
-# install robyn framework
-COPY requirements.txt /tmp/requirements.txt
-RUN pip install -r /tmp/requirements.txt && rm /tmp/requirements.txt
-
-# copy application
-COPY src/huggingface_inference_toolkit huggingface_inference_toolkit
-COPY src/huggingface_inference_toolkit/webservice_robyn.py webservice_robyn.py
-
-# run app
-CMD ["webservice_robyn.py"]
\ No newline at end of file
diff --git a/dockerfiles/robyn/Dockerfile.gpu b/dockerfiles/robyn/Dockerfile.gpu
deleted file mode 100644
index d0170a1f..00000000
--- a/dockerfiles/robyn/Dockerfile.gpu
+++ /dev/null
@@ -1,12 +0,0 @@
-FROM public.ecr.aws/t6m7g5n4/transformers-inference:4.20.1-pt1.11-cuda11.5
-
-# install robyn framework
-COPY requirements.txt /tmp/requirements.txt
-RUN pip install -r /tmp/requirements.txt && rm /tmp/requirements.txt
-
-# copy application
-COPY src/huggingface_inference_toolkit huggingface_inference_toolkit
-COPY src/huggingface_inference_toolkit/webservice_robyn.py webservice_robyn.py
-
-# run app
-CMD ["webservice_robyn.py"]
\ No newline at end of file
diff --git a/dockerfiles/starlette/pytorch/Dockerfile.cpu b/dockerfiles/starlette/pytorch/Dockerfile.cpu
deleted file mode 100644
index 7007c201..00000000
--- a/dockerfiles/starlette/pytorch/Dockerfile.cpu
+++ /dev/null
@@ -1,24 +0,0 @@
-FROM huggingface/transformers-inference:4.24.0-pt1.13-cpu
-
-# install starlette framework
-COPY starlette_requirements.txt /tmp/requirements.txt
-RUN pip install  --no-cache-dir -r /tmp/requirements.txt && rm /tmp/requirements.txt
-
-# Think about a better solution -> base contaienr has pt 1.13. thats why need below 0.14
-RUN pip install  --no-cache-dir sentence_transformers=="2.2.2" torchvision=="0.14.0" diffusers=="0.9.0" accelerate=="0.14.0"
-
-# Add upgrade due to issue in base container upgrade https://github.com/mamba-org/mamba/issues/2170
-RUN pip install transformers==4.25.1  --no-cache-dir --upgrade 
-
-# copy application
-COPY src/huggingface_inference_toolkit huggingface_inference_toolkit
-COPY src/huggingface_inference_toolkit/webservice_starlette.py webservice_starlette.py
-
-# copy entrypoint and change permissions
-COPY scripts/entrypoint.sh entrypoint.sh
-RUN chmod +x entrypoint.sh
-
-# run app
-ENTRYPOINT ["/bin/bash", "entrypoint.sh"]
-
-
diff --git a/dockerfiles/starlette/pytorch/Dockerfile.gpu b/dockerfiles/starlette/pytorch/Dockerfile.gpu
deleted file mode 100644
index 6e585775..00000000
--- a/dockerfiles/starlette/pytorch/Dockerfile.gpu
+++ /dev/null
@@ -1,22 +0,0 @@
-FROM huggingface/transformers-inference:4.24.0-pt1.13-cuda11.6
-
-# install starlette framework
-COPY starlette_requirements.txt /tmp/requirements.txt
-RUN pip install  --no-cache-dir -r /tmp/requirements.txt && rm /tmp/requirements.txt
-
-# Think about a better solution -> base contaienr has pt 1.13. thats why need below 0.14
-RUN pip install  --no-cache-dir sentence_transformers=="2.2.2" torchvision=="0.14.0" diffusers=="0.9.0" accelerate=="0.14.0" --extra-index-url https://download.pytorch.org/whl/cu116
-
-# Add upgrade due to issue in base container upgrade https://github.com/mamba-org/mamba/issues/2170
-RUN pip install transformers==4.25.1  --no-cache-dir --upgrade 
-
-# copy application
-COPY src/huggingface_inference_toolkit huggingface_inference_toolkit
-COPY src/huggingface_inference_toolkit/webservice_starlette.py webservice_starlette.py
-
-# copy entrypoint and change permissions
-COPY scripts/entrypoint.sh entrypoint.sh
-RUN chmod +x entrypoint.sh
-
-# run app
-ENTRYPOINT ["/bin/bash", "entrypoint.sh"]
\ No newline at end of file
diff --git a/dockerfiles/starlette/tensorflow/Dockerfile.cpu b/dockerfiles/starlette/tensorflow/Dockerfile.cpu
deleted file mode 100644
index e8061022..00000000
--- a/dockerfiles/starlette/tensorflow/Dockerfile.cpu
+++ /dev/null
@@ -1,16 +0,0 @@
-FROM huggingface/transformers-inference:4.24.0-tf2.9-cpu
-
-# install starlette framework
-COPY starlette_requirements.txt /tmp/requirements.txt
-RUN pip install  --no-cache-dir -r /tmp/requirements.txt && rm /tmp/requirements.txt
-
-# copy application
-COPY src/huggingface_inference_toolkit huggingface_inference_toolkit
-COPY src/huggingface_inference_toolkit/webservice_starlette.py webservice_starlette.py
-
-# copy entrypoint and change permissions
-COPY scripts/entrypoint.sh entrypoint.sh
-RUN chmod +x entrypoint.sh
-
-# run app
-ENTRYPOINT ["/bin/bash", "entrypoint.sh"]
\ No newline at end of file
diff --git a/dockerfiles/starlette/tensorflow/Dockerfile.gpu b/dockerfiles/starlette/tensorflow/Dockerfile.gpu
deleted file mode 100644
index 5ae9abd0..00000000
--- a/dockerfiles/starlette/tensorflow/Dockerfile.gpu
+++ /dev/null
@@ -1,16 +0,0 @@
-FROM huggingface/transformers-inference:4.24.0-tf2.9-cuda11.2
-
-# install starlette framework
-COPY starlette_requirements.txt /tmp/requirements.txt
-RUN pip install  --no-cache-dir -r /tmp/requirements.txt && rm /tmp/requirements.txt
-
-# copy application
-COPY src/huggingface_inference_toolkit huggingface_inference_toolkit
-COPY src/huggingface_inference_toolkit/webservice_starlette.py webservice_starlette.py
-
-# copy entrypoint and change permissions
-COPY scripts/entrypoint.sh entrypoint.sh
-RUN chmod +x entrypoint.sh
-
-# run app
-ENTRYPOINT ["/bin/bash", "entrypoint.sh"]
\ No newline at end of file
diff --git a/dockerfiles/tensorflow/cpu/Dockerfile b/dockerfiles/tensorflow/cpu/Dockerfile
new file mode 100644
index 00000000..b0443845
--- /dev/null
+++ b/dockerfiles/tensorflow/cpu/Dockerfile
@@ -0,0 +1,52 @@
+FROM ubuntu:22.04
+
+LABEL maintainer="Hugging Face"
+
+ENV DEBIAN_FRONTEND=noninteractive
+
+RUN apt-get update \
+    && apt-get -y upgrade --only-upgrade systemd openssl cryptsetup \
+    && apt-get install -y \
+    bzip2 \
+    curl \
+    git \
+    git-lfs \
+    tar \
+    gcc \
+    g++ \
+    # audio
+    libsndfile1-dev \
+    ffmpeg \
+    && apt-get clean autoremove --yes \
+    && rm -rf /var/lib/{apt,dpkg,cache,log}
+
+# install micromamba
+ENV MAMBA_ROOT_PREFIX=/opt/conda
+ENV PATH=/opt/conda/bin:$PATH
+RUN curl -L https://micromamba.snakepit.net/api/micromamba/linux-64/latest | tar -xj "bin/micromamba" \
+    && touch /root/.bashrc \
+    && ./bin/micromamba shell init -s bash -p /opt/conda  \
+    && grep -v '[ -z "\$PS1" ] && return' /root/.bashrc  > /opt/conda/bashrc
+
+WORKDIR /app
+
+# install base python dependencies
+COPY dockerfiles/tensorflow/cpu/environment.yaml /app/environment.yaml
+RUN micromamba install -y -n base -f environment.yaml \
+    && rm environment.yaml \
+    && micromamba clean --all --yes
+
+# install huggingface inference toolkit
+COPY requirements.txt /tmp/requirements.txt
+RUN pip install  --no-cache-dir -r /tmp/requirements.txt && rm /tmp/requirements.txt
+
+# copy application
+COPY src/huggingface_inference_toolkit huggingface_inference_toolkit
+COPY src/huggingface_inference_toolkit/webservice_starlette.py webservice_starlette.py
+
+# copy entrypoint and change permissions
+COPY scripts/entrypoint.sh entrypoint.sh
+RUN chmod +x entrypoint.sh
+
+# run app
+ENTRYPOINT ["/bin/bash", "entrypoint.sh"]
diff --git a/dockerfiles/tensorflow/cpu/environment.yaml b/dockerfiles/tensorflow/cpu/environment.yaml
new file mode 100644
index 00000000..a370380c
--- /dev/null
+++ b/dockerfiles/tensorflow/cpu/environment.yaml
@@ -0,0 +1,8 @@
+name: base
+channels:
+- conda-forge
+dependencies:
+- python=3.9.13
+- tensorflow=2.9.1=*cpu*py39*
+- pip:
+  - transformers[sklearn,sentencepiece,audio,vision]==4.27.2
\ No newline at end of file
diff --git a/dockerfiles/tensorflow/gpu/Dockerfile b/dockerfiles/tensorflow/gpu/Dockerfile
new file mode 100644
index 00000000..785bcc79
--- /dev/null
+++ b/dockerfiles/tensorflow/gpu/Dockerfile
@@ -0,0 +1,55 @@
+FROM nvidia/cuda:11.2.2-base-ubuntu20.04
+
+LABEL maintainer="Hugging Face"
+
+ENV DEBIAN_FRONTEND=noninteractive
+ENV CONDA_OVERRIDE_CUDA="11.2"
+
+RUN apt-get update \
+    && apt-get -y upgrade --only-upgrade systemd openssl cryptsetup \
+    && apt-get install -y \
+    bzip2 \
+    curl \
+    git \
+    git-lfs \
+    tar \
+    gcc \
+    g++ \
+    # audio
+    libsndfile1-dev \
+    ffmpeg \
+    && apt-get clean autoremove --yes \
+    && rm -rf /var/lib/{apt,dpkg,cache,log}
+
+# install micromamba
+ENV MAMBA_ROOT_PREFIX=/opt/conda
+ENV PATH=/opt/conda/bin:$PATH
+ENV LD_LIBRARY_PATH="/opt/conda/lib:${LD_LIBRARY_PATH}"
+
+RUN curl -L https://micromamba.snakepit.net/api/micromamba/linux-64/latest | tar -xj "bin/micromamba" \
+    && touch /root/.bashrc \
+    && ./bin/micromamba shell init -s bash -p /opt/conda  \
+    && grep -v '[ -z "\$PS1" ] && return' /root/.bashrc  > /opt/conda/bashrc
+
+WORKDIR /app
+
+# install base python dependencies
+COPY dockerfiles/tensorflow/gpu/environment.yaml /app/environment.yaml
+RUN micromamba install -y -n base -f environment.yaml \
+    && rm environment.yaml \
+    && micromamba clean --all --yes
+
+# install huggingface inference toolkit
+COPY requirements.txt /tmp/requirements.txt
+RUN pip install  --no-cache-dir -r /tmp/requirements.txt && rm /tmp/requirements.txt
+
+# copy application
+COPY src/huggingface_inference_toolkit huggingface_inference_toolkit
+COPY src/huggingface_inference_toolkit/webservice_starlette.py webservice_starlette.py
+
+# copy entrypoint and change permissions
+COPY scripts/entrypoint.sh entrypoint.sh
+RUN chmod +x entrypoint.sh
+
+# run app
+ENTRYPOINT ["/bin/bash", "entrypoint.sh"]
diff --git a/dockerfiles/tensorflow/gpu/environment.yaml b/dockerfiles/tensorflow/gpu/environment.yaml
new file mode 100644
index 00000000..1d886795
--- /dev/null
+++ b/dockerfiles/tensorflow/gpu/environment.yaml
@@ -0,0 +1,9 @@
+name: base
+channels:
+- conda-forge
+dependencies:
+- python=3.9.13
+- nvidia::cudatoolkit=11.7
+- tensorflow=2.9.1=*cuda112*py39*
+- pip:
+  - transformers[sklearn,sentencepiece,audio,vision]==4.27.2
\ No newline at end of file
diff --git a/makefile b/makefile
index fec9bc02..49855723 100644
--- a/makefile
+++ b/makefile
@@ -13,12 +13,9 @@ integ-test:
 # Check that source code meets quality standards
 
 quality:
-	black --check --line-length 119 --target-version py36 $(check_dirs)
-	isort --check-only $(check_dirs)
-	flake8 src
+	ruff $(check_dirs) 
 
 # Format source code automatically
 
-style:
-	black --line-length 119 --target-version py36 $(check_dirs)
-	isort $(check_dirs)
\ No newline at end of file
+style: 
+	ruff $(check_dirs) --fix
\ No newline at end of file
diff --git a/pyproject.toml b/pyproject.toml
index 211fdad8..96ef9084 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,3 +1,34 @@
-[tool.black]
+[tool.mypy]
+ignore_missing_imports = true
+no_implicit_optional = true
+scripts_are_modules = true
+
+[tool.ruff]
+select = [
+  "E", # pycodestyle errors
+  "W", # pycodestyle warnings
+  "F", # pyflakes
+  "I", # isort
+  "C", # flake8-comprehensions
+  "B", # flake8-bugbear
+]
+ignore = [
+  "E501", # line too long, handled by black
+  "B008", # do not perform function calls in argument defaults
+  "C901", # too complex
+]
+# Same as Black.
 line-length = 119
-target-version = ['py39']
+
+# Allow unused variables when underscore-prefixed.
+dummy-variable-rgx = "^(_+|(_+[a-zA-Z0-9_]*[a-zA-Z0-9]+?))$"
+
+# Assume Python 3.8.
+target-version = "py39"
+
+[tool.ruff.per-file-ignores]
+"__init__.py" = ["F401"]
+
+[tool.isort]
+profile = "black"
+known_third_party = ["transforemrs", "starlette", "huggingface_hub"]
diff --git a/requirements.txt b/requirements.txt
index 63c5474a..8a178f8d 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,2 +1,5 @@
 orjson
-robyn
\ No newline at end of file
+starlette
+uvicorn
+pandas
+huggingface_hub>=0.13.2
\ No newline at end of file
diff --git a/setup.py b/setup.py
index c11fa81d..92132915 100644
--- a/setup.py
+++ b/setup.py
@@ -17,6 +17,7 @@
 install_requires = [
     # transformers
     "transformers[sklearn,sentencepiece]>=4.25.1",
+    "huggingface_hub>=0.13.3",
     # api stuff
     "orjson",
     # "robyn",
@@ -54,6 +55,7 @@
     "black",
     "isort",
     "flake8",
+    "ruff",
 ]
 
 setup(
diff --git a/src/huggingface_inference_toolkit/async_utils.py b/src/huggingface_inference_toolkit/async_utils.py
index 8b73beba..5b6af3fd 100644
--- a/src/huggingface_inference_toolkit/async_utils.py
+++ b/src/huggingface_inference_toolkit/async_utils.py
@@ -5,7 +5,6 @@
 from anyio import Semaphore
 from typing_extensions import ParamSpec
 
-
 # To not have too many threads running (which could happen on too many concurrent
 # requests, we limit it with a semaphore.
 MAX_CONCURRENT_THREADS = 1
diff --git a/src/huggingface_inference_toolkit/const.py b/src/huggingface_inference_toolkit/const.py
index d7dc3688..993fea26 100644
--- a/src/huggingface_inference_toolkit/const.py
+++ b/src/huggingface_inference_toolkit/const.py
@@ -1,7 +1,6 @@
 import os
 from pathlib import Path
 
-
 HF_MODEL_DIR = os.environ.get("HF_MODEL_DIR", "/opt/huggingface/model")
 HF_MODEL_ID = os.environ.get("HF_MODEL_ID", None)
 HF_TASK = os.environ.get("HF_TASK", None)
diff --git a/src/huggingface_inference_toolkit/diffusers_utils.py b/src/huggingface_inference_toolkit/diffusers_utils.py
index 2bb27d81..36cdcb22 100644
--- a/src/huggingface_inference_toolkit/diffusers_utils.py
+++ b/src/huggingface_inference_toolkit/diffusers_utils.py
@@ -2,7 +2,6 @@
 import json
 import os
 
-
 _diffusers = importlib.util.find_spec("diffusers") is not None
 
 
@@ -12,7 +11,6 @@ def is_diffusers_available():
 
 if is_diffusers_available():
     import torch
-
     from diffusers import DPMSolverMultistepScheduler, StableDiffusionPipeline
 
 
diff --git a/src/huggingface_inference_toolkit/handler.py b/src/huggingface_inference_toolkit/handler.py
index e3fbb2e2..993e4967 100644
--- a/src/huggingface_inference_toolkit/handler.py
+++ b/src/huggingface_inference_toolkit/handler.py
@@ -1,16 +1,14 @@
 import logging
-from abc import ABC
 from pathlib import Path
 from typing import Optional, Union
 
 from huggingface_inference_toolkit.utils import check_and_register_custom_pipeline_from_directory, get_pipeline
 
-
 logger = logging.getLogger(__name__)
 logging.basicConfig(format="%(asctime)s | %(levelname)s | %(message)s", level=logging.INFO)
 
 
-class HuggingFaceHandler(ABC):
+class HuggingFaceHandler:
     """
     A Default Hugging Face Inference Handler which works with all transformers pipelines, Sentence Transformers and Optimum.
     """
diff --git a/src/huggingface_inference_toolkit/sentence_transformers_utils.py b/src/huggingface_inference_toolkit/sentence_transformers_utils.py
index a3f79ded..2a3c0055 100644
--- a/src/huggingface_inference_toolkit/sentence_transformers_utils.py
+++ b/src/huggingface_inference_toolkit/sentence_transformers_utils.py
@@ -1,6 +1,5 @@
 import importlib.util
 
-
 _sentence_transformers = importlib.util.find_spec("sentence_transformers") is not None
 
 
diff --git a/src/huggingface_inference_toolkit/serialization/base.py b/src/huggingface_inference_toolkit/serialization/base.py
index d3b936d9..eb965b64 100644
--- a/src/huggingface_inference_toolkit/serialization/base.py
+++ b/src/huggingface_inference_toolkit/serialization/base.py
@@ -2,7 +2,6 @@
 from huggingface_inference_toolkit.serialization.image_utils import Imager
 from huggingface_inference_toolkit.serialization.json_utils import Jsoner
 
-
 content_type_mapping = {
     "application/json": Jsoner,
     "text/csv": None,
diff --git a/src/huggingface_inference_toolkit/utils.py b/src/huggingface_inference_toolkit/utils.py
index e39c22af..a77dd655 100644
--- a/src/huggingface_inference_toolkit/utils.py
+++ b/src/huggingface_inference_toolkit/utils.py
@@ -1,13 +1,10 @@
 import importlib.util
 import logging
-import os
 import sys
 from pathlib import Path
 from typing import Optional, Union
 
-from huggingface_hub import HfApi, login
-from huggingface_hub.file_download import cached_download, hf_hub_url
-from huggingface_hub.utils import filter_repo_objects
+from huggingface_hub import login, snapshot_download
 from transformers import WhisperForConditionalGeneration, pipeline
 from transformers.file_utils import is_tf_available, is_torch_available
 from transformers.pipelines import Conversation, Pipeline
@@ -23,7 +20,6 @@
     is_sentence_transformers_available,
 )
 
-
 logger = logging.getLogger(__name__)
 logging.basicConfig(format="%(asctime)s | %(levelname)s | %(message)s", level=logging.INFO)
 
@@ -51,6 +47,11 @@ def is_optimum_available():
     "flax": "flax*",
     "rust": "rust*",
     "onnx": "*onnx",
+    "safetensors": "*safetensors",
+    "coreml": "*mlmodel",
+    "tflite": "*tflite",
+    "savedmodel": "*tar.gz",
+    "ckpt": "*ckpt",
 }
 
 
@@ -58,7 +59,18 @@ def create_artifact_filter(framework):
     """
     Returns a list of regex pattern based on the DL Framework. which will be to used to ignore files when downloading
     """
-    ignore_regex_list = ["pytorch*", "tf*", "flax*", "rust*", "*onnx"]
+    ignore_regex_list = [
+        "pytorch*",
+        "tf*",
+        "flax*",
+        "rust*",
+        "*onnx",
+        "*safetensors",
+        "*mlmodel",
+        "*tflite",
+        "*tar.gz",
+        "*ckpt",
+    ]
     pattern = framework2weight.get(framework, None)
     if pattern in ignore_regex_list:
         ignore_regex_list.remove(pattern)
@@ -147,46 +159,17 @@ def _load_repository_from_hf(
 
     # create regex to only include the framework specific weights
     ignore_regex = create_artifact_filter(framework)
+    logger.info(f"Ignore regex pattern for files, which are not downloaded: { ', '.join(ignore_regex) }")
 
-    # get image artifact files
-    _api = HfApi()
-    repo_info = _api.repo_info(
-        repo_id=repository_id,
-        repo_type="model",
+    # Download the repository to the workdir and filter out non-framework specific weights
+    snapshot_download(
+        repository_id,
         revision=revision,
-    )
-    # apply regex to filter out non-framework specific weights if args.framework is set
-    filtered_repo_files = filter_repo_objects(
-        items=[f.rfilename for f in repo_info.siblings],
+        local_dir=str(target_dir),
+        local_dir_use_symlinks=False,
         ignore_patterns=ignore_regex,
     )
 
-    # iterate over all files and download them
-    for repo_file in filtered_repo_files:
-        url = hf_hub_url(repo_id=repository_id, filename=repo_file, revision=revision)
-
-        # define values if repo has nested strucutre
-        if isinstance(repo_file, str):
-            repo_file = Path(repo_file)
-
-        repo_file_is_dir = repo_file.parent != Path(".")
-        real_target_dir = target_dir / repo_file.parent if repo_file_is_dir else target_dir
-        real_file_name = str(repo_file.name) if repo_file_is_dir else repo_file
-
-        # download files
-        path = cached_download(
-            url,
-            cache_dir=real_target_dir.as_posix(),
-            force_filename=real_file_name,
-            use_auth_token=hf_hub_token,
-        )
-        if os.path.exists(path + ".lock"):
-            os.remove(path + ".lock")
-
-    # create requirements.txt if not exists
-    if not (target_dir / "requirements.txt").exists():
-        target_dir.joinpath("requirements.txt").touch()
-
     return target_dir
 
 
diff --git a/src/huggingface_inference_toolkit/webservice_robyn.py b/src/huggingface_inference_toolkit/webservice_robyn.py
index aa677a93..a1c437af 100644
--- a/src/huggingface_inference_toolkit/webservice_robyn.py
+++ b/src/huggingface_inference_toolkit/webservice_robyn.py
@@ -6,7 +6,6 @@
 from huggingface_inference_toolkit.serialization.base import ContentType
 from huggingface_inference_toolkit.serialization.json_utils import Jsoner
 
-
 logger = logging.getLogger(__name__)
 logging.basicConfig(format="%(asctime)s | %(levelname)s | %(message)s", level=logging.INFO)
 
diff --git a/src/huggingface_inference_toolkit/webservice_starlette.py b/src/huggingface_inference_toolkit/webservice_starlette.py
index 9c3df0ee..a8581992 100644
--- a/src/huggingface_inference_toolkit/webservice_starlette.py
+++ b/src/huggingface_inference_toolkit/webservice_starlette.py
@@ -3,6 +3,10 @@
 from time import perf_counter
 
 import orjson
+from starlette.applications import Starlette
+from starlette.responses import PlainTextResponse, Response
+from starlette.routing import Route
+
 from huggingface_inference_toolkit.async_utils import async_handler_call
 from huggingface_inference_toolkit.const import (
     HF_FRAMEWORK,
@@ -16,9 +20,6 @@
 from huggingface_inference_toolkit.serialization.base import ContentType
 from huggingface_inference_toolkit.serialization.json_utils import Jsoner
 from huggingface_inference_toolkit.utils import _load_repository_from_hf
-from starlette.applications import Starlette
-from starlette.responses import PlainTextResponse, Response
-from starlette.routing import Route
 
 
 def config_logging(level=logging.INFO):
diff --git a/starlette_requirements.txt b/starlette_requirements.txt
deleted file mode 100644
index 6139a106..00000000
--- a/starlette_requirements.txt
+++ /dev/null
@@ -1,5 +0,0 @@
-orjson
-starlette
-uvicorn
-pandas
-huggingface_hub>=0.11.0
\ No newline at end of file
diff --git a/tests/unit/test_utils.py b/tests/unit/test_utils.py
index 9ac0603c..9d5052ee 100644
--- a/tests/unit/test_utils.py
+++ b/tests/unit/test_utils.py
@@ -44,8 +44,6 @@ def test_load_tensorflow_repository_from_hf():
         # folder contains all config files and pytorch_model.bin
         folder_contents = os.listdir(storage_folder)
         assert "pytorch_model.bin" not in folder_contents
-        # custom requirements.txt for custom handler
-        assert "requirements.txt" in folder_contents
         # filter framework
         assert "tf_model.h5" in folder_contents
         # revision doesn't have tokenizer
@@ -62,8 +60,6 @@ def test_load_onnx_repository_from_hf():
         # folder contains all config files and pytorch_model.bin
         folder_contents = os.listdir(storage_folder)
         assert "pytorch_model.bin" not in folder_contents
-        # custom requirements.txt for custom handler
-        assert "requirements.txt" in folder_contents
         # filter framework
         assert "tf_model.h5" not in folder_contents
         # onnx model
@@ -85,8 +81,6 @@ def test_load_pytorch_repository_from_hf():
         # folder contains all config files and pytorch_model.bin
         folder_contents = os.listdir(storage_folder)
         assert "pytorch_model.bin" in folder_contents
-        # custom requirements.txt for custom handler
-        assert "requirements.txt" in folder_contents
         # filter framework
         assert "tf_model.h5" not in folder_contents
         # revision doesn't have tokenizer
@@ -126,7 +120,6 @@ def test_get_pipeline():
 @require_torch
 def test_whisper_long_audio():
     with tempfile.TemporaryDirectory() as tmpdirname:
-
         storage_dir = _load_repository_from_hf("openai/whisper-tiny", tmpdirname, framework="pytorch")
         pipe = get_pipeline("automatic-speech-recognition", storage_dir.as_posix())
         res = pipe(os.path.join(os.getcwd(), "tests/resources/audio", "long_sample.mp3"))