add oneapi

finetune/Dockerfile

@@ -322,6 +322,15 @@ RUN <<EOT
 source /opt/conda/etc/profile.d/conda.sh
 conda activate ${CONDA_ENV_NAME}
 python -m pip install setuptools==${SETUPTOOLS_VERSION}
+# install oneapi for deepspeed
+git clone https://github.com/oneapi-src/oneCCL.git ${STAGE_DIR}/oneCCL
+cd ${STAGE_DIR}/oneCCL
+git checkout . 
+git checkout master
+mkdir build
+cd build 
+cmake .. -DCMAKE_INSTALL_PREFIX=/usr/local
+make -j install
 git clone https://github.com/microsoft/DeepSpeed-Kernels.git ${STAGE_DIR}/DeepSpeed-Kernels
 cd ${STAGE_DIR}/DeepSpeed-Kernels
 CUDA_ARCH_LIST=${CUDA_ARCH_LIST} python setup.py bdist_wheel

finetune/docker-compose_pytorch1.13.yml

@@ -25,7 +25,7 @@ services:
     pull_policy: if_not_present
     tty: true
     restart: unless-stopped
-    image: hotwa/deepspeed:test
+    image: hotwa/deepspeed:py113
     shm_size: '32gb'
     ports:
       - 3227:2222

finetune/docker-compose_pytorch2.3.yml

@@ -24,7 +24,7 @@ services:
     pull_policy: if_not_present
     tty: true
     restart: unless-stopped
-    image: hotwa/deepspeed:test
+    image: hotwa/deepspeed:pt23
     shm_size: '32gb'
     ports:
       - 3227:2222