1. refine cpu Dockerfile for better performance 2. add ipex_bert exam…

…ple (#257)

Signed-off-by: Yao, Matrix <matrix.yao@intel.com>

docker/cpu/Dockerfile

@@ -21,13 +21,13 @@ ENV DEBIAN_FRONTEND noninteractive
 ENV PATH="/home/user/.local/bin:${PATH}"
 RUN apt-get update && apt-get install -y --no-install-recommends \
     libgl1-mesa-dev libglib2.0-0 \
-    sudo build-essential git bash-completion \
+    sudo build-essential git bash-completion numactl \
     python3.10 python3-pip python3.10-dev google-perftools && \
     apt-get clean && rm -rf /var/lib/apt/lists/* && \
     update-alternatives --install /usr/bin/python python /usr/bin/python3.10 1 && \
-    pip install --no-cache-dir --upgrade pip setuptools wheel
+    pip install --no-cache-dir --upgrade pip setuptools wheel intel-openmp
 
-ENV LD_PRELOAD="/usr/lib/x86_64-linux-gnu/libtcmalloc_minimal.so.4"
+ENV LD_PRELOAD="/usr/lib/x86_64-linux-gnu/libtcmalloc_minimal.so.4:/usr/local/lib/libiomp5.so"
 
 # Install PyTorch
 ARG TORCH_VERSION=stable

examples/ipex_bert.yaml

@@ -0,0 +1,29 @@
+defaults:
+  - benchmark
+  - scenario: inference
+  - launcher: process
+  - backend: ipex
+  - _base_
+  - _self_
+
+name: ipex_bert
+
+launcher:
+  numactl: true
+  numactl_kwargs:
+    cpunodebind: 0
+    membind: 0
+
+scenario:
+  latency: true
+  memory: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 128
+
+backend:
+  device: cpu
+  no_weights: true
+  export: true
+  torch_dtype: bfloat16
+  model: bert-base-uncased