diff --git a/docker/Dockerfile.ubuntu b/docker/Dockerfile.ubuntu index 252a13a9..b163a42b 100644 --- a/docker/Dockerfile.ubuntu +++ b/docker/Dockerfile.ubuntu @@ -80,14 +80,14 @@ RUN if [ "$INSTALL_MS_DEPS" = "True" ]; then \ fi; \ if [ "$INSTALL_MEGATRON_DEPS" = "True" ]; then \ pip install "sglang[all]<0.4.7" math_verify -U && \ - pip install liger_kernel nvitop pre-commit "transformers<4.52" huggingface-hub -U && \ + pip install "liger_kernel<0.6" nvitop pre-commit "transformers<4.52" 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@release_v2.3; \ + pip install --no-build-isolation git+https://github.com/NVIDIA/TransformerEngine.git@release_v2.5#egg=transformer_engine[pytorch]; \ cd /tmp && GIT_LFS_SKIP_SMUDGE=1 git clone https://github.com/NVIDIA/apex && \ 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_r0.12.0; \ + 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 cde80086..8ebe2f8d 100644 --- a/docker/build_image.py +++ b/docker/build_image.py @@ -340,6 +340,11 @@ class LLMImageBuilder(Builder): class SwiftImageBuilder(LLMImageBuilder): + def init_args(self, args) -> Any: + if not args.lmdeploy_version: + args.lmdeploy_version = '0.8.0' + return super().init_args(args) + def generate_dockerfile(self) -> str: meta_file = './docker/install.sh' with open('docker/Dockerfile.extra_install', 'r') as f: @@ -348,7 +353,7 @@ class SwiftImageBuilder(LLMImageBuilder): self.args.python_version) extra_content += """ RUN pip install --no-cache-dir -U "deepspeed==0.16.*" --no-deps && \ - pip install --no-cache-dir -U icecream soundfile pybind11 + 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} '