update

finetune/Dockerfile

@@ -339,7 +339,7 @@ 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://ghproxy.dockless.eu.org/https://github.com/oneapi-src/oneCCL.git ${STAGE_DIR}/oneCCL
+git clone https://github.com/oneapi-src/oneCCL.git ${STAGE_DIR}/oneCCL
 cd ${STAGE_DIR}/oneCCL
 git checkout . 
 git checkout master
@@ -353,7 +353,7 @@ RUN <<EOT
 #!/bin/bash
 source /opt/conda/etc/profile.d/conda.sh
 conda activate ${CONDA_ENV_NAME}
-git clone https://ghproxy.dockless.eu.org/https://github.com/microsoft/DeepSpeed-Kernels.git ${STAGE_DIR}/DeepSpeed-Kernels
+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
 pip install dist/deepspeed_kernels-*.whl
@@ -364,7 +364,7 @@ RUN <<EOT
 #!/bin/bash
 source /opt/conda/etc/profile.d/conda.sh
 conda activate ${CONDA_ENV_NAME}
-git clone https://ghproxy.dockless.eu.org/https://github.com/microsoft/DeepSpeed.git ${STAGE_DIR}/DeepSpeed
+git clone https://github.com/microsoft/DeepSpeed.git ${STAGE_DIR}/DeepSpeed
 cd ${STAGE_DIR}/DeepSpeed 
 git checkout . 
 # git checkout v${DEEPSPEED_VERSION}

finetune/docker-compose_pytorch2.3.yml

@@ -19,11 +19,11 @@ services:
       context: .
       dockerfile: Dockerfile
       args: # PyTorch版本、Python版本与pytorch_lightning版本的对应关系表 https://blog.csdn.net/qq_41813454/article/details/137421822
-        PYTHON_VERSION: 3.10
-        CUDA_VERSION: 12.1.0
-        PYTORCH_VERSION: 2.3.1
-        TORCHVISION_VERSION: 0.18.0
-        TORCHAUDIO_VERSION: 2.3.0
+        PYTHON_VERSION: "3.10"
+        CUDA_VERSION: "12.1.0"
+        PYTORCH_VERSION: "2.3.0"
+        TORCHVISION_VERSION: "0.18.0"
+        TORCHAUDIO_VERSION: "2.3.0"
         DS_BUILD_OPS: 1
         DS_BUILD_SPARSE_ATTN: 0
         DS_BUILD_FUSED_ADAM: 1
@@ -32,7 +32,7 @@ services:
         USE_ROCM: 0
         USE_XPU: 0
         CUDA: cu121
-        CUDA_ARCH_LIST: "80;86;89;90;90a" # for RTX 4090, all : "80;86;89;90"
+        CUDA_ARCH_LIST: "80;86;89;90" # for RTX 4090, all : "80;86;89;90"
         SETUPTOOLS_VERSION: "69.5.1"
         DCUTLASS_NVCC_ARCHS: "90a" # 90a for H100 GPU 89:GeForce RTX 4090 
     volumes:
@@ -44,7 +44,7 @@ services:
     image: hotwa/deepspeed:pt23
     shm_size: '32gb'
     ports:
-      - 3227:2222
+      - 3228:2222
     environment:
       - NVIDIA_VISIBLE_DEVICES=all
       - NVIDIA_DRIVER_CAPABILITIES=compute,utility