From 2f3ac0fcd22a54050b65a9b3428f4980f98644f2 Mon Sep 17 00:00:00 2001
From: Jintao <huangjintao.hjt@alibaba-inc.com>
Date: Tue, 16 Sep 2025 12:49:29 +0800
Subject: [PATCH] [docker] update swift docker (#1502)

---
 docker/Dockerfile.ubuntu |  8 +++++---
 docker/build_image.py    | 20 ++++++++++++++++++--
 2 files changed, 23 insertions(+), 5 deletions(-)

diff --git a/docker/Dockerfile.ubuntu b/docker/Dockerfile.ubuntu
index f72e4cc9..e889c8d7 100644
--- a/docker/Dockerfile.ubuntu
+++ b/docker/Dockerfile.ubuntu
@@ -79,13 +79,15 @@ RUN if [ "$INSTALL_MS_DEPS" = "True" ]; then \
     pip install --no-cache-dir huggingface-hub transformers peft -U; \
 fi; \
 if [ "$INSTALL_MEGATRON_DEPS" = "True" ]; then \
-    pip install liger_kernel nvitop pre-commit transformers huggingface-hub -U && \
+    pip install "sglang[all]<0.5" "math_verify==0.5.2" "gradio<5.33" -U && \
+    pip install liger_kernel wandb swanlab nvitop pre-commit "transformers<4.57" "trl<0.21" huggingface-hub -U && \
     SITE_PACKAGES=$(python -c "import site; print(site.getsitepackages()[0])") && echo $SITE_PACKAGES && \
     CUDNN_PATH=$SITE_PACKAGES/nvidia/cudnn CPLUS_INCLUDE_PATH=$SITE_PACKAGES/nvidia/cudnn/include \
-    pip install git+https://github.com/NVIDIA/TransformerEngine.git@stable; \
+    pip install --no-build-isolation transformer_engine[pytorch]; \
     cd /tmp && GIT_LFS_SKIP_SMUDGE=1 git clone https://github.com/NVIDIA/apex && \
-    cd apex && pip install -v --disable-pip-version-check --no-cache-dir --no-build-isolation --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" ./ && \
+    cd apex && git checkout e13873debc4699d39c6861074b9a3b2a02327f92 && pip install -v --disable-pip-version-check --no-cache-dir --no-build-isolation --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" ./ && \
     cd / && rm -fr /tmp/apex && pip cache purge; \
+    pip install git+https://github.com/NVIDIA/Megatron-LM.git@core_r0.13.0; \
 fi
 
 # install nvm and set node version to 18
diff --git a/docker/build_image.py b/docker/build_image.py
index e1b6ca14..07c616f4 100644
--- a/docker/build_image.py
+++ b/docker/build_image.py
@@ -344,6 +344,23 @@ class LLMImageBuilder(Builder):
 
 class SwiftImageBuilder(LLMImageBuilder):
 
+    def init_args(self, args) -> Any:
+        if not args.base_image:
+            args.base_image = 'nvidia/cuda:12.6.3-devel-ubuntu22.04'
+        if not args.cuda_version:
+            args.cuda_version = '12.6.3'
+        if not args.torch_version:
+            args.torch_version = '2.7.1'
+            args.torchaudio_version = '2.7.1'
+            args.torchvision_version = '0.22.1'
+        if not args.vllm_version:
+            args.vllm_version = '0.10.1.1'
+        if not args.lmdeploy_version:
+            args.lmdeploy_version = '0.9.2.post1'
+        if not args.flashattn_version:
+            args.flashattn_version = '2.7.4.post1'
+        return super().init_args(args)
+
     def generate_dockerfile(self) -> str:
         meta_file = './docker/install.sh'
         with open('docker/Dockerfile.extra_install', 'r') as f:
@@ -351,8 +368,7 @@ class SwiftImageBuilder(LLMImageBuilder):
             extra_content = extra_content.replace('{python_version}',
                                                   self.args.python_version)
         extra_content += """
-RUN pip install --no-cache-dir deepspeed==0.14.5 --no-deps && \
-    pip install --no-cache-dir -U icecream soundfile pybind11
+RUN pip install --no-cache-dir -U icecream soundfile pybind11 py-spy
 """
         version_args = (
             f'{self.args.torch_version} {self.args.torchvision_version} {self.args.torchaudio_version} '