commit a203934033a2e8d141adf84f3fcdb2cbbf11fbe8 Author: wehub-resource-sync Date: Mon Jul 13 13:34:58 2026 +0800 chore: import upstream snapshot with attribution diff --git a/.dev_scripts/build_docs.sh b/.dev_scripts/build_docs.sh new file mode 100644 index 0000000..dc76e6f --- /dev/null +++ b/.dev_scripts/build_docs.sh @@ -0,0 +1,8 @@ +pip install -r requirements/docs.txt +cd docs +rm -rf build + +# update api rst +#rm -rf source/api/ +#sphinx-apidoc --module-first -o source/api/ ../modelscope/ +make html diff --git a/.dev_scripts/ci_container_test.sh b/.dev_scripts/ci_container_test.sh new file mode 100644 index 0000000..c648fec --- /dev/null +++ b/.dev_scripts/ci_container_test.sh @@ -0,0 +1,242 @@ +NPU_TORCH_VERSION=${NPU_TORCH_VERSION:-2.7.1} +NPU_TORCH_NPU_VERSION=${NPU_TORCH_NPU_VERSION:-2.7.1.post2} +NPU_MODELSCOPE_VERSION=${NPU_MODELSCOPE_VERSION:-1.37.0} +NPU_PIP_INDEX=${NPU_PIP_INDEX:-https://mirrors.aliyun.com/pypi/simple/} +NPU_CONSTRAINT_FILE=${NPU_CONSTRAINT_FILE:-/tmp/ms_swift_npu_constraints.txt} +NPU_PIP_BLOCK_CUDA_DEPS=${NPU_PIP_BLOCK_CUDA_DEPS:-True} + +print_npu_warning() { + echo "======================================================================" + echo "WARNING: NPU runtime is unavailable, tests will continue on CPU path" + echo "======================================================================" +} + +setup_npu_pip_constraints() { + cat >"$NPU_CONSTRAINT_FILE" <>"$NPU_CONSTRAINT_FILE" <<'EOF' +# NPU CI should not resolve CUDA runtime wheels. If a dependency starts requiring +# these packages, fail in pip's resolver instead of downloading hundreds of MB. +cuda-toolkit<0 +nvidia-cublas<0 +nvidia-cuda-runtime<0 +nvidia-cuda-nvrtc<0 +nvidia-cuda-cupti<0 +nvidia-cudnn<0 +nvidia-cufft<0 +nvidia-curand<0 +nvidia-cusolver<0 +nvidia-cusparse<0 +nvidia-nccl<0 +nvidia-nvjitlink<0 +nvidia-nvtx<0 +nvidia-cublas-cu12<0 +nvidia-cuda-runtime-cu12<0 +nvidia-cuda-nvrtc-cu12<0 +nvidia-cuda-cupti-cu12<0 +nvidia-cudnn-cu12<0 +nvidia-cufft-cu12<0 +nvidia-curand-cu12<0 +nvidia-cusolver-cu12<0 +nvidia-cusparse-cu12<0 +nvidia-cusparselt-cu12<0 +nvidia-nccl-cu12<0 +nvidia-nvjitlink-cu12<0 +nvidia-nvtx-cu12<0 +EOF + fi + export PIP_CONSTRAINT="$NPU_CONSTRAINT_FILE" + export UV_CONSTRAINT="$NPU_CONSTRAINT_FILE" + echo "Using NPU pip constraints: $PIP_CONSTRAINT" + cat "$PIP_CONSTRAINT" +} + +is_npu_runtime_matched() { + python - </dev/null 2>&1; then + npu-smi info || echo "WARNING: npu-smi info failed." + else + echo "WARNING: npu-smi not found." + fi + python - <<'PY' +import importlib.util +import os + +warning = 'WARNING: NPU runtime is unavailable, tests will continue on CPU path' +print(f"ASCEND_RT_VISIBLE_DEVICES={os.environ.get('ASCEND_RT_VISIBLE_DEVICES', '')}") +print(f"CUDA_VISIBLE_DEVICES={os.environ.get('CUDA_VISIBLE_DEVICES', '')}") + +try: + import torch + print(f"torch={torch.__version__}") +except Exception as e: + print(f"WARNING: failed to import torch: {e!r}") + print('=' * 70) + print(warning) + print('=' * 70) + raise SystemExit(0) + +try: + import transformers + print(f"transformers={transformers.__version__}") +except Exception as e: + print(f"WARNING: failed to import transformers: {e!r}") + +if importlib.util.find_spec('torch_npu') is None: + print('WARNING: torch_npu is not installed.') + print('=' * 70) + print(warning) + print('=' * 70) + raise SystemExit(0) + +try: + import torch_npu + print(f"torch_npu={getattr(torch_npu, '__version__', 'unknown')}") +except Exception as e: + print(f"WARNING: failed to import torch_npu: {e!r}") + print('=' * 70) + print(warning) + print('=' * 70) + raise SystemExit(0) + +try: + npu = getattr(torch, 'npu', None) + available = bool(npu is not None and npu.is_available()) + count = npu.device_count() if npu is not None else 0 + print(f"torch.npu.is_available={available}") + print(f"torch.npu.device_count={count}") + if not available: + print('=' * 70) + print(warning) + print('=' * 70) +except Exception as e: + print(f"WARNING: failed to query torch.npu status: {e!r}") + print('=' * 70) + print(warning) + print('=' * 70) +PY + echo "============================================================" +} + +if [ "$MODELSCOPE_SDK_DEBUG" == "True" ]; then + if [ "$SWIFT_CI_USE_NPU" == "True" ]; then + pip install uv -i https://mirrors.aliyun.com/pypi/simple/ + setup_npu_pip_constraints + ensure_npu_runtime + uv pip install -r requirements/tests.txt + git config --global --add safe.directory /ms-swift + git config --global user.email tmp + git config --global user.name tmp.com + # linter test + if [ `git remote -v | grep alibaba | wc -l` -gt 1 ]; then + pre-commit run -c .pre-commit-config_local.yaml --all-files + if [ $? -ne 0 ]; then + echo "linter test failed" + exit -1 + fi + fi + ensure_npu_runtime + uv pip install -r requirements/framework.txt -U "transformers<5.0" "peft<0.19" "modelscope==1.37.0" + ensure_npu_runtime + uv pip install decord einops -U + uv pip uninstall autoawq + uv pip install optimum + uv pip install diffusers + uv pip install math-verify -i "$NPU_PIP_INDEX" + uv pip install ray -i "$NPU_PIP_INDEX" + uv pip install msgspec -i "$NPU_PIP_INDEX" + uv pip install zmq -i "$NPU_PIP_INDEX" + uv pip install . + echo "NPU CI skips auto_gptq because it is a CUDA/GPTQ optional dependency." + uv pip install bitsandbytes deepspeed -U + if [ -f requirements/npu.txt ]; then + uv pip install -r requirements/npu.txt + fi + ensure_npu_runtime + report_npu_runtime + else + pip install -r requirements/tests.txt -i https://mirrors.aliyun.com/pypi/simple/ + git config --global --add safe.directory /ms-swift + git config --global user.email tmp + git config --global user.name tmp.com + # linter test + if [ `git remote -v | grep alibaba | wc -l` -gt 1 ]; then + pre-commit run -c .pre-commit-config_local.yaml --all-files + if [ $? -ne 0 ]; then + echo "linter test failed" + exit -1 + fi + fi + pip install -r requirements/framework.txt -U -i https://mirrors.aliyun.com/pypi/simple/ + pip install decord einops -U -i https://mirrors.aliyun.com/pypi/simple/ + pip uninstall autoawq -y + pip install optimum + pip install diffusers + pip install "transformers<5.0" "peft<0.19" + pip install . + pip install auto_gptq bitsandbytes deepspeed -U -i https://mirrors.aliyun.com/pypi/simple/ + fi +else + echo "Running case in release image, run case directly!" +fi +# remove torch_extensions folder to avoid ci hang. +rm -rf ~/.cache/torch_extensions +if [ $# -eq 0 ]; then + ci_command="python tests/run.py --subprocess" +else + ci_command="$@" +fi +echo "Running case with command: $ci_command" +$ci_command diff --git a/.dev_scripts/dockerci.sh b/.dev_scripts/dockerci.sh new file mode 100644 index 0000000..f42e819 --- /dev/null +++ b/.dev_scripts/dockerci.sh @@ -0,0 +1,89 @@ +#!/bin/bash +MODELSCOPE_CACHE_DIR_IN_CONTAINER=/modelscope_cache +CODE_DIR=$PWD +CODE_DIR_IN_CONTAINER=/ms-swift +MODELSCOPE_SDK_DEBUG=True +echo "$USER" +gpus='0,1 2,3' +cpu_sets='0-15 16-31' +cpu_sets_arr=($cpu_sets) +is_get_file_lock=false +CI_COMMAND=${CI_COMMAND:-bash .dev_scripts/ci_container_test.sh python tests/run.py --parallel 2 --run_config tests/run_config.yaml} +echo "ci command: $CI_COMMAND" +PR_CHANGED_FILES="${PR_CHANGED_FILES:-}" +echo "PR modified files: $PR_CHANGED_FILES" +PR_CHANGED_FILES=${PR_CHANGED_FILES//[ ]/#} +echo "PR_CHANGED_FILES: $PR_CHANGED_FILES" +idx=0 +for gpu in $gpus +do + exec {lock_fd}>"/tmp/gpu$gpu" || exit 1 + flock -n "$lock_fd" || { echo "WARN: gpu $gpu is in use!" >&2; idx=$((idx+1)); continue; } + echo "get gpu lock $gpu" + + CONTAINER_NAME="swift-ci-$idx" + let is_get_file_lock=true + + # pull image if there are update + docker pull ${IMAGE_NAME}:${IMAGE_VERSION} + if [ "$MODELSCOPE_SDK_DEBUG" == "True" ]; then + echo 'debugging' + docker run --rm --name $CONTAINER_NAME --shm-size=16gb \ + --cpuset-cpus=${cpu_sets_arr[$idx]} \ + --gpus='"'"device=$gpu"'"' \ + -v $CODE_DIR:$CODE_DIR_IN_CONTAINER \ + -v $MODELSCOPE_CACHE:$MODELSCOPE_CACHE_DIR_IN_CONTAINER \ + -v $MODELSCOPE_HOME_CACHE/$idx:/root \ + -v /home/admin/pre-commit:/home/admin/pre-commit \ + -e CI_TEST=True \ + -e TEST_LEVEL=$TEST_LEVEL \ + -e MODELSCOPE_CACHE=$MODELSCOPE_CACHE_DIR_IN_CONTAINER \ + -e MODELSCOPE_DOMAIN=$MODELSCOPE_DOMAIN \ + -e MODELSCOPE_SDK_DEBUG=True \ + -e HUB_DATASET_ENDPOINT=$HUB_DATASET_ENDPOINT \ + -e TEST_ACCESS_TOKEN_CITEST=$TEST_ACCESS_TOKEN_CITEST \ + -e TEST_ACCESS_TOKEN_SDKDEV=$TEST_ACCESS_TOKEN_SDKDEV \ + -e TEST_LEVEL=$TEST_LEVEL \ + -e MODELSCOPE_ENVIRONMENT='ci' \ + -e TEST_UPLOAD_MS_TOKEN=$TEST_UPLOAD_MS_TOKEN \ + -e MODEL_TAG_URL=$MODEL_TAG_URL \ + -e MODELSCOPE_API_TOKEN=$MODELSCOPE_API_TOKEN \ + -e PR_CHANGED_FILES=$PR_CHANGED_FILES \ + --workdir=$CODE_DIR_IN_CONTAINER \ + ${IMAGE_NAME}:${IMAGE_VERSION} \ + $CI_COMMAND + else + docker run --rm --name $CONTAINER_NAME --shm-size=16gb \ + --cpuset-cpus=${cpu_sets_arr[$idx]} \ + --gpus='"'"device=$gpu"'"' \ + -v $CODE_DIR:$CODE_DIR_IN_CONTAINER \ + -v $MODELSCOPE_CACHE:$MODELSCOPE_CACHE_DIR_IN_CONTAINER \ + -v $MODELSCOPE_HOME_CACHE/$idx:/root \ + -v /home/admin/pre-commit:/home/admin/pre-commit \ + -e CI_TEST=True \ + -e TEST_LEVEL=$TEST_LEVEL \ + -e MODELSCOPE_CACHE=$MODELSCOPE_CACHE_DIR_IN_CONTAINER \ + -e MODELSCOPE_DOMAIN=$MODELSCOPE_DOMAIN \ + -e HUB_DATASET_ENDPOINT=$HUB_DATASET_ENDPOINT \ + -e TEST_ACCESS_TOKEN_CITEST=$TEST_ACCESS_TOKEN_CITEST \ + -e TEST_ACCESS_TOKEN_SDKDEV=$TEST_ACCESS_TOKEN_SDKDEV \ + -e TEST_LEVEL=$TEST_LEVEL \ + -e MODELSCOPE_ENVIRONMENT='ci' \ + -e TEST_UPLOAD_MS_TOKEN=$TEST_UPLOAD_MS_TOKEN \ + -e MODEL_TAG_URL=$MODEL_TAG_URL \ + -e MODELSCOPE_API_TOKEN=$MODELSCOPE_API_TOKEN \ + -e PR_CHANGED_FILES=$PR_CHANGED_FILES \ + --workdir=$CODE_DIR_IN_CONTAINER \ + ${IMAGE_NAME}:${IMAGE_VERSION} \ + $CI_COMMAND + fi + if [ $? -ne 0 ]; then + echo "Running test case failed, please check the log!" + exit -1 + fi + break +done +if [ "$is_get_file_lock" = false ] ; then + echo 'No free GPU!' + exit 1 +fi diff --git a/.dev_scripts/dockerci_npu.sh b/.dev_scripts/dockerci_npu.sh new file mode 100644 index 0000000..ea76ce0 --- /dev/null +++ b/.dev_scripts/dockerci_npu.sh @@ -0,0 +1,58 @@ +#!/bin/bash +MODELSCOPE_CACHE_DIR=/modelscope_cache +CODE_DIR=$PWD +MODELSCOPE_SDK_DEBUG=True +echo "$USER" +npus='0,1 2,3' +is_get_file_lock=false +CI_COMMAND=${CI_COMMAND:-bash .dev_scripts/ci_container_test.sh python tests/run.py --parallel 2 --run_config tests/run_config.yaml} +echo "ci command: $CI_COMMAND" +PR_CHANGED_FILES="${PR_CHANGED_FILES:-}" +echo "PR modified files: $PR_CHANGED_FILES" +PR_CHANGED_FILES=${PR_CHANGED_FILES//[ ]/#} +echo "PR_CHANGED_FILES: $PR_CHANGED_FILES" +idx=0 +for npu in $npus +do + exec {lock_fd}>"/tmp/npu$npu" || exit 1 + flock -n "$lock_fd" || { echo "WARN: npu $npu is in use!" >&2; idx=$((idx+1)); continue; } + echo "get npu lock $npu" + + let is_get_file_lock=true + + # 设置环境变量 + export CI_TEST=True + export SWIFT_CI_USE_NPU=True + export TEST_LEVEL=$TEST_LEVEL + export MODELSCOPE_CACHE=${MODELSCOPE_CACHE:-$MODELSCOPE_CACHE_DIR} + export MODELSCOPE_DOMAIN=$MODELSCOPE_DOMAIN + export HUB_DATASET_ENDPOINT=$HUB_DATASET_ENDPOINT + export TEST_ACCESS_TOKEN_CITEST=$TEST_ACCESS_TOKEN_CITEST + export TEST_ACCESS_TOKEN_SDKDEV=$TEST_ACCESS_TOKEN_SDKDEV + export MODELSCOPE_ENVIRONMENT='ci' + export TEST_UPLOAD_MS_TOKEN=$TEST_UPLOAD_MS_TOKEN + export MODEL_TAG_URL=$MODEL_TAG_URL + export MODELSCOPE_API_TOKEN=$MODELSCOPE_API_TOKEN + export PR_CHANGED_FILES=$PR_CHANGED_FILES + export ASCEND_RT_VISIBLE_DEVICES=$npu + + if [ "$MODELSCOPE_SDK_DEBUG" == "True" ]; then + export MODELSCOPE_SDK_DEBUG=True + echo 'debugging' + fi + + # 切换到代码目录并执行命令 + cd $CODE_DIR + eval $CI_COMMAND + + if [ $? -ne 0 ]; then + echo "Running test case failed, please check the log!" + exit -1 + fi + break +done + +if [ "$is_get_file_lock" = false ] ; then + echo 'No free NPU!' + exit 1 +fi diff --git a/.github/ISSUE_TEMPLATE/1-bug-report.yml b/.github/ISSUE_TEMPLATE/1-bug-report.yml new file mode 100644 index 0000000..459e91f --- /dev/null +++ b/.github/ISSUE_TEMPLATE/1-bug-report.yml @@ -0,0 +1,49 @@ +name: "🐛 Bug Report" +description: Create a bug report to help us improve ms-swift +labels: ["bug"] + +body: + - type: markdown + attributes: + value: | + Thank you for supporting ms-swift and taking the time to submit this issue. + 感谢你对 ms-swift 的支持和抽出时间提交相关 issue。 + + - type: checkboxes + id: checklist + attributes: + label: Checklist / 检查清单 + options: + - label: I have searched existing issues, and this is a new bug report. / 我已经搜索过现有的 issues，确认这是一个新的 bug report。 + required: true + + + - type: textarea + id: bug-description + validations: + required: true + attributes: + label: Bug Description / Bug 描述 + description: | + Please describe the issue you encountered. It's better to include error screenshots or stack trace information. + 请详细描述你遇到的问题，最好包含报错截图或报错栈信息。 + + + - type: textarea + id: reproduction-steps + validations: + required: true + attributes: + label: How to Reproduce / 如何复现 + description: | + Please provide steps to reproduce the issue, including ms-swift version, runtime environment, and detailed reproduction steps. + 请提供复现问题的步骤，包括 ms-swift 的版本、运行环境、详细的复现步骤等。 + + + - type: textarea + id: additional-information + attributes: + label: Additional Information / 补充信息 + description: | + Please provide any additional information here. + 在这里补充其他相关信息。 diff --git a/.github/ISSUE_TEMPLATE/2-feature-request.yml b/.github/ISSUE_TEMPLATE/2-feature-request.yml new file mode 100644 index 0000000..07e4f5b --- /dev/null +++ b/.github/ISSUE_TEMPLATE/2-feature-request.yml @@ -0,0 +1,37 @@ +name: "🚀 Feature Request" +description: Submit a request for a new feature +labels: ["enhancement"] + +body: + - type: markdown + attributes: + value: | + Thank you for supporting ms-swift and taking the time to submit this issue. + 感谢你对 ms-swift 的支持和抽出时间提交相关 issue。 + + - type: checkboxes + id: checklist + attributes: + label: Checklist / 检查清单 + options: + - label: I have searched existing issues, and this is a new feature request. / 我已经搜索过现有的 issues，确认这是一个新的 Feature Request。 + required: true + + - type: textarea + id: feature-request-description + validations: + required: true + attributes: + label: Feature Request Description / Feature Request 描述 + description: | + Please provide a detailed description of the new feature you would like to see added. + 请详细描述您希望添加的新功能特性。 + + + - type: textarea + id: pull-request + attributes: + label: Pull Request / Pull Request 信息 + description: | + Have you already submitted or plan to submit a Pull Request? Please share your plans. + 你是否已经提交或即将提交 Pull Request？请说明你的计划。 diff --git a/.github/ISSUE_TEMPLATE/3-question-discussion.yml b/.github/ISSUE_TEMPLATE/3-question-discussion.yml new file mode 100644 index 0000000..02baa33 --- /dev/null +++ b/.github/ISSUE_TEMPLATE/3-question-discussion.yml @@ -0,0 +1,28 @@ +name: "🤔 Question & Discussion" +description: Create an issue for questions and discussions +labels: ["question"] + +body: + - type: markdown + attributes: + value: | + Thank you for supporting ms-swift and taking the time to submit this issue. + 感谢你对 ms-swift 的支持和抽出时间提交相关 issue。 + + - type: checkboxes + id: checklist + attributes: + label: Checklist / 检查清单 + options: + - label: I have searched existing issues, and this is a new question or discussion topic. / 我已经搜索过现有的 issues，确认这是一个新的问题与讨论。 + required: true + + - type: textarea + id: question-description + validations: + required: true + attributes: + label: Question Description / 问题描述 + description: | + Please describe the question or topic you would like to discuss. + 请描述你想要讨论的问题或话题。 diff --git a/.github/ISSUE_TEMPLATE/config.yml b/.github/ISSUE_TEMPLATE/config.yml new file mode 100644 index 0000000..3ba13e0 --- /dev/null +++ b/.github/ISSUE_TEMPLATE/config.yml @@ -0,0 +1 @@ +blank_issues_enabled: false diff --git a/.github/PULL_REQUEST_TEMPLATE.md b/.github/PULL_REQUEST_TEMPLATE.md new file mode 100644 index 0000000..a09bfad --- /dev/null +++ b/.github/PULL_REQUEST_TEMPLATE.md @@ -0,0 +1,13 @@ +# PR type +- [ ] Bug Fix +- [ ] New Feature +- [ ] Document Updates +- [ ] More Models or Datasets Support + +# PR information + +Write the detail information belongs to this PR. + +## Experiment results + +Paste your experiment result here(if needed). diff --git a/.github/SECURITY.md b/.github/SECURITY.md new file mode 100644 index 0000000..d549cbe --- /dev/null +++ b/.github/SECURITY.md @@ -0,0 +1,3 @@ +# Reporting Security Issues + +Usually security issues of a deep learning project come from non-standard 3rd packages or continuous running services. If you are suffering from security issues from our project, please consider reporting to us. We appreciate your efforts to responsibly disclose your findings, and will make every effort to acknowledge your contributions. diff --git a/.github/workflows/citest.yaml b/.github/workflows/citest.yaml new file mode 100644 index 0000000..101e9f3 --- /dev/null +++ b/.github/workflows/citest.yaml @@ -0,0 +1,77 @@ +name: citest + +on: + push: + branches: + - master + - "release/**" + paths-ignore: + - "setup.*" + - "requirements.txt" + - "requirements/**" + - "docs/**" + - "tools/**" + - ".dev_scripts/**" + - "README.md" + - "README_*.md" + - "NOTICE" + - ".github/workflows/lint.yaml" + - ".github/workflows/publish.yaml" + + pull_request: + paths-ignore: + - "setup.*" + - "requirements.txt" + - "requirements/**" + - "docs/**" + - "tools/**" + - ".dev_scripts/**" + - "README.md" + - "README_*.md" + - "NOTICE" + - ".github/workflows/lint.yaml" + - ".github/workflows/publish.yaml" + +concurrency: + group: ${{ github.workflow }}-${{ github.ref }} + cancel-in-progress: true + +jobs: + unittest: + # The type of runner that the job will run on + runs-on: [self-hosted] + timeout-minutes: 240 + steps: + - name: ResetFileMode + shell: bash + run: | + # reset filemode to allow action runner to delete files + # generated by root in docker + set -e + source ~/.bashrc + sudo chown -R $USER:$USER $GITHUB_WORKSPACE + + - name: Checkout + uses: actions/checkout@v3 + env: + GIT_CONFIG_PARAMETERS: "'core.hooksPath='" + with: + lfs: 'true' + submodules: 'false' + fetch-depth: ${{ github.event_name == 'pull_request' && 2 || 0 }} + - name: Get changed files + id: changed-files + run: | + if ${{ github.event_name == 'pull_request' }}; then + echo "PR_CHANGED_FILES=$(git diff --name-only -r HEAD^1 HEAD | xargs)" >> $GITHUB_ENV + else + echo "PR_CHANGED_FILES=$(git diff --name-only ${{ github.event.before }} ${{ github.event.after }} | xargs)" >> $GITHUB_ENV + fi + - name: Checkout LFS objects + run: git lfs checkout + - name: Run unittest + shell: bash + run: | + set -e + source /mnt/modelscope/ci_env.sh + bash .dev_scripts/dockerci.sh diff --git a/.github/workflows/citest_npu.yaml b/.github/workflows/citest_npu.yaml new file mode 100644 index 0000000..0411080 --- /dev/null +++ b/.github/workflows/citest_npu.yaml @@ -0,0 +1,98 @@ +name: citest-npu + +on: + push: + branches: + - master + - "release/**" + paths-ignore: + - "setup.*" + - "requirements.txt" + - "requirements/**" + - "docs/**" + - "tools/**" + - ".dev_scripts/**" + - "README.md" + - "README_*.md" + - "NOTICE" + - ".github/workflows/lint.yaml" + - ".github/workflows/publish.yaml" + + pull_request: + paths-ignore: + - "setup.*" + - "requirements.txt" + - "requirements/**" + - "docs/**" + - "tools/**" + - ".dev_scripts/**" + - "README.md" + - "README_*.md" + - "NOTICE" + - ".github/workflows/lint.yaml" + - ".github/workflows/publish.yaml" + +concurrency: + group: ${{ github.workflow }}-${{ github.ref }} + cancel-in-progress: true + +jobs: + unittest: + # The type of runner that the job will run on + runs-on: [linux-aarch64-a2-1] + timeout-minutes: 240 + container: + image: 'ascendai/cann:8.3.rc2-910b-ubuntu22.04-py3.11' + options: >- + --privileged + --device=/dev/davinci0 + --device=/dev/davinci1 + --device=/dev/davinci2 + --device=/dev/davinci3 + --device=/dev/davinci_manager + --device=/dev/devmm_svm + --device=/dev/hisi_hdc + --volume=/usr/local/Ascend/driver:/usr/local/Ascend/driver:ro + --volume=/usr/local/sbin/npu-smi:/usr/local/sbin/npu-smi:ro + --volume=/etc/ascend_install.info:/etc/ascend_install.info:ro + env: + UV_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple" + UV_EXTRA_INDEX_URL: "https://repo.huaweicloud.com/ascend/repos/pypi" + UV_INDEX_STRATEGY: "unsafe-best-match" + UV_INSECURE_HOST: "cache-service.nginx-pypi-cache.svc.cluster.local" + UV_HTTP_TIMEOUT: 120 + UV_NO_CACHE: 1 + UV_SYSTEM_PYTHON: 1 + PIP_INDEX_URL: "http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple" + PIP_TRUSTED_HOST: "cache-service.nginx-pypi-cache.svc.cluster.local" + steps: + - name: Config mirrors + run: | + sed -Ei 's@(ports|archive).ubuntu.com@cache-service.nginx-pypi-cache.svc.cluster.local:8081@g' /etc/apt/sources.list + pip config set global.index-url http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple + pip config set global.trusted-host cache-service.nginx-pypi-cache.svc.cluster.local + pip install uv + + - name: Checkout + uses: actions/checkout@v3 + with: + fetch-depth: ${{ github.event_name == 'pull_request' && 2 || 0 }} + - name: Get changed files + id: changed-files + run: | + if ${{ github.event_name == 'pull_request' }}; then + echo "PR_CHANGED_FILES=$(git diff --name-only -r HEAD^1 HEAD | xargs)" >> $GITHUB_ENV + else + echo "PR_CHANGED_FILES=$(git diff --name-only ${{ github.event.before }} ${{ github.event.after }} | xargs)" >> $GITHUB_ENV + fi + - name: Run unittest + shell: bash + run: | + set -e + export IMAGE_NAME=ascendai/cann + export IMAGE_VERSION=8.3.rc2-910b-ubuntu22.04-py3.11 + export TEST_LEVEL=0 + mkdir -p ~/.cache + export MODELSCOPE_CACHE=~/.cache + export CI_COMMAND='bash .dev_scripts/ci_container_test.sh python tests/run.py --parallel 2 --subprocess --run_config tests/run_config.yaml' + bash .dev_scripts/dockerci_npu.sh diff --git a/.github/workflows/close_tale_issue.yaml b/.github/workflows/close_tale_issue.yaml new file mode 100644 index 0000000..46a713f --- /dev/null +++ b/.github/workflows/close_tale_issue.yaml @@ -0,0 +1,20 @@ +name: Close Stale Issues +on: + schedule: + - cron: '0 0 * * *' + workflow_dispatch: + +jobs: + close-stale: + runs-on: ubuntu-latest + steps: + - name: Close stale issues + uses: actions/stale@v8 + with: + repo-token: ${{ secrets.GITHUB_TOKEN }} + days-before-stale: 90 + days-before-close: 7 + stale-issue-message: 'This issue has been inactive for over 3 months and will be automatically closed in 7 days. If this issue is still relevant, please reply to this message.' + close-issue-message: 'This issue has been automatically closed due to inactivity. If needed, it can be reopened.' + stale-issue-label: 'stale' + exempt-all-issue-labels: true diff --git a/.github/workflows/lint.yaml b/.github/workflows/lint.yaml new file mode 100644 index 0000000..6ff8451 --- /dev/null +++ b/.github/workflows/lint.yaml @@ -0,0 +1,22 @@ +name: Lint test + +on: [push, pull_request] + +concurrency: + group: ${{ github.workflow }}-${{ github.ref }} + cancel-in-progress: true + +jobs: + lint: + runs-on: ubuntu-latest + steps: + - uses: actions/checkout@v2 + - name: Set up Python 3.10 + uses: actions/setup-python@v2 + with: + python-version: '3.10' + - name: Install pre-commit hook + run: | + pip install pre-commit + - name: Linting + run: pre-commit run --all-files diff --git a/.github/workflows/publish.yaml b/.github/workflows/publish.yaml new file mode 100644 index 0000000..7a0c747 --- /dev/null +++ b/.github/workflows/publish.yaml @@ -0,0 +1,29 @@ +name: release + +on: + push: + tags: + - 'v**' + +concurrency: + group: ${{ github.workflow }}-${{ github.ref }}-publish + cancel-in-progress: true + +jobs: + build-n-publish: + runs-on: ubuntu-22.04 + #if: startsWith(github.event.ref, 'refs/tags') + steps: + - uses: actions/checkout@v2 + - name: Set up Python 3.10 + uses: actions/setup-python@v2 + with: + python-version: '3.10' + - name: Install wheel + run: pip install wheel packaging setuptools==69.5.1 + - name: Build ModelScope Swift + run: python setup.py sdist bdist_wheel + - name: Publish package to PyPI + run: | + pip install twine + twine upload dist/* --skip-existing -u __token__ -p ${{ secrets.PYPI_API_TOKEN }} diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000..f1fa258 --- /dev/null +++ b/.gitignore @@ -0,0 +1,156 @@ +# Byte-compiled / optimized / DLL files +tmp +*.ttf +__pycache__/ +*.py[cod] +*$py.class +test.py +# C extensions +*.so + +# Distribution / packaging +.Python +build/ +develop-eggs/ +dist/ +downloads/ +eggs/ +.eggs/ +lib/ +lib64/ +parts/ +sdist/ +var/ +wheels/ +*.egg-info/ +.installed.cfg +*.egg +/package +/temp +MANIFEST + +# PyInstaller +# Usually these files are written by a python script from a template +# before PyInstaller builds the exe, so as to inject date/other infos into it. +*.manifest +*.spec + +# Installer logs +pip-log.txt +pip-delete-this-directory.txt + +# Unit test / coverage reports +htmlcov/ +.tox/ +.coverage +.coverage.* +.cache +nosetests.xml +coverage.xml +*.cover +.hypothesis/ +.pytest_cache/ + +# Translations +*.mo +*.pot + +# Django stuff: +*.log +local_settings.py +db.sqlite3 + +# Flask stuff: +instance/ +.webassets-cache + +# Scrapy stuff: +.scrapy + +# Sphinx documentation +docs/_build/ + +# PyBuilder +target/ + +# Jupyter Notebook +.ipynb_checkpoints + +# pyenv +.python-version + +# celery beat schedule file +celerybeat-schedule + +# SageMath parsed files +*.sage.py + +# Environments +.env +.venv +env/ +venv/ +ENV/ +env.bak/ +venv.bak/ + +# Spyder project settings +.spyderproject +.spyproject + +# Rope project settings +.ropeproject + +# mkdocs documentation +/site + +# mypy +.mypy_cache/ + +.vscode +.idea +.run + +# custom +*.pkl +*.pkl.json +*.log.json +*.whl +*.tar.gz +*.swp +*.log +*.tar.gz +source.sh +tensorboard.sh +.DS_Store +replace.sh +result.png +result.jpg +result.mp4 +output/ +outputs/ +wandb/ +swanlog/ +*.out +benchmarks/ +eval_output/ +eval_outputs/ +vlmeval/ +my_model/ +/data +result/ +images +/custom/ +megatron_output/ +/*-mcore/ +/*-hf/ +/*_cached_dataset/ +/sample_output/ +.qoder/ + +# Pytorch +*.pth +*.pt + +# ast template +ast_index_file.py diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml new file mode 100644 index 0000000..34e2f96 --- /dev/null +++ b/.pre-commit-config.yaml @@ -0,0 +1,24 @@ +repos: + - repo: https://github.com/pycqa/flake8.git + rev: 7.3.0 + hooks: + - id: flake8 + - repo: https://github.com/PyCQA/isort.git + rev: 8.0.1 + hooks: + - id: isort + - repo: https://github.com/google/yapf.git + rev: v0.43.0 + hooks: + - id: yapf + - repo: https://github.com/pre-commit/pre-commit-hooks.git + rev: v6.0.0 + hooks: + - id: trailing-whitespace + - id: check-yaml + - id: end-of-file-fixer + - id: requirements-txt-fixer + - id: double-quote-string-fixer + - id: check-merge-conflict + - id: mixed-line-ending + args: ["--fix=lf"] diff --git a/CODE_OF_CONDUCT.md b/CODE_OF_CONDUCT.md new file mode 100644 index 0000000..804f949 --- /dev/null +++ b/CODE_OF_CONDUCT.md @@ -0,0 +1,132 @@ +# Contributor Covenant Code of Conduct + +## Our Pledge + +We as members, contributors, and leaders pledge to make participation in our +community a harassment-free experience for everyone, regardless of age, body +size, visible or invisible disability, ethnicity, sex characteristics, gender +identity and expression, level of experience, education, socio-economic status, +nationality, personal appearance, race, caste, color, religion, or sexual +identity and orientation. + +We pledge to act and interact in ways that contribute to an open, welcoming, +diverse, inclusive, and healthy community. + +## Our Standards + +Examples of behavior that contributes to a positive environment for our +community include: + +* Demonstrating empathy and kindness toward other people +* Being respectful of differing opinions, viewpoints, and experiences +* Giving and gracefully accepting constructive feedback +* Accepting responsibility and apologizing to those affected by our mistakes, + and learning from the experience +* Focusing on what is best not just for us as individuals, but for the overall + community + +Examples of unacceptable behavior include: + +* The use of sexualized language or imagery, and sexual attention or advances of + any kind +* Trolling, insulting or derogatory comments, and personal or political attacks +* Public or private harassment +* Publishing others' private information, such as a physical or email address, + without their explicit permission +* Other conduct which could reasonably be considered inappropriate in a + professional setting + +## Enforcement Responsibilities + +Community leaders are responsible for clarifying and enforcing our standards of +acceptable behavior and will take appropriate and fair corrective action in +response to any behavior that they deem inappropriate, threatening, offensive, +or harmful. + +Community leaders have the right and responsibility to remove, edit, or reject +comments, commits, code, wiki edits, issues, and other contributions that are +not aligned to this Code of Conduct, and will communicate reasons for moderation +decisions when appropriate. + +## Scope + +This Code of Conduct applies within all community spaces, and also applies when +an individual is officially representing the community in public spaces. +Examples of representing our community include using an official e-mail address, +posting via an official social media account, or acting as an appointed +representative at an online or offline event. + +## Enforcement + +Instances of abusive, harassing, or otherwise unacceptable behavior may be +reported to the community leaders responsible for enforcement at +contact@modelscope.cn. +All complaints will be reviewed and investigated promptly and fairly. + +All community leaders are obligated to respect the privacy and security of the +reporter of any incident. + +## Enforcement Guidelines + +Community leaders will follow these Community Impact Guidelines in determining +the consequences for any action they deem in violation of this Code of Conduct: + +### 1. Correction + +**Community Impact**: Use of inappropriate language or other behavior deemed +unprofessional or unwelcome in the community. + +**Consequence**: A private, written warning from community leaders, providing +clarity around the nature of the violation and an explanation of why the +behavior was inappropriate. A public apology may be requested. + +### 2. Warning + +**Community Impact**: A violation through a single incident or series of +actions. + +**Consequence**: A warning with consequences for continued behavior. No +interaction with the people involved, including unsolicited interaction with +those enforcing the Code of Conduct, for a specified period of time. This +includes avoiding interactions in community spaces as well as external channels +like social media. Violating these terms may lead to a temporary or permanent +ban. + +### 3. Temporary Ban + +**Community Impact**: A serious violation of community standards, including +sustained inappropriate behavior. + +**Consequence**: A temporary ban from any sort of interaction or public +communication with the community for a specified period of time. No public or +private interaction with the people involved, including unsolicited interaction +with those enforcing the Code of Conduct, is allowed during this period. +Violating these terms may lead to a permanent ban. + +### 4. Permanent Ban + +**Community Impact**: Demonstrating a pattern of violation of community +standards, including sustained inappropriate behavior, harassment of an +individual, or aggression toward or disparagement of classes of individuals. + +**Consequence**: A permanent ban from any sort of public interaction within the +community. + +## Attribution + +This Code of Conduct is adapted from the [Contributor Covenant][homepage], +version 2.1, available at +[https://www.contributor-covenant.org/version/2/1/code_of_conduct.html][v2.1]. + +Community Impact Guidelines were inspired by +[Mozilla's code of conduct enforcement ladder][Mozilla CoC]. + +For answers to common questions about this code of conduct, see the FAQ at +[https://www.contributor-covenant.org/faq][FAQ]. Translations are available at +[https://www.contributor-covenant.org/translations][translations]. + +[homepage]: https://www.contributor-covenant.org +[v2.1]: https://www.contributor-covenant.org/version/2/1/code_of_conduct.html +[Mozilla CoC]: https://github.com/mozilla/diversity +[FAQ]: https://www.contributor-covenant.org/faq +[translations]: https://www.contributor-covenant.org/translations diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md new file mode 100644 index 0000000..05ef837 --- /dev/null +++ b/CONTRIBUTING.md @@ -0,0 +1,59 @@ +# Contributor Guide + +_Welcome to offer PRs, bug reports, documentation supplements or other types of contributions to SWIFT!_ + +## Table of Contents +- [Code of Conduct](#-code-of-conduct) +- [Contribution Process](#-contribution-process) +- [Hardware support](#-Hardware-support) + +## 📖 Code of Conduct +Please refer to our [Code of Conduct documentation](./CODE_OF_CONDUCT.md). + +## 🔁 Contribution Process +### What We Need +- New Technologies and New Models: SWIFT needs to support more open-source models and datasets, or new technologies that we have not paid attention to. If you are interested please submit a PR to us. +- Technical Propagation: If you are interested in technical propagation, you are welcome to help us write tutorials, documents or videos on any website, and send us the link. +- Community Contribution: You can write technical articles related to SWIFT, and submit them to us. After review and approval, we will publish them on the official ModelScope accounts (Zhihu, WeChat, etc.), with your name assigned. + +### Incentives +- we will issue electronic certificates to contributors on behalf of the ModelScope community, to encourage your selfless contributions. +- We will offer small souvenirs related to the ModelScope Community. +- We will provide free A10 computing power during the development period. For more details, please refer to [Hardware-support](#-Hardware-support) section. + +### Submitting PR (Pull Requests) + +Any feature development is carried out in the form of Fork and then PR on GitHub. +1. Fork: Go to the [ms-swift](https://github.com/modelscope/ms-swift) page and click the **Fork button**. After completion, a SWIFT code repository will be cloned under your personal organization. +2. Clone: Clone the code repository generated in the first step to your local machine and **create a new branch** for development. During development, please click the **Sync Fork button** in time to synchronize with the `main` branch to prevent code expiration and conflicts. +3. Submit PR: After development and testing, push the code to the remote branch. On GitHub, go to the **Pull Requests page**, create a new PR, select your code branch as the source branch, and the `modelscope/ms-swift:main` branch as the target branch. + +4. Write Description: It is necessary to provide a good feature description in the PR, so that the reviewers know the content of your modification. +5. Review: We hope that the code to be merged is concise and efficient, so we may raise some questions and discuss them. Please note that any issues raised in the review are aimed at the code itself, not at you personally. Once all issues are discussed and resolved, your code will be approved. + +### Code Standards and Development Approach +SWIFT has conventional variable naming conventions and development approaches. Please follow these approaches as much as possible during development. +1. Variable names are separated by underscores, and class names are named with the first letter of each word capitalized. +2. All Python indentation uses four spaces instead of a tab. +3. Choose well-known open-source libraries, avoid using closed-source libraries or unstable open-source libraries, and avoid repeating the existing code. + +After the PR is submitted, SWIFT will perform two types of tests: +- Code Lint Test: A static code compliance check test. please make sure that you have performed code lint locally in advance. +```shell +pip install pre-commit # In the swift folder +pre-commit run --all-files # Fix the errors reported by pre-commit until all checks are successful +``` +- CI Tests: Smoke tests and unit tests, please refer to the next section. + +### Running CI Tests +Before submitting the PR, please ensure that your development code is protected by test cases, such as smoke tests for new features, or unit tests for various edge cases. Reviewers will also pay attention to this during code review. At the same time, there will be dedicated services running CI Tests, running all test cases, and the code can only be merged after the test cases pass. + +## ✅ Hardware support + +ModelScope provides developers with free A10 GPU computing resources. For more details, please refer to the [ModelScope Notebook](https://modelscope.cn/my/mynotebook). + +ms-swift Training WeChat Group: + +

+ +

diff --git a/CONTRIBUTING_CN.md b/CONTRIBUTING_CN.md new file mode 100644 index 0000000..982c2f7 --- /dev/null +++ b/CONTRIBUTING_CN.md @@ -0,0 +1,75 @@ +# 贡献者指引 + +*欢迎帮SWIFT提供Feature PR、Bug反馈、文档补充或其他类型的贡献！* + +## 目录 + +- [代码规约](#-代码规约) +- [贡献流程](#-贡献流程) +- [资源支持](#-资源支持) + +## 📖 代码规约 + +请查看我们的[代码规约文档](./CODE_OF_CONDUCT.md). + +## 🔁 贡献流程 + +### 我们需要什么 +- 新技术和新模型：SWIFT需要支持更多的开源模型和数据集，或我们没有关注到的新技术，如果您对此有兴趣，可以提交PR给我们。 +- 技术布道：如果您对技术布道有兴趣，欢迎在任何网站上帮我们撰写教程文档或视频等，并将链接发给我们。 +- 社区供稿：您可以撰写和SWIFT有关的技术文章，并供稿给我们，我们审核通过后会在魔搭官方账号（知乎、公众号等）上进行发布，并属上您的名字。 + +### 激励 + +- 我们会以魔搭社区的身份给贡献者颁发电子证书，以鼓励您的无私贡献。 +- 我们会赠送相关魔搭社区相关周边小礼品。 +- 我们会赠送开发期间的免费A10算力，具体可以查看[资源支持](#-资源支持)章节。 + +### 提交PR（Pull Requests） + +任何feature开发都在github上以先Fork后PR的形式进行。 + +1. Fork：进入[ms-swift](https://github.com/modelscope/ms-swift)页面后，点击**Fork按钮**执行。完成后会在您的个人组织下克隆出一个SWIFT代码库 + +2. Clone：将第一步产生的代码库clone到本地并**拉新分支**进行开发，开发中请及时点击**Sync Fork按钮**同步`main`分支，防止代码过期并冲突 + +3. 提交PR：开发、测试完成后将代码推送到远程分支。在github上点击**Pull Requests页面**，新建一个PR，源分支选择您提交的代码分支，目标分支选择`modelscope/ms-swift:main`分支 + +4. 撰写描述：在PR中填写良好的feature描述是必要的，让Reviewers知道您的修改内容 + +5. Review：我们希望合入的代码简洁高效，因此可能会提出一些问题并讨论。请注意，任何review中提出的问题是针对代码本身，而非您个人。在所有问题讨论通过后，您的代码会被通过 + +### 代码规范和开发方式 + +SWIFT有约定俗成的变量命名方式和开发方式。在开发中请尽量遵循这些方式。 + +1. 变量命名以下划线分割，类名以所有单词首字母大写方式命名 +2. 所有的python缩进都是四个空格取代一个tab +3. 选用知名的开源库，避免使用闭源库或不稳定的开源库，避免重复造轮子 + +SWIFT在PR提交后会进行两类测试： + +- Code Lint测试 对代码进行静态规范走查的测试，为保证改测试通过，请保证本地预先进行了Code lint。方法是： + + ```shell + pip install pre-commit + # 在swift文件夹内 + pre-commit run --all-files + # 对pre-commit报的错误进行修改，直到所有的检查都是成功状态 + ``` + +- CI Tests 冒烟测试和单元测试，请查看下一章节 + +### Running CI Tests + +在提交PR前，请保证您的开发代码已经受到了测试用例的保护。例如，对新功能的冒烟测试，或者各种边缘case的单元测试等。在代码review时Reviewers也会关注这一点。同时，也会有服务专门运行CI Tests，运行所有的测试用例，测试用例通过后代码才可以合并。 + +## ✅ 资源支持 + +魔搭为开发者提供了免费的A10 GPU算力支持，具体参考[魔搭 Notebook](https://modelscope.cn/my/mynotebook)。 + +ms-swift训练微信群： + +

+ +

diff --git a/LICENSE b/LICENSE new file mode 100644 index 0000000..261eeb9 --- /dev/null +++ b/LICENSE @@ -0,0 +1,201 @@ + Apache License + Version 2.0, January 2004 + http://www.apache.org/licenses/ + + TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION + + 1. Definitions. + + "License" shall mean the terms and conditions for use, reproduction, + and distribution as defined by Sections 1 through 9 of this document. + + "Licensor" shall mean the copyright owner or entity authorized by + the copyright owner that is granting the License. + + "Legal Entity" shall mean the union of the acting entity and all + other entities that control, are controlled by, or are under common + control with that entity. For the purposes of this definition, + "control" means (i) the power, direct or indirect, to cause the + direction or management of such entity, whether by contract or + otherwise, or (ii) ownership of fifty percent (50%) or more of the + outstanding shares, or (iii) beneficial ownership of such entity. + + "You" (or "Your") shall mean an individual or Legal Entity + exercising permissions granted by this License. + + "Source" form shall mean the preferred form for making modifications, + including but not limited to software source code, documentation + source, and configuration files. + + "Object" form shall mean any form resulting from mechanical + transformation or translation of a Source form, including but + not limited to compiled object code, generated documentation, + and conversions to other media types. + + "Work" shall mean the work of authorship, whether in Source or + Object form, made available under the License, as indicated by a + copyright notice that is included in or attached to the work + (an example is provided in the Appendix below). + + "Derivative Works" shall mean any work, whether in Source or Object + form, that is based on (or derived from) the Work and for which the + editorial revisions, annotations, elaborations, or other modifications + represent, as a whole, an original work of authorship. For the purposes + of this License, Derivative Works shall not include works that remain + separable from, or merely link (or bind by name) to the interfaces of, + the Work and Derivative Works thereof. + + "Contribution" shall mean any work of authorship, including + the original version of the Work and any modifications or additions + to that Work or Derivative Works thereof, that is intentionally + submitted to Licensor for inclusion in the Work by the copyright owner + or by an individual or Legal Entity authorized to submit on behalf of + the copyright owner. For the purposes of this definition, "submitted" + means any form of electronic, verbal, or written communication sent + to the Licensor or its representatives, including but not limited to + communication on electronic mailing lists, source code control systems, + and issue tracking systems that are managed by, or on behalf of, the + Licensor for the purpose of discussing and improving the Work, but + excluding communication that is conspicuously marked or otherwise + designated in writing by the copyright owner as "Not a Contribution." + + "Contributor" shall mean Licensor and any individual or Legal Entity + on behalf of whom a Contribution has been received by Licensor and + subsequently incorporated within the Work. + + 2. Grant of Copyright License. Subject to the terms and conditions of + this License, each Contributor hereby grants to You a perpetual, + worldwide, non-exclusive, no-charge, royalty-free, irrevocable + copyright license to reproduce, prepare Derivative Works of, + publicly display, publicly perform, sublicense, and distribute the + Work and such Derivative Works in Source or Object form. + + 3. Grant of Patent License. Subject to the terms and conditions of + this License, each Contributor hereby grants to You a perpetual, + worldwide, non-exclusive, no-charge, royalty-free, irrevocable + (except as stated in this section) patent license to make, have made, + use, offer to sell, sell, import, and otherwise transfer the Work, + where such license applies only to those patent claims licensable + by such Contributor that are necessarily infringed by their + Contribution(s) alone or by combination of their Contribution(s) + with the Work to which such Contribution(s) was submitted. If You + institute patent litigation against any entity (including a + cross-claim or counterclaim in a lawsuit) alleging that the Work + or a Contribution incorporated within the Work constitutes direct + or contributory patent infringement, then any patent licenses + granted to You under this License for that Work shall terminate + as of the date such litigation is filed. + + 4. Redistribution. You may reproduce and distribute copies of the + Work or Derivative Works thereof in any medium, with or without + modifications, and in Source or Object form, provided that You + meet the following conditions: + + (a) You must give any other recipients of the Work or + Derivative Works a copy of this License; and + + (b) You must cause any modified files to carry prominent notices + stating that You changed the files; and + + (c) You must retain, in the Source form of any Derivative Works + that You distribute, all copyright, patent, trademark, and + attribution notices from the Source form of the Work, + excluding those notices that do not pertain to any part of + the Derivative Works; and + + (d) If the Work includes a "NOTICE" text file as part of its + distribution, then any Derivative Works that You distribute must + include a readable copy of the attribution notices contained + within such NOTICE file, excluding those notices that do not + pertain to any part of the Derivative Works, in at least one + of the following places: within a NOTICE text file distributed + as part of the Derivative Works; within the Source form or + documentation, if provided along with the Derivative Works; or, + within a display generated by the Derivative Works, if and + wherever such third-party notices normally appear. The contents + of the NOTICE file are for informational purposes only and + do not modify the License. You may add Your own attribution + notices within Derivative Works that You distribute, alongside + or as an addendum to the NOTICE text from the Work, provided + that such additional attribution notices cannot be construed + as modifying the License. + + You may add Your own copyright statement to Your modifications and + may provide additional or different license terms and conditions + for use, reproduction, or distribution of Your modifications, or + for any such Derivative Works as a whole, provided Your use, + reproduction, and distribution of the Work otherwise complies with + the conditions stated in this License. + + 5. Submission of Contributions. Unless You explicitly state otherwise, + any Contribution intentionally submitted for inclusion in the Work + by You to the Licensor shall be under the terms and conditions of + this License, without any additional terms or conditions. + Notwithstanding the above, nothing herein shall supersede or modify + the terms of any separate license agreement you may have executed + with Licensor regarding such Contributions. + + 6. Trademarks. This License does not grant permission to use the trade + names, trademarks, service marks, or product names of the Licensor, + except as required for reasonable and customary use in describing the + origin of the Work and reproducing the content of the NOTICE file. + + 7. Disclaimer of Warranty. Unless required by applicable law or + agreed to in writing, Licensor provides the Work (and each + Contributor provides its Contributions) on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or + implied, including, without limitation, any warranties or conditions + of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A + PARTICULAR PURPOSE. You are solely responsible for determining the + appropriateness of using or redistributing the Work and assume any + risks associated with Your exercise of permissions under this License. + + 8. Limitation of Liability. In no event and under no legal theory, + whether in tort (including negligence), contract, or otherwise, + unless required by applicable law (such as deliberate and grossly + negligent acts) or agreed to in writing, shall any Contributor be + liable to You for damages, including any direct, indirect, special, + incidental, or consequential damages of any character arising as a + result of this License or out of the use or inability to use the + Work (including but not limited to damages for loss of goodwill, + work stoppage, computer failure or malfunction, or any and all + other commercial damages or losses), even if such Contributor + has been advised of the possibility of such damages. + + 9. Accepting Warranty or Additional Liability. While redistributing + the Work or Derivative Works thereof, You may choose to offer, + and charge a fee for, acceptance of support, warranty, indemnity, + or other liability obligations and/or rights consistent with this + License. However, in accepting such obligations, You may act only + on Your own behalf and on Your sole responsibility, not on behalf + of any other Contributor, and only if You agree to indemnify, + defend, and hold each Contributor harmless for any liability + incurred by, or claims asserted against, such Contributor by reason + of your accepting any such warranty or additional liability. + + END OF TERMS AND CONDITIONS + + APPENDIX: How to apply the Apache License to your work. + + To apply the Apache License to your work, attach the following + boilerplate notice, with the fields enclosed by brackets "[]" + replaced with your own identifying information. (Don't include + the brackets!) The text should be enclosed in the appropriate + comment syntax for the file format. We also recommend that a + file or class name and description of purpose be included on the + same "printed page" as the copyright notice for easier + identification within third-party archives. + + Copyright [yyyy] [name of copyright owner] + + Licensed under the Apache License, Version 2.0 (the "License"); + you may not use this file except in compliance with the License. + You may obtain a copy of the License at + + http://www.apache.org/licenses/LICENSE-2.0 + + Unless required by applicable law or agreed to in writing, software + distributed under the License is distributed on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + See the License for the specific language governing permissions and + limitations under the License. diff --git a/MANIFEST.in b/MANIFEST.in new file mode 100644 index 0000000..4f91846 --- /dev/null +++ b/MANIFEST.in @@ -0,0 +1 @@ +recursive-include requirements *.txt diff --git a/Makefile b/Makefile new file mode 100644 index 0000000..9653219 --- /dev/null +++ b/Makefile @@ -0,0 +1,25 @@ +WHL_BUILD_DIR :=package +DOC_BUILD_DIR :=docs/build/ + +# default rule +default: whl docs + +.PHONY: docs +docs: + bash .dev_scripts/build_docs.sh + +.PHONY: linter +linter: + bash .dev_scripts/linter.sh + +.PHONY: test +test: + bash .dev_scripts/citest.sh + +.PHONY: whl +whl: + python setup.py sdist bdist_wheel + +.PHONY: clean +clean: + rm -rf $(WHL_BUILD_DIR) $(DOC_BUILD_DIR) diff --git a/README.md b/README.md new file mode 100644 index 0000000..16f7487 --- /dev/null +++ b/README.md @@ -0,0 +1,508 @@ +# SWIFT (Scalable lightWeight Infrastructure for Fine-Tuning) + +

+
+ +
+

+

+ModelScope Community Website +
+ 中文   ｜   English   +

+ +

+ + + + + + + +

+ +

+modelscope/ms-swift | Trendshift +

+ +

+ Paper   ｜ English Documentation   ｜   中文文档   +

+ +## 📖 Table of Contents +- [Groups](#-Groups) +- [Introduction](#-introduction) +- [News](#-news) +- [Installation](#%EF%B8%8F-installation) +- [Quick Start](#-quick-Start) +- [Usage](#-Usage) +- [License](#-License) +- [Citation](#-citation) + + +## ☎ Groups + +You can contact us and communicate with us by adding our group: + + +[Discord Group](https://discord.gg/yeN59wxjwe) | WeChat Group +:-------------------------:|:-------------------------: + | + + +## 📝 Introduction +🍲 **ms-swift** is a large model and multimodal large model fine-tuning and deployment framework provided by the ModelScope community. It now supports training (pre-training, fine-tuning, human alignment), inference, evaluation, quantization, and deployment for 600+ text-only large models and 400+ multimodal large models. Large models include: Qwen3, Qwen3.5, InternLM3, GLM4.5, Mistral, DeepSeek-R1, Llama4, etc. Multimodal large models include: Qwen3-VL, Qwen3-Omni, Llava, InternVL3.5, MiniCPM-V-4, Ovis2.5, GLM4.5-V, DeepSeek-VL2, etc. + +🍔 In addition, ms-swift integrates the latest training technologies, including Megatron parallelism techniques such as TP, PP, CP, EP to accelerate training, as well as numerous GRPO algorithm family reinforcement learning algorithms including: GRPO, DAPO, GSPO, SAPO, CISPO, RLOO, Reinforce++, etc. to enhance model intelligence. ms-swift supports a wide range of training tasks, including preference learning algorithms such as DPO, KTO, RM, CPO, SimPO, ORPO, as well as Embedding, Reranker, and sequence classification tasks. ms-swift provides full-pipeline support for large model training, including acceleration for inference, evaluation, and deployment modules using vLLM, SGLang, and LMDeploy, as well as model quantization using GPTQ, AWQ, BNB, and FP8 technologies. + +**Why Choose ms-swift?** + +- 🍎 **Model Types**: Supports **600+ text-only large models**, **400+ multimodal large models**, and All-to-All full modality models from training to deployment full pipeline, with Day-0 support for popular models. +- **Dataset Types**: Built-in 150+ datasets for pre-training, fine-tuning, human alignment, multimodal and various other tasks, with support for custom datasets. Users only need to prepare datasets for one-click training. +- **Hardware Support**: Supports A10/A100/H100, RTX series, T4/V100, AMD GPU (MI300 series, etc.), CPU, MPS, and domestic hardware Ascend NPU, etc. +- **Lightweight Training**: Supports lightweight fine-tuning methods such as LoRA, QLoRA, DoRA, LoRA+, LLaMAPro, LongLoRA, LoRA-GA, ReFT, RS-LoRA, Adapter, LISA, etc. +- **Quantized Training**: Supports training on BNB, AWQ, GPTQ, AQLM, HQQ, EETQ quantized models, requiring only 9GB training resources for 7B models. +- **Memory Optimization**: GaLore, Q-Galore, UnSloth, Liger-Kernel, Flash-Attention 2/3, and **Ulysses and Ring-Attention sequence parallelism techniques** support, reducing memory consumption for long-text training. +- **Distributed Training**: Supports distributed data parallelism (DDP), device_map simple model parallelism, DeepSpeed ZeRO2 ZeRO3, FSDP/FSDP2, and Megatron distributed training technologies. +- 🍓 **Multimodal Training**: Supports multimodal packing technology to improve training speed by 100%+, supports mixed modality data training with text, images, video and audio, and supports independent control of vit/aligner/llm. +- **Agent Training**: Supports Agent templates, allowing one dataset to be used for training different models. +- 🍊 **Training Tasks**: Supports pre-training and instruction fine-tuning, as well as training tasks such as DPO, GKD, KTO, RM, CPO, SimPO, ORPO, and supports **Embedding/Reranker** and sequence classification tasks. +- 🥥 **Megatron Parallelism**: Provides TP/PP/SP/CP/ETP/EP/VPP parallel strategies to significantly boost **MoE model training speed**. Supports full-parameter and LoRA training methods for 300+ pure text large models and 100+ multimodal large models. Supports CPT/SFT/GRPO/DPO/KTO/RM training tasks. +- 🍉 **Reinforcement Learning**: Built-in **rich GRPO family algorithms**, including GRPO, DAPO, GSPO, SAPO, CISPO, CHORD, RLOO, Reinforce++, etc. Supports synchronous and asynchronous vLLM engine inference acceleration, with extensible reward functions, multi-turn inference Schedulers, and environments through plugins. +- **Full-Pipeline Capabilities**: Covers the entire workflow of training, inference, evaluation, quantization, and deployment. +- **UI Training**: Provides Web-UI interface for training, inference, evaluation, and quantization, completing the full pipeline for large models. +- **Inference Acceleration**: Supports Transformers, vLLM, SGLang, and LmDeploy inference acceleration engines, providing OpenAI interfaces for accelerating inference, deployment, and evaluation modules. +- **Model Evaluation**: Uses EvalScope as the evaluation backend, supporting 100+ evaluation datasets for evaluating text-only and multimodal models. +- **Model Quantization**: Supports quantization export for AWQ, GPTQ, FP8, and BNB. Exported models support inference acceleration using vLLM/SGLang/LmDeploy. + + +## 🎉 News +- 🎁 2026.06.10: Megatron-Ray now supports GRPO and GKD training. See [docs](./docs/source_en/Instruction/Ray.md) and [examples](examples/ray). +- 🎁 2026.03.03: **ms-swift v4.0** major version is officially released. For release notes, please refer to [here](https://github.com/modelscope/ms-swift/releases/tag/v4.0.0). You can provide your suggestions to us in [this issue](https://github.com/modelscope/ms-swift/issues/7250). Thank you for your support. +- 🎁 2025.11.14: Megatron GRPO is now available! Check out the [docs](./docs/source_en/Megatron-SWIFT/GRPO.md) and [examples](examples/megatron/grpo). +- 🎁 2025.11.04: Support for [Mcore-Bridge](docs/source_en/Megatron-SWIFT/Mcore-Bridge.md), making Megatron training as simple and easy to use as transformers. +- 🎁 2025.10.28: Ray [here](docs/source_en/Instruction/Ray.md). +- 🎁 2025.09.07: Added support for CHORD training algorithm. See the [documentation](./docs/source_en/Instruction/GRPO/AdvancedResearch/CHORD.md). +- 🎁 2025.09.06: Ulysses can now be used with ring-attention, allowing sequences to be sharded into any number of chunks (no longer limited by the number of heads). The argument remains `--sequence_parallel_size N`. +- 🎁 2025.09.02: Megatron-SWIFT now supports multimodal model training. Documentation can be found [here](./docs/source_en/Megatron-SWIFT/Multimodal-Model.md). +- 🎁 2025.08.12: Support [Dynamic Fine-Tuning](https://arxiv.org/abs/2508.05629)(DFT) in SFT training, use parameter `--enable_dft_loss true`. Training scripts can be found [here](https://github.com/modelscope/ms-swift/blob/main/examples/train/full/dft.sh). +- 🎁 2025.07.09: Megatron-SWIFT supports LoRA training. Compared to ms-swift, it achieves significant speedup on MoE models. Training scripts can be found [here](https://github.com/modelscope/ms-swift/blob/main/examples/megatron/lora). +- 🎁 2025.06.23: Fine-tuning of reranker models is supported. Training scripts can be found here: [Reranker](https://github.com/modelscope/ms-swift/blob/main/examples/train/reranker/train_reranker.sh). +- 🎁 2025.06.15: Support for GKD training on both pure text large models and multimodal models. Training scripts can be found here: [Pure Text](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/gkd), [Multimodal](https://github.com/modelscope/ms-swift/blob/main/examples/train/multimodal/rlhf/gkd). + +
More + +- 🎁 2025.06.11: Support for using Megatron parallelism techniques for RLHF training. The training script can be found [here](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/rlhf). +- 🎁 2025.05.29: Support sequence parallel in pretrain, sft, dpo and grpo, check script [here](https://github.com/modelscope/ms-swift/tree/main/examples/train/sequence_parallel). +- 🎁 2025.05.11: GRPO now supports custom processing logic for reward models. See the GenRM example [here](./docs/source_en/Instruction/GRPO/DeveloperGuide/reward_model.md). +- 🎁 2025.04.15: The ms-swift paper has been accepted by AAAI 2025. You can find the paper at [this link](https://ojs.aaai.org/index.php/AAAI/article/view/35383). +- 🎁 2025.03.23: Multi-round GRPO is now supported for training multi-turn dialogue scenarios (e.g., agent tool calling). Please refer to the [doc](./docs/source_en/Instruction/GRPO/DeveloperGuide/multi_turn.md). +- 🎁 2025.03.16: Support for Megatron's parallel training techniques is now available. Please see the [Megatron-SWIFT training documentation](https://swift.readthedocs.io/en/latest/Megatron-SWIFT/Quick-start.html). +- 🎁 2025.03.15: Fine-tuning of embedding models for both pure text and multimodal models is supported. Please check the [training script](examples/train/embedding). +- 🎁 2025.03.05: The hybrid mode for GRPO is supported, with a script for training a 72B model on 4 GPUs (4*80G) available [here](examples/train/grpo/internal/vllm_72b_4gpu.sh). Tensor parallelism with vllm is also supported, with the training script available [here](examples/train/grpo/internal). +- 🎁 2025.02.21: The GRPO algorithm now supports LMDeploy, with the training script available [here](examples/train/grpo/internal/full_lmdeploy.sh). Additionally, the performance of the GRPO algorithm has been tested, achieving a training speed increase of up to 300% using various tricks. Please check the WanDB table [here](https://wandb.ai/tastelikefeet/grpo_perf_test?nw=nwuseryuzezyz). +- 🎁 2025.02.21: The `swift sample` command is now supported. The reinforcement fine-tuning script can be found [here](docs/source_en/Instruction/Reinforced-Fine-tuning.md), and the large model API distillation sampling script is available [here](examples/sampler/distill/distill.sh). +- 🔥 2025.02.12: Support for the GRPO (Group Relative Policy Optimization) training algorithm has been added. Documentation is available [here](docs/source_en/Instruction/GRPO/GetStarted/GRPO.md). +- 🎁 2024.12.04: Major update to **ms-swift 3.0**. Please refer to the [release notes and changes](docs/source_en/Instruction/ReleaseNote3.0.md). +- 🎉 2024.08.12: The ms-swift paper has been published on arXiv and can be read [here](https://arxiv.org/abs/2408.05517). +- 🔥 2024.08.05: Support for using [evalscope](https://github.com/modelscope/evalscope/) as a backend for evaluating large models and multimodal models. +- 🔥 2024.07.29: Support for using [vllm](https://github.com/vllm-project/vllm) and [lmdeploy](https://github.com/InternLM/lmdeploy) to accelerate inference for large models and multimodal models. When performing infer/deploy/eval, you can specify `--infer_backend vllm/lmdeploy`. +- 🔥 2024.07.24: Support for human preference alignment training for multimodal large models, including DPO/ORPO/SimPO/CPO/KTO/RM/PPO. +- 🔥 2024.02.01: Support for Agent training! The training algorithm is derived from [this paper](https://arxiv.org/pdf/2309.00986.pdf). +
+ +## 🛠️ Installation +To install using pip: +```shell +pip install ms-swift -U + +# Using uv +pip install uv +uv pip install ms-swift -U --torch-backend=auto +``` + +To install from source: +```shell +# pip install git+https://github.com/modelscope/ms-swift.git + +git clone https://github.com/modelscope/ms-swift.git +cd ms-swift +# The main branch is for swift 4.x. To install swift 3.x, please run the following command: +# git checkout release/3.12 +pip install -e . + +# Using uv +uv pip install -e . --torch-backend=auto +``` + +Running Environment: + +| | Range | Recommended | Notes | +|--------------|--------------|---------------------|-------------------------------------------| +| python | >=3.10 | 3.12 | | +| cuda | | cuda12.8/13.0 | No need to install if using CPU, NPU, MPS | +| torch | >=2.0 | 2.8.0/2.11.0 | | +| transformers | >=4.33 | 4.57.6/5.12.1 | | +| modelscope | >=1.23 | | | +| datasets | >=3.0,<4.8.5 | 3.6.0/4.8.4 | | +| peft | >=0.11,<0.20 | | | +| flash_attn | | 2.8.3/4.0.0b15 | | +| trl | >=0.15,<1.0 | 0.29.1 | RLHF | +| deepspeed | >=0.14 | 0.18.9 | Training | +| vllm | >=0.5.1 | 0.11.0/0.23.0 | Inference/Deployment | +| sglang | >=0.4.6 | | Inference/Deployment | +| evalscope | >=1.0 | | Evaluation | +| gradio | | 5.32.1 | Web-UI/App | + +For more optional dependencies, you can refer to [here](https://github.com/modelscope/ms-swift/blob/main/requirements/install_all.sh). + + +## 🚀 Quick Start + +10 minutes of self-cognition fine-tuning of Qwen3-4B-Instruct-2507 on a single 3090 GPU: + +### Command Line Interface (Recommended) + +```shell +# 13GB +CUDA_VISIBLE_DEVICES=0 \ +swift sft \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --tuner_type lora \ + --dataset 'AI-ModelScope/alpaca-gpt4-data-zh#500' \ + 'AI-ModelScope/alpaca-gpt4-data-en#500' \ + 'swift/self-cognition#500' \ + --torch_dtype bfloat16 \ + --num_train_epochs 1 \ + --per_device_train_batch_size 1 \ + --per_device_eval_batch_size 1 \ + --learning_rate 1e-4 \ + --lora_rank 8 \ + --lora_alpha 32 \ + --target_modules all-linear \ + --gradient_accumulation_steps 16 \ + --eval_steps 50 \ + --save_steps 50 \ + --save_total_limit 2 \ + --logging_steps 5 \ + --max_length 2048 \ + --output_dir output \ + --warmup_ratio 0.05 \ + --dataloader_num_workers 4 \ + --model_author swift \ + --model_name swift-robot +``` + +Tips: + +- If you want to train with a custom dataset, you can refer to [this guide](https://swift.readthedocs.io/en/latest/Customization/Custom-dataset.html) to organize your dataset format and specify `--dataset `. +- The `--model_author` and `--model_name` parameters are only effective when the dataset includes `swift/self-cognition`. +- To train with a different model, simply modify `--model `. +- By default, **ModelScope** is used for downloading models and datasets. If you want to use HuggingFace, simply specify `--use_hf true`. + +After training is complete, use the following command to infer with the trained weights: + +- Here, `--adapters` should be replaced with the last checkpoint folder generated during training. Since the adapters folder contains the training parameter file `args.json`, there is no need to specify `--model`, `--system` separately; Swift will automatically read these parameters. To disable this behavior, you can set `--load_args false`. + +```shell +# Using an interactive command line for inference. +CUDA_VISIBLE_DEVICES=0 \ +swift infer \ + --adapters output/vx-xxx/checkpoint-xxx \ + --stream true \ + --temperature 0 \ + --max_new_tokens 2048 + +# merge-lora and use vLLM for inference acceleration +CUDA_VISIBLE_DEVICES=0 \ +swift infer \ + --adapters output/vx-xxx/checkpoint-xxx \ + --stream true \ + --merge_lora true \ + --infer_backend vllm \ + --vllm_max_model_len 8192 \ + --temperature 0 \ + --max_new_tokens 2048 +``` + +Finally, use the following command to push the model to ModelScope: + +```shell +CUDA_VISIBLE_DEVICES=0 \ +swift export \ + --adapters output/vx-xxx/checkpoint-xxx \ + --push_to_hub true \ + --hub_model_id '' \ + --hub_token '' \ + --use_hf false +``` + + +### Web-UI +The Web-UI is a **zero-threshold** training and deployment interface solution based on Gradio interface technology. For more details, you can check [here](https://swift.readthedocs.io/en/latest/GetStarted/Web-UI.html). + +```shell +SWIFT_UI_LANG=en swift web-ui +``` + +![image.png](./docs/resources/web-ui-en.jpg) + +### Using Python + +ms-swift also supports training and inference using Python. Below is pseudocode for training and inference. For more details, you can refer to [here](https://github.com/modelscope/ms-swift/blob/main/examples/notebook/qwen2_5-self-cognition/self-cognition-sft.ipynb). + +Training: + +```python +from peft import LoraConfig, get_peft_model +from swift import get_model_processor, get_template, load_dataset, EncodePreprocessor +from swift.trainers import Seq2SeqTrainer, Seq2SeqTrainingArguments +# Retrieve the model and template, and add a trainable LoRA module +model, tokenizer = get_model_processor(model_id_or_path, ...) +template = get_template(tokenizer, ...) +lora_config = LoraConfig(...) +model = get_peft_model(model, lora_config) + +# Download and load the dataset, and encode the text into tokens +train_dataset, val_dataset = load_dataset(dataset_id_or_path, ...) +train_dataset = EncodePreprocessor(template=template)(train_dataset, num_proc=num_proc) +val_dataset = EncodePreprocessor(template=template)(val_dataset, num_proc=num_proc) + +# Train the model +training_args = Seq2SeqTrainingArguments(...) +trainer = Seq2SeqTrainer( + model=model, + args=training_args, + template=template, + train_dataset=train_dataset, + eval_dataset=val_dataset, +) +trainer.train() +``` +Inference: + +```python +from swift import TransformersEngine, InferRequest, RequestConfig +# Perform inference using the native Transformers engine +engine = TransformersEngine(model_id_or_path, adapters=[lora_checkpoint]) +infer_request = InferRequest(messages=[{'role': 'user', 'content': 'who are you?'}]) +request_config = RequestConfig(max_tokens=max_new_tokens, temperature=temperature) + +resp_list = engine.infer([infer_request], request_config) +print(f'response: {resp_list[0].choices[0].message.content}') +``` + +## ✨ Usage +Here is a minimal example of training to deployment using ms-swift. For more details, you can check the [examples](https://github.com/modelscope/ms-swift/tree/main/examples). + +- If you want to use other models or datasets (including multimodal models and datasets), you only need to modify `--model` to specify the corresponding model's ID or path, and modify `--dataset` to specify the corresponding dataset's ID or path. +- By default, ModelScope is used for downloading models and datasets. If you want to use HuggingFace, simply specify `--use_hf true`. + +| Useful Links | +| ------ | +| [🔥Command Line Parameters](https://swift.readthedocs.io/en/latest/Instruction/Command-line-parameters.html) | +| [Megatron-SWIFT](https://swift.readthedocs.io/en/latest/Megatron-SWIFT/Quick-start.html) | +| [GRPO](https://swift.readthedocs.io/en/latest/Instruction/GRPO/GetStarted/GRPO.html) | +| [Supported Models and Datasets](https://swift.readthedocs.io/en/latest/Instruction/Supported-models-and-datasets.html) | +| [Custom Models](https://swift.readthedocs.io/en/latest/Customization/Custom-model.html), [🔥Custom Datasets](https://swift.readthedocs.io/en/latest/Customization/Custom-dataset.html) | +| [LLM Tutorial](https://github.com/modelscope/modelscope-classroom/tree/main/LLM-tutorial) | + +### Training + +Supported Training Methods: + +| Method | Full-Parameter | LoRA | QLoRA | Deepspeed | Multi-Machine | Multimodal | +| ------------------------------------------------------------ | ------------------------------------------------------------ | ---- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | +| [Pre-training](https://github.com/modelscope/ms-swift/blob/main/examples/train/pretrain) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [Supervised Fine-Tuning](https://github.com/modelscope/ms-swift/blob/main/examples/train/lora_sft.sh) | [✅](https://github.com/modelscope/ms-swift/blob/main/examples/train/full/train.sh) | ✅ | [✅](https://github.com/modelscope/ms-swift/tree/main/examples/train/qlora) | [✅](https://github.com/modelscope/ms-swift/tree/main/examples/train/multi-gpu/deepspeed) | [✅](https://github.com/modelscope/ms-swift/tree/main/examples/train/multi-node) | [✅](https://github.com/modelscope/ms-swift/tree/main/examples/train/multimodal) | +| [GRPO](https://github.com/modelscope/ms-swift/blob/main/examples/train/grpo) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [GKD](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/gkd) | ✅ | ✅ | ✅ | ✅ | ✅ | [✅](https://github.com/modelscope/ms-swift/blob/main/examples/train/multimodal/rlhf/gkd) | +| [PPO](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/ppo) | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | +| [DPO](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/dpo) | ✅ | ✅ | ✅ | ✅ | ✅ | [✅](https://github.com/modelscope/ms-swift/blob/main/examples/train/multimodal/rlhf/dpo) | +| [KTO](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/kto.sh) | ✅ | ✅ | ✅ | ✅ | ✅ | [✅](https://github.com/modelscope/ms-swift/blob/main/examples/train/multimodal/rlhf/kto.sh) | +| [Reward Model](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/rm.sh) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [CPO](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/cpo.sh) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [SimPO](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/simpo.sh) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [ORPO](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/orpo.sh) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [Embedding](https://github.com/modelscope/ms-swift/blob/main/examples/train/embedding) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [Reranker](https://github.com/modelscope/ms-swift/tree/main/examples/train/reranker) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [Sequence Classification](https://github.com/modelscope/ms-swift/blob/main/examples/train/seq_cls) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | + + +Pre-training: +```shell +# 8*A100 +NPROC_PER_NODE=8 \ +CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \ +swift pt \ + --model Qwen/Qwen3-4B-Base \ + --dataset swift/chinese-c4 \ + --streaming true \ + --tuner_type full \ + --deepspeed zero2 \ + --output_dir output \ + --max_steps 10000 \ + ... +``` + +Fine-tuning: +```shell +CUDA_VISIBLE_DEVICES=0 swift sft \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --dataset AI-ModelScope/alpaca-gpt4-data-en \ + --tuner_type lora \ + --output_dir output \ + ... +``` + +RLHF: +```shell +CUDA_VISIBLE_DEVICES=0 swift rlhf \ + --rlhf_type dpo \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --dataset hjh0119/shareAI-Llama3-DPO-zh-en-emoji \ + --tuner_type lora \ + --output_dir output \ + ... +``` + + +### Megatron-SWIFT + +ms-swift supports using Megatron parallelism techniques to accelerate training, including large-scale cluster training and MoE model training. The following training methods are supported: + +| Method | Full-Parameter | LoRA | MoE | Multimodal | FP8 | +| ---------------------- | -------------- | ---- | ---- | ---------- | ---- | +| Pre-training | ✅ | ✅ | ✅ | ✅ | ✅ | +| [Supervised Fine-Tuning](https://github.com/modelscope/ms-swift/tree/main/examples/megatron) | ✅ | ✅ | ✅ | ✅ | ✅ | +| [GRPO](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/grpo) | ✅ | ✅ | ✅ | ✅ | ✅ | +| [GKD](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/rlhf/gkd) | ✅ | ✅ | ✅ | ✅ | ✅ | +| [DPO](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/rlhf/dpo) | ✅ | ✅ | ✅ | ✅ | ✅ | +| [KTO](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/rlhf/kto) | ✅ | ✅ | ✅ | ✅ | ✅ | +| [RM](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/rlhf/rm) | ✅ | ✅ | ✅ | ✅ | ✅ | +| [Embedding](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/embedding) | ✅ | ✅| ✅ | ✅ | ✅ | +| [Reranker](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/reranker) | ✅ | ✅| ✅ | ✅ | ✅ | +| [Sequence Classification](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/seq_cls) | ✅ | ✅ | ✅ | ✅ | ✅ | + + +```shell +NPROC_PER_NODE=2 CUDA_VISIBLE_DEVICES=0,1 megatron sft \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --save_safetensors true \ + --dataset AI-ModelScope/alpaca-gpt4-data-zh \ + --tuner_type lora \ + --output_dir output \ + ... +``` + +### Reinforcement Learning + +ms-swift supports a rich set of GRPO family algorithms: + +| Method | Full-Parameter | LoRA | Multimodal | Multi-Machine | +| ------------------------------------------------------------ | -------------- | ---- | ---------- | ------------- | +| [GRPO](https://swift.readthedocs.io/en/latest/Instruction/GRPO/GetStarted/GRPO.html) | ✅ | ✅ | ✅ | ✅ | +| [DAPO](https://swift.readthedocs.io/en/latest/Instruction/GRPO/AdvancedResearch/DAPO.html) | ✅ | ✅ | ✅ | ✅ | +| [GSPO](https://swift.readthedocs.io/en/latest/Instruction/GRPO/AdvancedResearch/GSPO.html) | ✅ | ✅ | ✅ | ✅ | +| [SAPO](https://swift.readthedocs.io/en/latest/Instruction/GRPO/AdvancedResearch/SAPO.html) | ✅ | ✅ | ✅ | ✅ | +| [CISPO](https://swift.readthedocs.io/en/latest/Instruction/GRPO/AdvancedResearch/CISPO.html) | ✅ | ✅ | ✅ | ✅ | +| [CHORD](https://swift.readthedocs.io/en/latest/Instruction/GRPO/AdvancedResearch/CHORD.html) | ✅ | ✅ | ✅ | ✅ | +| [RLOO](https://swift.readthedocs.io/en/latest/Instruction/GRPO/AdvancedResearch/RLOO.html) | ✅ | ✅ | ✅ | ✅ | +| [Reinforce++](https://swift.readthedocs.io/en/latest/Instruction/GRPO/AdvancedResearch/REINFORCEPP.html) | ✅ | ✅ | ✅ | ✅ | + +```shell +CUDA_VISIBLE_DEVICES=0,1,2,3 NPROC_PER_NODE=4 \ +swift rlhf \ + --rlhf_type grpo \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --tuner_type lora \ + --use_vllm true \ + --vllm_mode colocate \ + --dataset AI-MO/NuminaMath-TIR#10000 \ + --output_dir output \ + ... +``` + + +### Inference +```shell +CUDA_VISIBLE_DEVICES=0 swift infer \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --stream true \ + --infer_backend transformers \ + --max_new_tokens 2048 +``` + +### Interface Inference +```shell +CUDA_VISIBLE_DEVICES=0 swift app \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --stream true \ + --infer_backend transformers \ + --max_new_tokens 2048 +``` + +### Deployment +```shell +CUDA_VISIBLE_DEVICES=0 swift deploy \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --infer_backend vllm +``` + +### Sampling +```shell +CUDA_VISIBLE_DEVICES=0 swift sample \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --sampler_engine transformers \ + --num_return_sequences 5 \ + --dataset AI-ModelScope/alpaca-gpt4-data-zh#5 +``` + +### Evaluation +```shell +CUDA_VISIBLE_DEVICES=0 swift eval \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --infer_backend sglang \ + --eval_backend OpenCompass \ + --eval_dataset ARC_c +``` + +### Quantization +```shell +CUDA_VISIBLE_DEVICES=0 swift export \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --quant_method fp8 \ + --dataset AI-ModelScope/alpaca-gpt4-data-zh \ + --output_dir Qwen3-4B-Instruct-2507-FP8 +``` + +### Push Model +```shell +swift export \ + --model \ + --push_to_hub true \ + --hub_model_id '' \ + --hub_token '' +``` + +## 🏛 License + +This framework is licensed under the [Apache License (Version 2.0)](https://github.com/modelscope/ms-swift/blob/master/LICENSE). For models and datasets, please refer to the original resource page and follow the corresponding License. + +## 📎 Citation + +```bibtex +@misc{zhao2024swiftascalablelightweightinfrastructure, + title={SWIFT:A Scalable lightWeight Infrastructure for Fine-Tuning}, + author={Yuze Zhao and Jintao Huang and Jinghan Hu and Xingjun Wang and Yunlin Mao and Daoze Zhang and Zeyinzi Jiang and Zhikai Wu and Baole Ai and Ang Wang and Wenmeng Zhou and Yingda Chen}, + year={2024}, + eprint={2408.05517}, + archivePrefix={arXiv}, + primaryClass={cs.CL}, + url={https://arxiv.org/abs/2408.05517}, +} +``` + +## Star History + +[![Star History Chart](https://api.star-history.com/svg?repos=modelscope/ms-swift&type=Date)](https://star-history.com/#modelscope/ms-swift&Date) diff --git a/README.wehub.md b/README.wehub.md new file mode 100644 index 0000000..a96bfa6 --- /dev/null +++ b/README.wehub.md @@ -0,0 +1,7 @@ +# WeHub 来源说明 + +- 原始项目：`modelscope/ms-swift` +- 原始仓库：https://github.com/modelscope/ms-swift +- 导入方式：上游默认分支的最新快照 +- 原作者、版权和许可证信息以原始仓库及本仓库 LICENSE 为准 +- 本文件仅用于记录来源，不代表 WeHub 是原项目作者 diff --git a/README_CN.md b/README_CN.md new file mode 100644 index 0000000..e791361 --- /dev/null +++ b/README_CN.md @@ -0,0 +1,497 @@ +# SWIFT (Scalable lightWeight Infrastructure for Fine-Tuning) + +

+
+ +
+

+

+魔搭社区官网 +
+ 中文  ｜  English  +

+ + +

+ + + + + + + +

+ +

+modelscope/ms-swift | Trendshift +

+ +

+ 论文   ｜ English Documentation   ｜   中文文档   +

+ +## 📖 目录 +- [用户群](#-用户群) +- [简介](#-简介) +- [新闻](#-新闻) +- [安装](#%EF%B8%8F-安装) +- [快速开始](#-快速开始) +- [如何使用](#-如何使用) +- [License](#-license) +- [引用](#-引用) + +## ☎ 用户群 + +请扫描下面的二维码来加入我们的交流群： + +[Discord Group](https://discord.gg/yeN59wxjwe) | 微信群 +:-------------------------:|:-------------------------: + | + +## 📝 简介 +🍲 **ms-swift**是魔搭社区提供的大模型与多模态大模型微调部署框架，现已支持600+纯文本大模型与400+多模态大模型的训练（预训练、微调、人类对齐）、推理、评测、量化与部署。其中大模型包括：Qwen3、Qwen3.5、InternLM3、GLM4.5、Mistral、DeepSeek-R1、Llama4等模型，多模态大模型包括：Qwen3-VL、Qwen3-Omni、Llava、InternVL3.5、MiniCPM-V-4、Ovis2.5、GLM4.5-V、DeepSeek-VL2等模型。 + +🍔 除此之外，ms-swift汇集了最新的训练技术，包括集成Megatron并行技术，包括TP、PP、CP、EP等为训练提供加速，以及众多GRPO算法族强化学习的算法，包括：GRPO、DAPO、GSPO、SAPO、CISPO、RLOO、Reinforce++等提升模型智能。ms-swift支持广泛的训练任务，包括DPO、KTO、RM、CPO、SimPO、ORPO等偏好学习算法，以及Embedding、Reranker、序列分类任务。ms-swift提供了大模型训练全链路的支持，包括使用vLLM、SGLang和LMDeploy对推理、评测、部署模块提供加速，以及使用GPTQ、AWQ、BNB、FP8技术对大模型进行量化。 + +**为什么选择ms-swift？** +- 🍎 **模型类型**：支持**600+纯文本大模型**、**400+多模态大模型**以及All-to-All全模态模型训练到部署全流程，热门模型Day0支持。 +- **数据集类型**：内置150+预训练、微调、人类对齐、多模态等各种任务数据集，并支持自定义数据集，用户只需准备数据集即可一键训练。 +- **硬件支持**：支持A10/A100/H100、RTX系列、T4/V100、CPU、MPS以及国产硬件Ascend NPU等。 +- **轻量训练**：支持了LoRA、QLoRA、DoRA、LoRA+、LLaMAPro、LongLoRA、LoRA-GA、ReFT、RS-LoRA、Adapter、LISA等轻量微调方式。 +- **量化训练**：支持对BNB、AWQ、GPTQ、AQLM、HQQ、EETQ量化模型进行训练，7B模型训练只需9GB训练资源。 +- **显存优化**: GaLore、Q-Galore、UnSloth、Liger-Kernel、Flash-Attention 2/3 以及 **Ulysses和Ring-Attention序列并行技术**支持，降低长文本训练显存占用。 +- **分布式训练**：支持分布式数据并行（DDP）、device_map简易模型并行、DeepSpeed ZeRO2 ZeRO3、FSDP/FSDP2以及Megatron等分布式训练技术。 +- 🍓 **多模态训练**：支持多模态packing技术提升训练速度100%+，支持文本、图像、视频和语音混合模态数据训练，支持vit/aligner/llm单独控制。 +- **Agent训练**：支持Agent template，准备一套数据集可用于不同模型的训练。 +- 🍊 **训练任务**：支持预训练和指令微调，以及DPO、GKD、KTO、RM、CPO、SimPO、ORPO等训练任务，支持**Embedding/Reranker**和序列分类任务。 +- 🥥 **Megatron并行技术**：提供TP/PP/SP/CP/ETP/EP/VPP并行策略，显著提升**MoE模型训练速度**。支持300+纯文本大模型和100+多模态大模型的全参数和LoRA训练方法。支持CPT/SFT/GRPO/DPO/KTO/RM训练任务。 +- 🍉 **强化学习**：内置**丰富GRPO族算法**，包括GRPO、DAPO、GSPO、SAPO、CISPO、CHORD、RLOO、Reinforce++等，支持同步和异步vLLM引擎推理加速，可使用插件拓展奖励函数、多轮推理调度器以及环境等。 +- **全链路能力**：覆盖训练、推理、评测、量化和部署全流程。 +- **界面训练**：提供使用Web-UI界面的方式进行训练、推理、评测、量化，完成大模型的全链路。 +- **推理加速**：支持Transformers、vLLM、SGLang和LmDeploy推理加速引擎，并提供OpenAI接口，为推理、部署和评测模块提供加速。 +- **模型评测**：以EvalScope作为评测后端，支持100+评测数据集对纯文本和多模态模型进行评测。 +- **模型量化**：支持AWQ、GPTQ、FP8和BNB的量化导出，导出的模型支持使用vLLM/SGLang/LmDeploy推理加速。 + +## 🎉 新闻 +- 🎁 2026.06.10: Megatron-Ray支持GRPO和GKD训练，查看[文档](docs/source/Instruction/Ray.md)和[示例](examples/ray)。 +- 🎁 2026.03.03: **ms-swift v4.0**大版本正式发布，release note参考[这里](https://github.com/modelscope/ms-swift/releases/tag/v4.0.0)，您的建议可以在[这个issue](https://github.com/modelscope/ms-swift/issues/7250)中反馈给我们，感谢您的支持。 +- 🎁 2025.11.14: Megatron GRPO现已支持！查看[文档](./docs/source/Megatron-SWIFT/GRPO.md)和[示例](examples/megatron/grpo)。 +- 🎁 2025.11.04: 支持[Mcore-Bridge](docs/source/Megatron-SWIFT/Mcore-Bridge.md)，使Megatron训练像transformers一样简单易用。 +- 🎁 2025.10.28: Ray [已支持](docs/source/Instruction/Ray.md)。 +- 🎁 2025.09.07: 支持CHORD训练算法，请查看[文档](docs/source/Instruction/GRPO/AdvancedResearch/CHORD.md)。 +- 🎁 2025.09.06: Ulysses现已支持与ring-attention结合使用，使得输入序列可以被切分成任意数量的块（不再受限于num_heads），命令参数仍然是`--sequence_parallel_size N`。 +- 🎁 2025.09.02: Megatron-SWIFT支持多模态模型训练。文档参考[这里](./docs/source/Megatron-SWIFT/Multimodal-Model.md)。 +- 🎁 2025.08.12: 支持在SFT训练中使用[Dynamic Fine-Tuning](https://arxiv.org/abs/2508.05629)(DFT)，使用参数 `--enable_dft_loss true`。训练脚本参考[这里](https://github.com/modelscope/ms-swift/blob/main/examples/train/full/dft.sh) +- 🎁 2025.07.09: Megatron-SWIFT支持LoRA训练。相比ms-swift，在MoE模型提速显著。训练脚本参考[这里](https://github.com/modelscope/ms-swift/blob/main/examples/megatron/lora)。 +- 🎁 2025.06.23: 支持Reranker模型训练，训练脚本参考[这里](https://github.com/modelscope/ms-swift/blob/main/examples/train/reranker/train_reranker.sh)。 +- 🎁 2025.06.15: 支持对纯文本大模型和多模态模型进行GKD训练。训练脚本参考这里：[纯文本](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/gkd), [多模态](https://github.com/modelscope/ms-swift/blob/main/examples/train/multimodal/rlhf/gkd)。 + +
更多 + +- 🎁 2025.06.11: 支持使用Megatron并行技术进行RLHF训练，训练脚本参考[这里](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/rlhf)。 +- 🎁 2025.05.29: 支持pt、sft、dpo、grpo的序列并行，具体请查看[脚本](https://github.com/modelscope/ms-swift/tree/main/examples/train/sequence_parallel)。 +- 🎁 2025.05.11: GRPO中的奖励模型支持自定义处理逻辑，GenRM的例子参考[这里](./docs/source/Instruction/GRPO/DeveloperGuide/reward_model.md)。 +- 🎁 2025.04.15: ms-swift论文已经被AAAI 2025接收，论文地址在[这里](https://ojs.aaai.org/index.php/AAAI/article/view/35383)。 +- 🎁 2025.03.23: 支持了多轮GRPO，用于构建多轮对话场景的训练(例如agent tool calling)，请查看[文档](docs/source/Instruction/GRPO/DeveloperGuide/multi_turn.md)。 +- 🎁 2025.03.16: 支持了Megatron的并行技术进行训练，请查看[Megatron-SWIFT训练文档](https://swift.readthedocs.io/zh-cn/latest/Megatron-SWIFT/Quick-start.html)。 +- 🎁 2025.03.15: 支持纯文本和多模态模型的embedding模型的微调，请查看[训练脚本](examples/train/embedding)。 +- 🎁 2025.03.05: 支持GRPO的hybrid模式，4GPU(4*80G)训练72B模型的脚本参考[这里](examples/train/grpo/internal/vllm_72b_4gpu.sh)。同时支持vllm的tensor并行，训练脚本参考[这里](examples/train/grpo/internal)。 +- 🎁 2025.02.21: GRPO算法支持使用LMDeploy，训练脚本参考[这里](examples/train/grpo/internal/full_lmdeploy.sh)。此外测试了GRPO算法的性能，使用一些tricks使训练速度提高到300%。WanDB表格请查看[这里](https://wandb.ai/tastelikefeet/grpo_perf_test?nw=nwuseryuzezyz)。 +- 🎁 2025.02.21: 支持`swift sample`命令。强化微调脚本参考[这里](docs/source/Instruction/Reinforced-Fine-tuning.md)，大模型API蒸馏采样脚本参考[这里](examples/sampler/distill/distill.sh)。 +- 🔥 2025.02.12: 支持GRPO (Group Relative Policy Optimization) 训练算法，文档参考[这里](docs/source/Instruction/GRPO/GetStarted/GRPO.md)。 +- 🎁 2024.12.04: **ms-swift3.0**大版本更新。请查看[发布说明和更改](docs/source/Instruction/ReleaseNote3.0.md)。 +- 🎉 2024.08.12: ms-swift论文已经发布到arXiv上，可以点击[这里](https://arxiv.org/abs/2408.05517)阅读。 +- 🔥 2024.08.05: 支持使用[evalscope](https://github.com/modelscope/evalscope/)作为后端进行大模型和多模态模型的评测。 +- 🔥 2024.07.29: 支持使用[vllm](https://github.com/vllm-project/vllm), [lmdeploy](https://github.com/InternLM/lmdeploy)对大模型和多模态大模型进行推理加速，在infer/deploy/eval时额外指定`--infer_backend vllm/lmdeploy`即可。 +- 🔥 2024.07.24: 支持对多模态大模型进行人类偏好对齐训练，包括DPO/ORPO/SimPO/CPO/KTO/RM/PPO。 +- 🔥 2024.02.01: 支持Agent训练！训练算法源自这篇[论文](https://arxiv.org/pdf/2309.00986.pdf)。 +
+ +## 🛠️ 安装 +使用pip进行安装： +```shell +pip install ms-swift -U + +# 使用uv +pip install uv +uv pip install ms-swift -U --torch-backend=auto +``` + +从源代码安装： +```shell +# pip install git+https://github.com/modelscope/ms-swift.git + +git clone https://github.com/modelscope/ms-swift.git +cd ms-swift +# main分支为swift4.x。若安装swift3.x，请运行以下命令 +# git checkout release/3.12 +pip install -e . + +# 使用uv +uv pip install -e . --torch-backend=auto +``` + +运行环境： + +| | 范围 | 推荐 | 备注 | +|--------------|--------------|---------------------|--------------------| +| python | >=3.10 | 3.12 | | +| cuda | | cuda12.8/13.0 | 使用cpu、npu、mps则无需安装 | +| torch | >=2.0 | 2.8.0/2.11.0 | | +| transformers | >=4.33 | 4.57.6/5.12.1 | | +| modelscope | >=1.23 | | | +| datasets | >=3.0,<4.8.5 | 3.6.0/4.8.4 | | +| peft | >=0.11,<0.20 | | | +| flash_attn | | 2.8.3/4.0.0b15 | | +| trl | >=0.15,<1.0 | 0.29.1 | RLHF | +| deepspeed | >=0.14 | 0.18.9 | 训练 | +| vllm | >=0.5.1 | 0.11.0/0.23.0 | 推理/部署 | +| sglang | >=0.4.6 | | 推理/部署 | +| evalscope | >=1.0 | | 评测 | +| gradio | | 5.32.1 | Web-UI/App | + +更多可选依赖可以参考[这里](https://github.com/modelscope/ms-swift/blob/main/requirements/install_all.sh)。 + + +## 🚀 快速开始 + +**10分钟**在单卡3090上对Qwen3-4B-Instruct-2507进行自我认知微调： + +### 命令行（推荐） +```shell +# 13GB +CUDA_VISIBLE_DEVICES=0 \ +swift sft \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --tuner_type lora \ + --dataset 'AI-ModelScope/alpaca-gpt4-data-zh#500' \ + 'AI-ModelScope/alpaca-gpt4-data-en#500' \ + 'swift/self-cognition#500' \ + --torch_dtype bfloat16 \ + --num_train_epochs 1 \ + --per_device_train_batch_size 1 \ + --per_device_eval_batch_size 1 \ + --learning_rate 1e-4 \ + --lora_rank 8 \ + --lora_alpha 32 \ + --target_modules all-linear \ + --gradient_accumulation_steps 16 \ + --eval_steps 50 \ + --save_steps 50 \ + --save_total_limit 2 \ + --logging_steps 5 \ + --max_length 2048 \ + --output_dir output \ + --warmup_ratio 0.05 \ + --dataloader_num_workers 4 \ + --model_author swift \ + --model_name swift-robot +``` + +小贴士： +- 如果要使用自定义数据集进行训练，你可以参考[这里](https://swift.readthedocs.io/zh-cn/latest/Customization/Custom-dataset.html)组织数据集格式，并指定`--dataset `。 +- `--model_author`和`--model_name`参数只有当数据集中包含`swift/self-cognition`时才生效。 +- 如果要使用其他模型进行训练，你只需要修改`--model `即可。 +- 默认使用**ModelScope**进行模型和数据集的下载。如果要使用HuggingFace，指定`--use_hf true`即可。 + +训练完成后，使用以下命令对训练后的权重进行推理： +- 这里的`--adapters`需要替换成训练生成的last checkpoint文件夹。由于adapters文件夹中包含了训练的参数文件`args.json`，因此不需要额外指定`--model`，`--system`，swift会自动读取这些参数。如果要关闭此行为，可以设置`--load_args false`。 + +```shell +# 使用交互式命令行进行推理 +CUDA_VISIBLE_DEVICES=0 \ +swift infer \ + --adapters output/vx-xxx/checkpoint-xxx \ + --stream true \ + --temperature 0 \ + --max_new_tokens 2048 + +# merge-lora并使用vLLM进行推理加速 +CUDA_VISIBLE_DEVICES=0 \ +swift infer \ + --adapters output/vx-xxx/checkpoint-xxx \ + --stream true \ + --merge_lora true \ + --infer_backend vllm \ + --vllm_max_model_len 8192 \ + --temperature 0 \ + --max_new_tokens 2048 +``` + +最后，使用以下命令将模型推送到ModelScope： +```shell +CUDA_VISIBLE_DEVICES=0 \ +swift export \ + --adapters output/vx-xxx/checkpoint-xxx \ + --push_to_hub true \ + --hub_model_id '' \ + --hub_token '' \ + --use_hf false +``` + +### Web-UI + +Web-UI是基于gradio界面技术的**零门槛**训练、部署界面方案，具体可以查看[这里](https://swift.readthedocs.io/zh-cn/latest/GetStarted/Web-UI.html)。 + +```shell +swift web-ui +``` +![image.png](./docs/resources/web-ui.jpg) + +### 使用Python +ms-swift也支持使用python的方式进行训练和推理。下面给出训练和推理的**伪代码**，具体可以查看[这里](https://github.com/modelscope/ms-swift/blob/main/examples/notebook/qwen2_5-self-cognition/self-cognition-sft.ipynb)。 + +训练： +```python +from peft import LoraConfig, get_peft_model +from swift import get_model_processor, get_template, load_dataset, EncodePreprocessor +from swift.trainers import Seq2SeqTrainer, Seq2SeqTrainingArguments +# 获取模型和template，并加入可训练的LoRA模块 +model, tokenizer = get_model_processor(model_id_or_path, ...) +template = get_template(tokenizer, ...) +lora_config = LoraConfig(...) +model = get_peft_model(model, lora_config) + +# 下载并载入数据集，并将文本encode成tokens +train_dataset, val_dataset = load_dataset(dataset_id_or_path, ...) +train_dataset = EncodePreprocessor(template=template)(train_dataset, num_proc=num_proc) +val_dataset = EncodePreprocessor(template=template)(val_dataset, num_proc=num_proc) + +# 进行训练 +training_args = Seq2SeqTrainingArguments(...) +trainer = Seq2SeqTrainer( + model=model, + args=training_args, + template=template, + train_dataset=train_dataset, + eval_dataset=val_dataset, +) +trainer.train() +``` + +推理： +```python +from swift import TransformersEngine, InferRequest, RequestConfig +# 使用原生 transformers 引擎进行推理 +engine = TransformersEngine(model_id_or_path, adapters=[lora_checkpoint]) +infer_request = InferRequest(messages=[{'role': 'user', 'content': 'who are you?'}]) +request_config = RequestConfig(max_tokens=max_new_tokens, temperature=temperature) + +resp_list = engine.infer([infer_request], request_config) +print(f'response: {resp_list[0].choices[0].message.content}') +``` + +## ✨ 如何使用 + +这里给出使用ms-swift进行训练到部署的最简示例，具体可以查看[examples](https://github.com/modelscope/ms-swift/tree/main/examples)。 + +- 若想使用其他模型或者数据集（含多模态模型和数据集），你只需要修改`--model`指定对应模型的id或者path，修改`--dataset`指定对应数据集的id或者path即可。 +- 默认使用ModelScope进行模型和数据集的下载。如果要使用HuggingFace，指定`--use_hf true`即可。 + +| 常用链接 | +| ------ | +| [🔥命令行参数](https://swift.readthedocs.io/zh-cn/latest/Instruction/Command-line-parameters.html) | +| [Megatron-SWIFT](https://swift.readthedocs.io/zh-cn/latest/Megatron-SWIFT/Quick-start.html) | +| [GRPO](https://swift.readthedocs.io/zh-cn/latest/Instruction/GRPO/GetStarted/GRPO.html) | +| [支持的模型和数据集](https://swift.readthedocs.io/zh-cn/latest/Instruction/Supported-models-and-datasets.html) | +| [自定义模型](https://swift.readthedocs.io/zh-cn/latest/Customization/Custom-model.html), [🔥自定义数据集](https://swift.readthedocs.io/zh-cn/latest/Customization/Custom-dataset.html) | +| [大模型教程](https://github.com/modelscope/modelscope-classroom/tree/main/LLM-tutorial) | + +### 训练 +支持的训练方法： + +| 方法 | 全参数 | LoRA | QLoRA | Deepspeed | 多机 | 多模态 | +| ------ | ------ |---------------------------------------------------------------------------------------------| ----- | ------ | ------ |----------------------------------------------------------------------------------------------| +| [预训练](https://github.com/modelscope/ms-swift/blob/main/examples/train/pretrain) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [指令监督微调](https://github.com/modelscope/ms-swift/blob/main/examples/train/lora_sft.sh) | [✅](https://github.com/modelscope/ms-swift/blob/main/examples/train/full/train.sh) | ✅ | [✅](https://github.com/modelscope/ms-swift/tree/main/examples/train/qlora) | [✅](https://github.com/modelscope/ms-swift/tree/main/examples/train/multi-gpu/deepspeed) | [✅](https://github.com/modelscope/ms-swift/tree/main/examples/train/multi-node) | [✅](https://github.com/modelscope/ms-swift/tree/main/examples/train/multimodal) | +| [GRPO](https://github.com/modelscope/ms-swift/blob/main/examples/train/grpo) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [GKD](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/gkd) | ✅ | ✅ | ✅ | ✅ | ✅ | [✅](https://github.com/modelscope/ms-swift/blob/main/examples/train/multimodal/rlhf/gkd) | +| [PPO](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/ppo) | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | +| [DPO](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/dpo) | ✅ | ✅ | ✅ | ✅ | ✅ | [✅](https://github.com/modelscope/ms-swift/blob/main/examples/train/multimodal/rlhf/dpo) | +| [KTO](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/kto.sh) | ✅ | ✅ | ✅ | ✅ | ✅ | [✅](https://github.com/modelscope/ms-swift/blob/main/examples/train/multimodal/rlhf/kto.sh) | +| [奖励模型](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/rm.sh) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [CPO](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/cpo.sh) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [SimPO](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/simpo.sh) | ✅ | ✅ | ✅ | ✅| ✅ | ✅ | +| [ORPO](https://github.com/modelscope/ms-swift/blob/main/examples/train/rlhf/orpo.sh) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [Embedding](https://github.com/modelscope/ms-swift/blob/main/examples/train/embedding) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [Reranker](https://github.com/modelscope/ms-swift/tree/main/examples/train/reranker) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | +| [序列分类](https://github.com/modelscope/ms-swift/blob/main/examples/train/seq_cls) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | + + +预训练： +```shell +# 8*A100 +NPROC_PER_NODE=8 \ +CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \ +swift pt \ + --model Qwen/Qwen3-4B-Base \ + --dataset swift/chinese-c4 \ + --streaming true \ + --tuner_type full \ + --deepspeed zero2 \ + --output_dir output \ + --max_steps 10000 \ + ... +``` + +微调： +```shell +CUDA_VISIBLE_DEVICES=0 swift sft \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --dataset AI-ModelScope/alpaca-gpt4-data-zh \ + --tuner_type lora \ + --output_dir output \ + ... +``` + +RLHF： +```shell +CUDA_VISIBLE_DEVICES=0 swift rlhf \ + --rlhf_type dpo \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --dataset hjh0119/shareAI-Llama3-DPO-zh-en-emoji \ + --tuner_type lora \ + --output_dir output \ + ... +``` + +### Megatron-SWIFT + +ms-swift支持使用Megatron并行技术加速训练，包括大规模集群训练和MoE模型训练。以下为支持的训练方法： + +| 方法 | 全参数 | LoRA | MoE | 多模态 | FP8 | +| ------ | ------ | ---- | ----- | ----- | ----- | +| 预训练 | ✅ | ✅| ✅ | ✅ | ✅ | +| [指令监督微调](https://github.com/modelscope/ms-swift/tree/main/examples/megatron) | ✅ | ✅| ✅ | ✅ | ✅ | +| [GRPO](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/grpo) | ✅ | ✅| ✅ | ✅ | ✅ | +| [GKD](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/rlhf/gkd) | ✅ | ✅| ✅ | ✅ | ✅ | +| [DPO](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/rlhf/dpo) | ✅ | ✅| ✅ | ✅ | ✅ | +| [KTO](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/rlhf/kto) | ✅ | ✅| ✅ | ✅ | ✅ | +| [RM](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/rlhf/rm) | ✅ | ✅| ✅ | ✅ | ✅ | +| [Embedding](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/embedding) | ✅ | ✅| ✅ | ✅ | ✅ | +| [Reranker](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/reranker) | ✅ | ✅| ✅ | ✅ | ✅ | +| [序列分类](https://github.com/modelscope/ms-swift/tree/main/examples/megatron/seq_cls) | ✅ | ✅| ✅ | ✅ | ✅ | + + +```shell +NPROC_PER_NODE=2 CUDA_VISIBLE_DEVICES=0,1 megatron sft \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --save_safetensors true \ + --dataset AI-ModelScope/alpaca-gpt4-data-zh \ + --tuner_type lora \ + --output_dir output \ + ... +``` + +### 强化学习 + +ms-swift支持丰富GRPO族算法： + +| 方法 | 全参数 | LoRA | 多模态 | 多机 | +| ------ | ------ | ---- | ----- | ----- | +| [GRPO](https://swift.readthedocs.io/zh-cn/latest/Instruction/GRPO/GetStarted/GRPO.html) | ✅ | ✅| ✅ | ✅ | +| [DAPO](https://swift.readthedocs.io/zh-cn/latest/Instruction/GRPO/AdvancedResearch/DAPO.html) | ✅ | ✅| ✅ | ✅ | +| [GSPO](https://swift.readthedocs.io/zh-cn/latest/Instruction/GRPO/AdvancedResearch/GSPO.html) | ✅ | ✅| ✅ | ✅ | +| [SAPO](https://swift.readthedocs.io/zh-cn/latest/Instruction/GRPO/AdvancedResearch/SAPO.html) | ✅ | ✅| ✅ | ✅ | +| [CISPO](https://swift.readthedocs.io/zh-cn/latest/Instruction/GRPO/AdvancedResearch/CISPO.html) | ✅ | ✅| ✅ | ✅ | +| [CHORD](https://swift.readthedocs.io/zh-cn/latest/Instruction/GRPO/AdvancedResearch/CHORD.html) | ✅ | ✅| ✅ | ✅ | +| [RLOO](https://swift.readthedocs.io/zh-cn/latest/Instruction/GRPO/AdvancedResearch/RLOO.html) | ✅ | ✅| ✅ | ✅ | +| [Reinforce++](https://swift.readthedocs.io/zh-cn/latest/Instruction/GRPO/AdvancedResearch/REINFORCEPP.html) | ✅ | ✅| ✅ | ✅ | + +```shell +CUDA_VISIBLE_DEVICES=0,1,2,3 NPROC_PER_NODE=4 \ +swift rlhf \ + --rlhf_type grpo \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --tuner_type lora \ + --use_vllm true \ + --vllm_mode colocate \ + --dataset AI-MO/NuminaMath-TIR#10000 \ + --output_dir output \ + ... +``` + +### 推理 +```shell +CUDA_VISIBLE_DEVICES=0 swift infer \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --stream true \ + --infer_backend transformers \ + --max_new_tokens 2048 +``` + +### 界面推理 +```shell +CUDA_VISIBLE_DEVICES=0 swift app \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --stream true \ + --infer_backend transformers \ + --max_new_tokens 2048 \ + --lang zh +``` + +### 部署 +```shell +CUDA_VISIBLE_DEVICES=0 swift deploy \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --infer_backend vllm +``` + +### 采样 +```shell +CUDA_VISIBLE_DEVICES=0 swift sample \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --sampler_engine transformers \ + --num_return_sequences 5 \ + --dataset AI-ModelScope/alpaca-gpt4-data-zh#5 +``` + +### 评测 +```shell +CUDA_VISIBLE_DEVICES=0 swift eval \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --infer_backend sglang \ + --eval_backend OpenCompass \ + --eval_dataset ARC_c +``` + +### 量化 +```shell +CUDA_VISIBLE_DEVICES=0 swift export \ + --model Qwen/Qwen3-4B-Instruct-2507 \ + --quant_method fp8 \ + --dataset AI-ModelScope/alpaca-gpt4-data-zh \ + --output_dir Qwen3-4B-Instruct-2507-FP8 +``` + +### 推送模型 +```shell +swift export \ + --model \ + --push_to_hub true \ + --hub_model_id '' \ + --hub_token '' +``` + + +## 🏛 License + +本框架使用[Apache License (Version 2.0)](https://github.com/modelscope/ms-swift/blob/master/LICENSE)进行许可。模型和数据集请查看原资源页面并遵守对应License。 + +## 📎 引用 + +```bibtex +@misc{zhao2024swiftascalablelightweightinfrastructure, + title={SWIFT:A Scalable lightWeight Infrastructure for Fine-Tuning}, + author={Yuze Zhao and Jintao Huang and Jinghan Hu and Xingjun Wang and Yunlin Mao and Daoze Zhang and Zeyinzi Jiang and Zhikai Wu and Baole Ai and Ang Wang and Wenmeng Zhou and Yingda Chen}, + year={2024}, + eprint={2408.05517}, + archivePrefix={arXiv}, + primaryClass={cs.CL}, + url={https://arxiv.org/abs/2408.05517}, +} +``` + +## Star History + +[![Star History Chart](https://api.star-history.com/svg?repos=modelscope/ms-swift&type=Date)](https://star-history.com/#modelscope/ms-swift&Date) diff --git a/asset/banner.png b/asset/banner.png new file mode 100644 index 0000000..cf20b8d Binary files /dev/null and b/asset/banner.png differ diff --git a/asset/ding.png b/asset/ding.png new file mode 100644 index 0000000..36b4a2c Binary files /dev/null and b/asset/ding.png differ diff --git a/asset/discord_qr.jpg b/asset/discord_qr.jpg new file mode 100644 index 0000000..4f2dabc Binary files /dev/null and b/asset/discord_qr.jpg differ diff --git a/asset/wechat.png b/asset/wechat.png new file mode 100644 index 0000000..3cfc075 Binary files /dev/null and b/asset/wechat.png differ diff --git a/docs/Makefile b/docs/Makefile new file mode 100644 index 0000000..d0c3cbf --- /dev/null +++ b/docs/Makefile @@ -0,0 +1,20 @@ +# Minimal makefile for Sphinx documentation +# + +# You can set these variables from the command line, and also +# from the environment for the first two. +SPHINXOPTS ?= +SPHINXBUILD ?= sphinx-build +SOURCEDIR = source +BUILDDIR = build + +# Put it first so that "make" without argument is like "make help". +help: + @$(SPHINXBUILD) -M help "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O) + +.PHONY: help Makefile + +# Catch-all target: route all unknown targets to Sphinx using the new +# "make mode" option. $(O) is meant as a shortcut for $(SPHINXOPTS). +%: Makefile + @$(SPHINXBUILD) -M $@ "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O) diff --git a/docs/README.md b/docs/README.md new file mode 100644 index 0000000..8ccd292 --- /dev/null +++ b/docs/README.md @@ -0,0 +1,37 @@ +## maintain docs +1. build docs + ```shell + # in root directory: + make docs + ``` + +2. doc string format + + We adopt the google style docstring format as the standard, please refer to the following documents. + 1. Google Python style guide docstring [link](http://google.github.io/styleguide/pyguide.html#381-docstrings) + 2. Google docstring example [link](https://sphinxcontrib-napoleon.readthedocs.io/en/latest/example_google.html) + 3. sample：torch.nn.modules.conv [link](https://pytorch.org/docs/stable/_modules/torch/nn/modules/conv.html#Conv1d) + 4. load function as an example： + + ```python + def load(file, file_format=None, **kwargs): + """Load data from json/yaml/pickle files. + + This method provides a unified api for loading data from serialized files. + + Args: + file (str or :obj:`Path` or file-like object): Filename or a file-like + object. + file_format (str, optional): If not specified, the file format will be + inferred from the file extension, otherwise use the specified one. + Currently supported formats include "json", "yaml/yml". + + Examples: + >>> load('/path/of/your/file') # file is stored in disk + >>> load('https://path/of/your/file') # file is stored on internet + >>> load('oss://path/of/your/file') # file is stored in petrel + + Returns: + The content from the file. + """ + ``` diff --git a/docs/make.bat b/docs/make.bat new file mode 100644 index 0000000..9534b01 --- /dev/null +++ b/docs/make.bat @@ -0,0 +1,35 @@ +@ECHO OFF + +pushd %~dp0 + +REM Command file for Sphinx documentation + +if "%SPHINXBUILD%" == "" ( + set SPHINXBUILD=sphinx-build +) +set SOURCEDIR=source +set BUILDDIR=build + +if "%1" == "" goto help + +%SPHINXBUILD% >NUL 2>NUL +if errorlevel 9009 ( + echo. + echo.The 'sphinx-build' command was not found. Make sure you have Sphinx + echo.installed, then set the SPHINXBUILD environment variable to point + echo.to the full path of the 'sphinx-build' executable. Alternatively you + echo.may add the Sphinx directory to PATH. + echo. + echo.If you don't have Sphinx installed, grab it from + echo.http://sphinx-doc.org/ + exit /b 1 +) + +%SPHINXBUILD% -M %1 %SOURCEDIR% %BUILDDIR% %SPHINXOPTS% %O% +goto end + +:help +%SPHINXBUILD% -M help %SOURCEDIR% %BUILDDIR% %SPHINXOPTS% %O% + +:end +popd diff --git a/docs/resources/asyncengine.png b/docs/resources/asyncengine.png new file mode 100644 index 0000000..0fbc373 Binary files /dev/null and b/docs/resources/asyncengine.png differ diff --git a/docs/resources/deepeyes.png b/docs/resources/deepeyes.png new file mode 100644 index 0000000..1a9f801 Binary files /dev/null and b/docs/resources/deepeyes.png differ diff --git a/docs/resources/deepseek_v4/infer_result.png b/docs/resources/deepseek_v4/infer_result.png new file mode 100644 index 0000000..5b1908e Binary files /dev/null and b/docs/resources/deepseek_v4/infer_result.png differ diff --git a/docs/resources/deepseek_v4/loss.png b/docs/resources/deepseek_v4/loss.png new file mode 100644 index 0000000..0757624 Binary files /dev/null and b/docs/resources/deepseek_v4/loss.png differ diff --git a/docs/resources/deepseek_v4/memory.png b/docs/resources/deepseek_v4/memory.png new file mode 100644 index 0000000..f76cf7a Binary files /dev/null and b/docs/resources/deepseek_v4/memory.png differ diff --git a/docs/resources/deepseek_v4/precision.png b/docs/resources/deepseek_v4/precision.png new file mode 100644 index 0000000..734d736 Binary files /dev/null and b/docs/resources/deepseek_v4/precision.png differ diff --git a/docs/resources/dpo_data.png b/docs/resources/dpo_data.png new file mode 100644 index 0000000..6560967 Binary files /dev/null and b/docs/resources/dpo_data.png differ diff --git a/docs/resources/grpo.png b/docs/resources/grpo.png new file mode 100644 index 0000000..9d076a7 Binary files /dev/null and b/docs/resources/grpo.png differ diff --git a/docs/resources/grpo_clevr_count.png b/docs/resources/grpo_clevr_count.png new file mode 100644 index 0000000..ba331e3 Binary files /dev/null and b/docs/resources/grpo_clevr_count.png differ diff --git a/docs/resources/grpo_code.png b/docs/resources/grpo_code.png new file mode 100644 index 0000000..32c6953 Binary files /dev/null and b/docs/resources/grpo_code.png differ diff --git a/docs/resources/grpo_countdown.png b/docs/resources/grpo_countdown.png new file mode 100644 index 0000000..3126e99 Binary files /dev/null and b/docs/resources/grpo_countdown.png differ diff --git a/docs/resources/grpo_countdown_1.png b/docs/resources/grpo_countdown_1.png new file mode 100644 index 0000000..2bb4a87 Binary files /dev/null and b/docs/resources/grpo_countdown_1.png differ diff --git a/docs/resources/grpo_geoqa.png b/docs/resources/grpo_geoqa.png new file mode 100644 index 0000000..c600b0a Binary files /dev/null and b/docs/resources/grpo_geoqa.png differ diff --git a/docs/resources/grpo_multi_turn.png b/docs/resources/grpo_multi_turn.png new file mode 100644 index 0000000..98ec478 Binary files /dev/null and b/docs/resources/grpo_multi_turn.png differ diff --git a/docs/resources/grpo_openr1_multimodal.png b/docs/resources/grpo_openr1_multimodal.png new file mode 100644 index 0000000..7d3fac7 Binary files /dev/null and b/docs/resources/grpo_openr1_multimodal.png differ diff --git a/docs/resources/kto_data.png b/docs/resources/kto_data.png new file mode 100644 index 0000000..802ed2b Binary files /dev/null and b/docs/resources/kto_data.png differ diff --git a/docs/resources/multiturn_pipeline.png b/docs/resources/multiturn_pipeline.png new file mode 100644 index 0000000..f3511ab Binary files /dev/null and b/docs/resources/multiturn_pipeline.png differ diff --git a/docs/resources/real.png b/docs/resources/real.png new file mode 100644 index 0000000..ab1951d Binary files /dev/null and b/docs/resources/real.png differ diff --git a/docs/resources/real_framework.png b/docs/resources/real_framework.png new file mode 100644 index 0000000..d86d374 Binary files /dev/null and b/docs/resources/real_framework.png differ diff --git a/docs/resources/sapo_tau.png b/docs/resources/sapo_tau.png new file mode 100644 index 0000000..6b0f911 Binary files /dev/null and b/docs/resources/sapo_tau.png differ diff --git a/docs/resources/treepo.png b/docs/resources/treepo.png new file mode 100644 index 0000000..d5bd508 Binary files /dev/null and b/docs/resources/treepo.png differ diff --git a/docs/resources/web-ui-en.jpg b/docs/resources/web-ui-en.jpg new file mode 100644 index 0000000..daf180a Binary files /dev/null and b/docs/resources/web-ui-en.jpg differ diff --git a/docs/resources/web-ui.jpg b/docs/resources/web-ui.jpg new file mode 100644 index 0000000..15b45c7 Binary files /dev/null and b/docs/resources/web-ui.jpg differ diff --git a/docs/resources/wechat/grpo.png b/docs/resources/wechat/grpo.png new file mode 100644 index 0000000..4b394eb Binary files /dev/null and b/docs/resources/wechat/grpo.png differ diff --git a/docs/resources/wechat/megatron.png b/docs/resources/wechat/megatron.png new file mode 100644 index 0000000..46598f0 Binary files /dev/null and b/docs/resources/wechat/megatron.png differ diff --git a/docs/resources/wechat/npu.png b/docs/resources/wechat/npu.png new file mode 100644 index 0000000..64964c0 Binary files /dev/null and b/docs/resources/wechat/npu.png differ diff --git a/docs/source/.readthedocs.yaml b/docs/source/.readthedocs.yaml new file mode 100644 index 0000000..c9707f4 --- /dev/null +++ b/docs/source/.readthedocs.yaml @@ -0,0 +1,29 @@ +# .readthedocs.yaml +# Read the Docs configuration file +# See https://docs.readthedocs.io/en/stable/config-file/v2.html for details + +# Required +version: 2 + +# Set the OS, Python version and other tools you might need +build: + os: ubuntu-22.04 + tools: + python: "3.10" + +# Build documentation in the "docs/" directory with Sphinx +sphinx: + configuration: docs/source/conf.py + +# Optionally build your docs in additional formats such as PDF and ePub +# formats: +# - pdf +# - epub + +# Optional but recommended, declare the Python requirements required +# to build your documentation +# See https://docs.readthedocs.io/en/stable/guides/reproducible-builds.html +python: + install: + - requirements: requirements/docs.txt + - requirements: requirements/framework.txt diff --git a/docs/source/BestPractices/AMD-support.md b/docs/source/BestPractices/AMD-support.md new file mode 100644 index 0000000..34a1557 --- /dev/null +++ b/docs/source/BestPractices/AMD-support.md @@ -0,0 +1,321 @@ +# AMD GPU 支持 + +## 1. 环境配置 +### 1.1 基础环境 +拉取适配了 AMD ROCm 生态的 ms-swift 镜像，并参照以下命令启动容器。 + +如果用户需要运行更新版本的 ms-swift，可以使用 pip 升级或基于源码安装更新（建议安装时添加 `--no-deps` 选项以避免自动升级其他依赖可能引起的问题）。 + +```bash +IMAGE_NAME=amdagi/modelscope:ubuntu22.04-rocm7.2.0-py312-torch2.10.0-vllm0.18.1-modelscope1.35.1-swift4.1.0 +docker pull ${IMAGE_NAME} + +CONTAINER_NAME=swift_test +docker run -it --network=host --ipc=host --privileged --group-add video \ + --device=/dev/dri --device=/dev/kfd \ + --shm-size 512G --ulimit memlock=-1 \ + --security-opt seccomp=unconfined --cap-add SYS_PTRACE \ + --name ${CONTAINER_NAME} \ + ${IMAGE_NAME} \ + /bin/bash +``` + +### 1.2 环境检查 + +- 确认 container 环境中 pytorch 正确识别 AMD GPU。 + +```bash +python -c "import torch;print(torch.cuda.is_available())" # output: True +``` + +- 检查 GPU 的拓扑连接及 NUMA：`rocm-smi --showtopo` + +``` +============================ ROCm System Management Interface ============================ +WARNING: AMD GPU device(s) is/are in a low-power state. Check power control/runtime_status + +================================ Weight between two GPUs ================================= + GPU0 GPU1 GPU2 GPU3 GPU4 GPU5 GPU6 GPU7 +GPU0 0 15 15 15 15 15 15 15 +GPU1 15 0 15 15 15 15 15 15 +GPU2 15 15 0 15 15 15 15 15 +GPU3 15 15 15 0 15 15 15 15 +GPU4 15 15 15 15 0 15 15 15 +GPU5 15 15 15 15 15 0 15 15 +GPU6 15 15 15 15 15 15 0 15 +GPU7 15 15 15 15 15 15 15 0 + +================================= Hops between two GPUs ================================== + GPU0 GPU1 GPU2 GPU3 GPU4 GPU5 GPU6 GPU7 +GPU0 0 1 1 1 1 1 1 1 +GPU1 1 0 1 1 1 1 1 1 +GPU2 1 1 0 1 1 1 1 1 +GPU3 1 1 1 0 1 1 1 1 +GPU4 1 1 1 1 0 1 1 1 +GPU5 1 1 1 1 1 0 1 1 +GPU6 1 1 1 1 1 1 0 1 +GPU7 1 1 1 1 1 1 1 0 + +=============================== Link Type between two GPUs =============================== + GPU0 GPU1 GPU2 GPU3 GPU4 GPU5 GPU6 GPU7 +GPU0 0 XGMI XGMI XGMI XGMI XGMI XGMI XGMI +GPU1 XGMI 0 XGMI XGMI XGMI XGMI XGMI XGMI +GPU2 XGMI XGMI 0 XGMI XGMI XGMI XGMI XGMI +GPU3 XGMI XGMI XGMI 0 XGMI XGMI XGMI XGMI +GPU4 XGMI XGMI XGMI XGMI 0 XGMI XGMI XGMI +GPU5 XGMI XGMI XGMI XGMI XGMI 0 XGMI XGMI +GPU6 XGMI XGMI XGMI XGMI XGMI XGMI 0 XGMI +GPU7 XGMI XGMI XGMI XGMI XGMI XGMI XGMI 0 + +======================================= Numa Nodes ======================================= +GPU[0] : (Topology) Numa Node: 0 +GPU[0] : (Topology) Numa Affinity: 0 +GPU[1] : (Topology) Numa Node: 0 +GPU[1] : (Topology) Numa Affinity: 0 +GPU[2] : (Topology) Numa Node: 0 +GPU[2] : (Topology) Numa Affinity: 0 +GPU[3] : (Topology) Numa Node: 0 +GPU[3] : (Topology) Numa Affinity: 0 +GPU[4] : (Topology) Numa Node: 1 +GPU[4] : (Topology) Numa Affinity: 1 +GPU[5] : (Topology) Numa Node: 1 +GPU[5] : (Topology) Numa Affinity: 1 +GPU[6] : (Topology) Numa Node: 1 +GPU[6] : (Topology) Numa Affinity: 1 +GPU[7] : (Topology) Numa Node: 1 +GPU[7] : (Topology) Numa Affinity: 1 +================================== End of ROCm SMI Log =================================== +``` + +- 查看 GPU 利用率及显存占用等信息（`rocm-smi` 或者 `rocm-smi -u --showmeminfo vram`） + +``` +# output of 'rocm-smi' +============================================ ROCm System Management Interface ============================================ +====================================================== Concise Info ====================================================== +Device Node IDs Temp Power Partitions SCLK MCLK Fan Perf PwrCap VRAM% GPU% + (DID, GUID) (Junction) (Socket) (Mem, Compute, ID) +========================================================================================================================== +0 2 0x74a2, 1017 43.0°C 155.0W NPS1, SPX, 0 94Mhz 900Mhz 0% auto 650.0W 0% 0% +1 3 0x74a2, 47713 41.0°C 155.0W NPS1, SPX, 0 91Mhz 900Mhz 0% auto 650.0W 0% 0% +2 4 0x74a2, 37449 45.0°C 159.0W NPS1, SPX, 0 95Mhz 900Mhz 0% auto 650.0W 0% 0% +3 5 0x74a2, 11217 41.0°C 155.0W NPS1, SPX, 0 95Mhz 900Mhz 0% auto 650.0W 0% 0% +4 6 0x74a2, 41880 44.0°C 160.0W NPS1, SPX, 0 91Mhz 900Mhz 0% auto 650.0W 0% 0% +5 7 0x74a2, 6656 42.0°C 157.0W NPS1, SPX, 0 95Mhz 900Mhz 0% auto 650.0W 0% 0% +6 8 0x74a2, 12840 45.0°C 160.0W NPS1, SPX, 0 96Mhz 900Mhz 0% auto 650.0W 0% 0% +7 9 0x74a2, 35760 43.0°C 158.0W NPS1, SPX, 0 107Mhz 900Mhz 0% auto 650.0W 0% 0% +========================================================================================================================== +================================================== End of ROCm SMI Log =================================================== + +# output of 'rocm-smi -u --showmeminfo vram' +============================ ROCm System Management Interface ============================ +=================================== % time GPU is busy =================================== +GPU[0] : GPU use (%): 0 +GPU[0] : GFX Activity: 3862538534 +GPU[1] : GPU use (%): 0 +GPU[1] : GFX Activity: 4053246251 +GPU[2] : GPU use (%): 0 +GPU[2] : GFX Activity: 3114103535 +GPU[3] : GPU use (%): 0 +GPU[3] : GFX Activity: 4026776444 +GPU[4] : GPU use (%): 0 +GPU[4] : GFX Activity: 1224255679 +GPU[5] : GPU use (%): 0 +GPU[5] : GFX Activity: 1191191242 +GPU[6] : GPU use (%): 0 +GPU[6] : GFX Activity: 1184652679 +GPU[7] : GPU use (%): 0 +GPU[7] : GFX Activity: 2145209382 +========================================================================================== +================================== Memory Usage (Bytes) ================================== +GPU[0] : VRAM Total Memory (B): 206141652992 +GPU[0] : VRAM Total Used Memory (B): 297611264 +GPU[1] : VRAM Total Memory (B): 206141652992 +GPU[1] : VRAM Total Used Memory (B): 297623552 +GPU[2] : VRAM Total Memory (B): 206141652992 +GPU[2] : VRAM Total Used Memory (B): 297623552 +GPU[3] : VRAM Total Memory (B): 206141652992 +GPU[3] : VRAM Total Used Memory (B): 297623552 +GPU[4] : VRAM Total Memory (B): 206141652992 +GPU[4] : VRAM Total Used Memory (B): 297623552 +GPU[5] : VRAM Total Memory (B): 206141652992 +GPU[5] : VRAM Total Used Memory (B): 297623552 +GPU[6] : VRAM Total Memory (B): 206141652992 +GPU[6] : VRAM Total Used Memory (B): 297623552 +GPU[7] : VRAM Total Memory (B): 206141652992 +GPU[7] : VRAM Total Used Memory (B): 297623552 +========================================================================================== +================================== End of ROCm SMI Log =================================== +``` + +## 2. 运行示例 + +### 2.1 使用 Megatron-Swift 全量微调 Qwen3.5 模型 +AMD GPU 显存相对较大，因此可以通过同时对以下参数进行联合调优，以提升训练吞吐性能。 +- 并行度调优（TP/PP/EP等）：GPU 单卡显存较大使得用户可以尽可能减小并行度切分带来的通信开销（优先级 PP/EP > TP） +- 显存允许的情况下关闭 optimizer cpu offload：设置 `--optimizer_cpu_offload false` +- 显存允许的情况下调整 activation/gradient checkpointing：设置 `--recompute_granularity none`，或者 `--recompute_granularity selective` 配合 `--recompute_modules` 进行细粒度的控制 +- 对于 MoE 模型，建议设置 `export NVTE_USE_GROUPED_GEMM_TRITON=1` 以使用 triton 实现的 grouped gemm kernel +- 对于带有 GatedDeltaNet 结构的模型，建议设置 `USE_MCORE_GDN=1` 使用 mcore 的实现版本 +- 为避免在某些 AMD GPU 上可能出现的问题，保证性能更稳定，建议设置 `export HSA_NO_SCRATCH_RECLAIM=1` + +单机训练： + +```bash +export HSA_NO_SCRATCH_RECLAIM=1 +export NVTE_USE_GROUPED_GEMM_TRITON=1 + +output_dir=${PWD}/megatron_output/Qwen3.5-35B-A3B +mkdir -p ${output_dir} +current_time=$(date "+%Y.%m.%d-%H.%M.%S") +log_file=${output_dir}/"1node_full_megatron_Qwen3.5-35B-A3B_${current_time}.log" + +PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True' \ +NPROC_PER_NODE=8 \ +MAX_PIXELS=1003520 \ +VIDEO_MAX_PIXELS=50176 \ +FPS_MAX_FRAMES=12 \ +SKIP_MULTIMODAL_MTP_VALIDATION=1 \ +USE_MCORE_GDN=1 \ +megatron sft \ + --model Qwen/Qwen3.5-35B-A3B \ + --dataset 'AI-ModelScope/LongAlpaca-12k' \ + --save_safetensors true \ + --load_from_cache_file true \ + --tuner_type full \ + --add_non_thinking_prefix true \ + --split_dataset_ratio 0.01 \ + --tensor_model_parallel_size 1 \ + --pipeline_model_parallel_size 1 \ + --expert_model_parallel_size 8 \ + --sequence_parallel true \ + --moe_permute_fusion true \ + --moe_grouped_gemm true \ + --moe_shared_expert_overlap true \ + --moe_aux_loss_coeff 1e-6 \ + --moe_expert_capacity_factor 2 \ + --micro_batch_size 1 \ + --global_batch_size 8 \ + --recompute_granularity selective \ + --recompute_modules core_attn mlp moe \ + --gradient_accumulation_fusion false \ + --num_train_epochs 500 \ + --group_by_length true \ + --finetune true \ + --freeze_llm false \ + --freeze_vit false \ + --freeze_aligner false \ + --cross_entropy_loss_fusion true \ + --lr 1e-5 \ + --lr_warmup_fraction 0.05 \ + --min_lr 1e-6 \ + --weight_decay 0.1 \ + --adam_beta2 0.95 \ + --eval_steps 500 \ + --save_steps 500 \ + --save_total_limit 10 \ + --logging_steps 1 \ + --max_length 16384 \ + --dataloader_num_workers 8 \ + --dataset_num_proc 8 \ + --no_save_optim true \ + --no_save_rng true \ + --optimizer_cpu_offload false \ + --attention_backend flash \ + --padding_free false \ + --output_dir ${output_dir} \ + 2>&1 | tee ${log_file} +``` + +多机训练： + +```bash +export NNODES=2 # 此处以 2 节点为例 +export NODE_RANK=0 # 主节点设置为 0，从节点设置为 1 +export MASTER_ADDR= # 根据主节点 ip 设置 +export MASTER_PORT=29500 # 设置通信端口 +export NCCL_SOCKET_IFNAME=ens50f1np1 # 根据机器实际通信网口名设置，可通过 ifconfig 查看 +export GLOO_SOCKET_IFNAME=ens50f1np1 # 根据机器实际通信网口名设置，可通过 ifconfig 查看 +export NCCL_IB_HCA=mlx5_0,mlx5_1,mlx5_2,mlx5_3 # 根据实际IB网卡名设置，可通过 ibv_devices 查看 +export NCCL_IB_GID_INDEX=3 + +# 训练脚本主体：参照单机训练脚本 +... +``` + +### 2.2 使用 Megatron-Swift 对 Qwen3.5 模型做强化学习训练 + +```bash +# 单机训练样例 +export HSA_NO_SCRATCH_RECLAIM=1 +export NVTE_USE_GROUPED_GEMM_TRITON=1 + +SYSTEM_PROMPT="""You are a helpful math assistant. Solve the problem step by step and put your final answer within \\boxed{}.""" + +CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \ +NPROC_PER_NODE=8 \ +PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True' \ +megatron rlhf \ + --rlhf_type grpo \ + --model Qwen/Qwen3.5-35B-A3B \ + --save_safetensors true \ + --enable_thinking false \ + --merge_lora true \ + --context_parallel_size 1 \ + --tensor_model_parallel_size 1 \ + --expert_model_parallel_size 8 \ + --pipeline_model_parallel_size 1 \ + --moe_permute_fusion true \ + --dataset open-r1/DAPO-Math-17k-Processed \ + --system "$SYSTEM_PROMPT" \ + --num_train_epochs 1 \ + --global_batch_size 64 \ + --micro_batch_size 1 \ + --steps_per_generation 2 \ + --num_generations 8 \ + --reward_funcs accuracy \ + --use_vllm true \ + --vllm_mode colocate \ + --vllm_gpu_memory_utilization 0.5 \ + --vllm_tensor_parallel_size 2 \ + --vllm_max_model_len 9192 \ + --max_length 1000 \ + --max_completion_length 8192 \ + --tuner_type lora \ + --target_modules all-linear \ + --lr 5e-5 \ + --bf16 true \ + --beta 0.00 \ + --epsilon 0.2 \ + --epsilon_high 0.28 \ + --dynamic_sample false \ + --overlong_filter true \ + --loss_type grpo \ + --sleep_level 1 \ + --offload_model true \ + --offload_bridge false \ + --offload_optimizer true \ + --logging_steps 1 \ + --recompute_granularity none \ + --gradient_accumulation_fusion false \ + --finetune \ + --dataloader_num_workers 8 \ + --dataset_num_proc 8 \ + --no_save_optim \ + --no_save_rng \ + --save_steps 20 \ + --attention_backend flash \ + --moe_expert_capacity_factor 2 \ + --temperature 1.0 \ + --padding_free false \ + --sequence_parallel true \ + --log_completions true \ + --report_to tensorboard +``` + +## 已知问题 +- 强化学习训练： + - 在强化学习训练中，如果使用 vLLM 作为推理引擎，需要 vLLM>=0.11.0。建议使用 ROCm7.0 或者我们提供的镜像以避免出现 sleep mode memory leak 问题。 + - 在使用 [Ray Megatron](../../source/Instruction/Ray.md) 而非 `torchrun` 的方式进行多 GPU/Node 训练时，不设置 `CUDA_VISIBLE_DEVICES`/`HIP_VISIBLE_DEVICES`等，以避免冲突问题。 +- MoE 模型训练： + - MoE 模型建议增加环境变量 `NVTE_USE_GROUPED_GEMM_TRITON=1` 和 `--gradient_accumulation_fusion false` 以避免偶发的GPU卡死问题。 diff --git a/docs/source/BestPractices/Elastic.md b/docs/source/BestPractices/Elastic.md new file mode 100644 index 0000000..3a78239 --- /dev/null +++ b/docs/source/BestPractices/Elastic.md @@ -0,0 +1,211 @@ +# Elastic + + + +## 安装依赖 + +集群部署K8S,并在集群中部署DLrover,[DLRover](https://github.com/intelligent-machine-learning/dlrover), +`pip install dlrover && pip install tornado && pip install kubernetes && pip install ms-swift` + +经过反复测试验证的训练镜像中的其它依赖以及版本： +deepspeed 0.16.5（需参考https://github.com/deepspeedai/DeepSpeed/pull/7585/files 修复universal checkpoint 相关问题） +pytorch 2.6.0 + + +## 如何启动 + +通过在`--callbacks`中添加`deepspeed_elastic`（可选`graceful_exit`）启用弹性训练，并配置DeepSpeed弹性参数。 +命令组成=dlrover-run +dlrover 命令参数+swift 启动命令 +swift参数,dlrover-run除自定义的参数外，其他参数与torchrun一致； +dlrover-run 参数如下: +``` +usage: dlrover-run [-h] [--nnodes NNODES] [--nproc-per-node NPROC_PER_NODE] + [--rdzv-backend RDZV_BACKEND] [--rdzv-endpoint RDZV_ENDPOINT] [--rdzv-id RDZV_ID] + [--rdzv-conf RDZV_CONF] [--standalone] [--max-restarts MAX_RESTARTS] + [--monitor-interval MONITOR_INTERVAL] [--start-method {spawn,fork,forkserver}] + [--role ROLE] [-m] [--no-python] [--run-path] [--log-dir LOG_DIR] [-r REDIRECTS] + [-t TEE] [--local-ranks-filter LOCAL_RANKS_FILTER] [--node-rank NODE_RANK] + [--master-addr MASTER_ADDR] [--master-port MASTER_PORT] [--local-addr LOCAL_ADDR] + [--logs-specs LOGS_SPECS] [--precheck {0,1,2}] [--node_unit NODE_UNIT] + [--auto_config] [--auto_tunning] [--exclude-straggler] [--save_at_breakpoint] + [--accelerator {nvidia.com/gpu,ascend-npu}] [--training_port TRAINING_PORT] + [--switchbox-check] [--box-pairs PAIR [PAIR ...]] [--min-bandwidth MIN_BANDWIDTH] + [--min-channels MIN_CHANNELS] [--numa-affinity] [--network-check] + [--comm-perf-test] [--ucp_device_type UCP_DEVICE_TYPE] + training_script + +``` +在弹性训练中我们需要关注的参数为： + +--nnodes NNODES Number of nodes, or the range of nodes in form + :. + +--nproc-per-node NPROC_PER_NODE Number of processes per node. +示例: + +```bash +model=your model path +dataset=your dataset +output= your output dir +export CUDA_VISIBLE_DEVICES=0 根据实际使用的GPU情况设置 +deepspeed_config_or_type=deepspeed类型或者配置文件的路径，如 zero1 或者/xxx/ms-swift/swift/llm/ds_config/zero1.json + +dlrover-run --nnodes 1:$NODE_NUM --nproc_per_node=1 \ +/opt/conda/lib/python3.10/site-packages/swift/cli/sft.py --model $model \ +--model_type qwen3 \ +--tuner_type lora \ +--torch_dtype bfloat16 \ +--dataset $dataset \ +--num_train_epochs 4 \ +--per_device_train_batch_size 1 \ +--per_device_eval_batch_size 1 \ +--learning_rate 5e-7 \ +--gradient_accumulation_steps 8 \ +--eval_steps 500 \ +--save_steps 10 \ +--save_total_limit 20 \ +--logging_steps 1 \ +--output_dir $output \ +--warmup_ratio 0.01 \ +--dataloader_num_workers 4 \ +--temperature 1.0 \ +--system 'You are a helpful assistant.' \ +--lora_rank 8 \ +--lora_alpha 32 \ +--target_modules all-linear \ +--dataset_num_proc 1 \ +--use_flash_ckpt true \ +--callbacks deepspeed_elastic graceful_exit \ +--deepspeed $deepspeed_config_or_type \ +``` + +## 配置文件示例 +默认情况下的zero1为以下示例配置, + +```json +{ + "fp16": { + "enabled": "auto", + "loss_scale": 0, + "loss_scale_window": 1000, + "initial_scale_power": 16, + "hysteresis": 2, + "min_loss_scale": 1 + }, + + "bf16": { + "enabled": "auto" + }, + + "zero_optimization": { + "stage": 1, + "offload_optimizer": { + "device": "none", + "pin_memory": true + }, + "allgather_partitions": true, + "allgather_bucket_size": 2e8, + "overlap_comm": false, + "reduce_scatter": true, + "reduce_bucket_size": 2e8, + "contiguous_gradients": true + }, + + "gradient_accumulation_steps": "auto", + "gradient_clipping": "auto", + "steps_per_print": 2000, + "train_batch_size": "auto", + "train_micro_batch_size_per_gpu": "auto", + "wall_clock_breakdown": false, + "elasticity": { + "ignore_non_elastic_batch_info": true, + "enabled": true, + "max_train_batch_size": 8, + "micro_batch_sizes": [ + 4, + 2 + ], + "min_gpus": 1, + "max_gpus": 4, + "min_time": 20, + "version": 0.1 + } +} +``` + +如果用户需要自定义，可以在启动命令中deepspeed_config_or_type指定自定义的zero1.json的存放路径,其中弹性相关的配置为： +```json +... + + "elasticity": { + "ignore_non_elastic_batch_info": true, + "enabled": true, + "max_train_batch_size": 8, + "micro_batch_sizes": [ + 4, + 2 + ], + "min_gpus": 1, + "max_gpus": 4, + "min_time": 20, + "version": 0.1 + } +``` + +- ignore_non_elastic_batch_info：代表在elasticity里的配置会忽略外层的batch_size相关的配置，训练过程中会根据实际的训练进程个数实时修改batch_size等相关的参数 +计算原则为： + global-training-batch-size = micro-batch-size * gradient-accumulation-steps * world-size +- max_train_batch_size：最大batch_size数 +- micro_batch_sizes：elasticity下允许的每卡micro-batch size列表，相当于train_micro_batch_size_per_gpu的候选值 +- min_gpus：最小gpu数目 +- max_gpus：最大gpu数目 +更详细的内容见：[Deepspeed](https://www.deepspeed.ai/docs/config-json/#elastic-training-config-v01-and-v02) + + +## 启动训练 + +```yaml +--- +apiVersion: elastic.iml.github.io/v1alpha1 +kind: ElasticJob +metadata: + name: deepspeed-elastic-swift + namespace: dlrover +spec: + distributionStrategy: AllreduceStrategy + optimizeMode: single-job + replicaSpecs: + worker: + replicas: 1 #【这里需要与启动命令中的--nnodes NNODES的最大值一致】 + template: + spec: + restartPolicy: Never + containers: + - name: main + image: #【训练镜像，需要安装deepspeed,dlrover 和swift 】 + imagePullPolicy: IfNotPresent + command: + - /bin/bash + - -c + - sh start.sh # 启动脚本 + resources: + limits: + cpu: '8' + memory: 16Gi + nvidia.com/gpu: '1' + volumeMounts: + - mountPath: /model + name: volume-model + - mountPath: /dev/shm + name: volume-shm + restartPolicy: Never + volumes: + - hostPath: + path: /model + type: Directory + name: volume-model + - emptyDir: + medium: Memory + sizeLimit: 200Gi + name: volume-shm + +``` diff --git a/docs/source/BestPractices/Embedding.md b/docs/source/BestPractices/Embedding.md new file mode 100644 index 0000000..e59df21 --- /dev/null +++ b/docs/source/BestPractices/Embedding.md @@ -0,0 +1,211 @@ +# Embedding训练 + +SWIFT已经支持Embedding模型的训练，包括纯文本和多模态两个类型。目前已经支持的模型有： + +1. modernbert embedding模型 + - [ModelScope](https://modelscope.cn/models/iic/gte-modernbert-base) [Hugging Face](https://huggingface.co/Alibaba-NLP/gte-modernbert-base) +2. gte embedding模型 + - 1.5B: [ModelScope](https://www.modelscope.cn/models/iic/gte_Qwen2-1.5B-instruct) [Hugging Face](https://huggingface.co/Alibaba-NLP/gte-Qwen2-1.5B-instruct) + - 7B: [ModelScope](https://www.modelscope.cn/models/iic/gte_Qwen2-7B-instruct) [Hugging Face](https://huggingface.co/Alibaba-NLP/gte-Qwen2-7B-instruct) +3. gme embedding模型 + - 2B: [ModelScope](https://www.modelscope.cn/models/iic/gme-Qwen2-VL-2B-Instruct) [Hugging Face](https://huggingface.co/Alibaba-NLP/gme-Qwen2-VL-2B-Instruct) + - 7B: [ModelScope](https://www.modelscope.cn/models/iic/gme-Qwen2-VL-7B-Instruct) [Hugging Face](https://huggingface.co/Alibaba-NLP/gme-Qwen2-VL-7B-Instruct) +4. qwen3-embedding模型 + - 0.6B: [ModelScope](https://www.modelscope.cn/models/Qwen/Qwen3-Embedding-0.6B) [Hugging Face](https://huggingface.co/Qwen/Qwen3-Embedding-0.6B) + - 4B: [ModelScope](https://www.modelscope.cn/models/Qwen/Qwen3-Embedding-4B) [Hugging Face](https://huggingface.co/Qwen/Qwen3-Embedding-4B) + - 8B: [ModelScope](https://www.modelscope.cn/models/Qwen/Qwen3-Embedding-8B) [Hugging Face](https://huggingface.co/Qwen/Qwen3-Embedding-8B) +5. qwen3-vl-embedding模型 + - 2B: [ModelScope](https://www.modelscope.cn/models/Qwen/Qwen3-VL-Embedding-2B) [Hugging Face](https://huggingface.co/Qwen/Qwen3-VL-Embedding-2B) + - 8B: [ModelScope](https://www.modelscope.cn/models/Qwen/Qwen3-VL-Embedding-8B) [Hugging Face](https://huggingface.co/Qwen/Qwen3-VL-Embedding-8B) + + +开发者可以自行集成自己的模型，模型forward输出值需要满足： + +```text +{"last_hidden_state": some-embedding-tensor} +``` + +返回值是一个json，具有`last_hidden_state` key，value是embedding tensor即可，输入部分可以使用我们已经支持的template。用户也可以通过指定 + +```shell + --task_type embedding +``` +参数来将任意一个其他模型转换为embedding模型进行训练。 + +需要注意的是，SWIFT目前支持的embedding模型均为符合纯文本或多模态LLM，目前并不支持CLIP类型的模型训练。 + +此外，SWIFT支持的所有embedding模型在模型forward最后都增加了normalize，如自行增加新模型请注意增加normalize层。 + +## loss + +目前SWIFT支持的Embedding模型可以使用的loss有： + +- cosine_similarity: cosine相似度loss，计算两个embedding的相似度，并根据label的值拟合，实际为MSE loss +- contrastive: 可调margin的对比学习loss，label仅支持0和1两个值 +- online_contrastive: 考虑hard negative和hard positive部分的contrastive loss，label仅支持0和1两个值 +- infonce: 在同一个batch中不同row两两计算cosine相似度，并使row内部相似度最大，不同row相似度最小，不需要label + +loss的源代码可以在[这里](https://github.com/modelscope/ms-swift/blob/main/swift/loss/mapping.py)找到。 + +## 数据集格式 + +> 注： +> 1. ``标签可以出现在`messages`/`positive_messages`/`negative_messages`的任意位置；它们各自拥有独立的`images`/`positive_images`/`negative_images`字段用于提供图片路径或URL。 +> 2. 不再需要跨字段的“对应顺序”。对齐规则为：`images`的长度等于`messages`中``标签的数量；`positive_images`与`negative_images`均为“list of list”，其外层长度分别等于`positive_messages`与`negative_messages`的长度；并且外层每一项的内层列表长度等于该条消息序列中``标签的数量。 +> 3. `messages`代表anchor样本（anchor sample）；`positive_messages`/`negative_messages`为“list of messages”（因此多一层`[]`）；相应地，`positive_images`/`negative_images`也多一层`[]`并与之逐项对齐。 +> 4. 也支持`