Upload folder using huggingface_hub

.github/CODEOWNERS

@@ -0,0 +1,18 @@
+/docs @eric-haibin-lin @zhaochenyang20 @hongpeng-guo
+/docs/amd_tutorial @yushengsu-thu
+/docs/slang_multiturn @zhaochenyang20 @SwordFaith
+
+/recipe/dapo @tongyx361 @PeterSH6
+/recipe/spin @zhaochenyang20
+/recipe/sppo @zhaochenyang20
+
+/third_party/sglang @zhaochenyang20 @SwordFaith
+/third_party/vllm @PeterSH6 @wuxibin89
+/verl/single_controller @zw0610 @wuxibin89
+/verl/trainer @eric-haibin-lin @vermouth1992 @tongyx361 @PeterSH6
+/verl/workers/rollout/vllm_rollout @wuxibin89 @PeterSH6 @chenhaiq
+/verl/workers/rollout/sglang_rollout @zhaochenyang20 @SwordFaith @chenhaiq
+
+/tests/single_controller @zw0610 @wuxibin89
+/tests/trainer @eric-haibin-lin @vermouth1992 @tongyx361 @PeterSH6
+/tests/workers/rollout/vllm_rollout @wuxibin89 @PeterSH6 @chenhaiq

.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,44 @@
+### What does this PR do?
+
+> Add **concise** overview of what this PR aims to achieve or accomplish. Reference related GitHub issues and PRs that help with the review.
+
+### Checklist Before Starting
+
+- [ ] Search for similar PRs. Paste at least one query link here: ...
+- [ ] Format the PR title as `[{modules}] {type}: {description}` (This will be checked by the CI)
+  - `{modules}` include `fsdp`, `megatron`, `sglang`, `vllm`, `rollout`, `trainer`, `ci`, `training_utils`, `recipe`, `hardware`, `deployment`, `ray`, `worker`, `single_controller`, `misc`, `perf`, `model`, `algo`, `env`, `tool`, `ckpt`, `doc`, `data`
+  - If this PR involves multiple modules, separate them with `,` like `[megatron, fsdp, doc]`
+  - `{type}` is in `feat`, `fix`, `refactor`, `chore`, `test`
+  - If this PR breaks any API (CLI arguments, config, function signature, etc.), add `[BREAKING]` to the beginning of the title.
+  - Example: `[BREAKING][fsdp, megatron] feat: dynamic batching`
+
+### Test
+
+> For changes that can not be tested by CI (e.g., algorithm implementation, new model support), validate by experiment(s) and show results like training curve plots, evaluation results, etc.
+
+### API and Usage Example
+
+> Demonstrate how the API changes if any, and provide usage example(s) if possible.
+
+```python
+# Add code snippet or script demonstrating how to use this
+```
+
+### High-Level Design
+
+> Demonstrate the high-level design if this PR is complex.
+
+### Specific Changes
+
+> List the specific changes.
+
+### Checklist Before Submitting
+
+> [!IMPORTANT]
+> Please check all the following items before requesting a review, otherwise the reviewer might deprioritize this PR for review.
+
+- [ ] Read the [Contribute Guide](https://github.com/volcengine/verl?tab=readme-ov-file#contribution-guide).
+- [ ] Apply [pre-commit checks](https://github.com/volcengine/verl?tab=readme-ov-file#code-linting-and-formatting): `pre-commit install && pre-commit run --all-files --show-diff-on-failure --color=always`
+- [ ] Add / Update [the documentation](https://github.com/volcengine/verl/tree/main/docs).
+- [ ] Add unit or end-to-end test(s) to [the CI workflow](https://github.com/volcengine/verl/tree/main/.github/workflows) to cover all the code. If not feasible, explain why: ...
+- [ ] Once your PR is ready for CI, send a message in [the `ci-request` channel](https://verl-project.slack.com/archives/C091TCESWB1) in [the `verl` Slack workspace](https://join.slack.com/t/verl-project/shared_invite/zt-3855yhg8g-CTkqXu~hKojPCmo7k_yXTQ).

.github/dependabot.yml

@@ -0,0 +1,9 @@
+## Enabled the dependabot to check the dependencies of the project
+## Dependabot will open pull requests to update dependencies automatically
+
+version: 2
+updates:
+  - package-ecosystem: pip
+    directory: "/"
+    schedule:
+      interval: weekly

.github/workflows/check-pr-title.yml

@@ -0,0 +1,52 @@
+# # Tests layout
+
+# Each folder under tests/ corresponds to a test category for a sub-namespace in verl. For instance:
+# - `tests/trainer` for testing functionality related to `verl/trainer`
+# - `tests/models` for testing functionality related to `verl/models`
+# - ...
+
+# There are a few folders with `special_` prefix, created for special purposes:
+# - `special_distributed`: unit tests that must run with multiple GPUs
+# - `special_e2e`: end-to-end tests with training/generation scripts
+# - `special_npu`: tests for NPUs
+# - `special_sanity`: a suite of quick sanity tests
+# - `special_standalone`: a set of test that are designed to run in dedicated environments
+
+# Accelerators for tests 
+# - By default tests are run with GPU available, except for the ones under `special_npu`, and any test script whose name ends with `on_cpu.py`.
+# - For test scripts with `on_cpu.py` name suffix would be tested on CPU resources in linux environment.
+
+# # Workflow layout
+
+# All CI tests are configured by yaml files in `.github/workflows/`. Here's an overview of all test configs:
+# 1. A list of always triggered CPU sanity tests: `check-pr-title.yml`, `secrets_scan.yml`, `check-pr-title,yml`, `pre-commit.yml`, `doc.yml`
+# 2. Some heavy multi-GPU unit tests, such as `model.yml`, `vllm.yml`, `sgl.yml`
+# 3. End-to-end tests: `e2e_*.yml`
+# 4. Unit tests
+#   - `cpu_unit_tests.yml`, run pytest on all scripts with file name pattern `tests/**/test_*_on_cpu.py`
+#   - `gpu_unit_tests.yml`, run pytest on all scripts with file without the `on_cpu.py` suffix.
+#   - Since cpu/gpu unit tests by default runs all tests under `tests`, please make sure tests are manually excluded in them when
+#     - new workflow yaml is added to `.github/workflows`
+#     - new tests are added to workflow mentioned in 2.
+
+
+on:
+  pull_request:
+    types: [opened, edited, synchronize]
+
+jobs:
+  check-title:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+
+      - name: Run PR title checker
+        run: python3 tests/special_sanity/check_pr_title.py
+        env:
+          PR_TITLE: ${{ github.event.pull_request.title }}

.github/workflows/checkpoint_converter.yml

@@ -0,0 +1,136 @@
+# # Tests layout
+
+# Each folder under tests/ corresponds to a test category for a sub-namespace in verl. For instance:
+# - `tests/trainer` for testing functionality related to `verl/trainer`
+# - `tests/models` for testing functionality related to `verl/models`
+# - ...
+
+# There are a few folders with `special_` prefix, created for special purposes:
+# - `special_distributed`: unit tests that must run with multiple GPUs
+# - `special_e2e`: end-to-end tests with training/generation scripts
+# - `special_npu`: tests for NPUs
+# - `special_sanity`: a suite of quick sanity tests
+# - `special_standalone`: a set of test that are designed to run in dedicated environments
+
+# Accelerators for tests 
+# - By default tests are run with GPU available, except for the ones under `special_npu`, and any test script whose name ends with `on_cpu.py`.
+# - For test scripts with `on_cpu.py` name suffix would be tested on CPU resources in linux environment.
+
+# # Workflow layout
+
+# All CI tests are configured by yaml files in `.github/workflows/`. Here's an overview of all test configs:
+# 1. A list of always triggered CPU sanity tests: `check-pr-title.yml`, `secrets_scan.yml`, `check-pr-title,yml`, `pre-commit.yml`, `doc.yml`
+# 2. Some heavy multi-GPU unit tests, such as `model.yml`, `vllm.yml`, `sgl.yml`
+# 3. End-to-end tests: `e2e_*.yml`
+# 4. Unit tests
+#   - `cpu_unit_tests.yml`, run pytest on all scripts with file name pattern `tests/**/test_*_on_cpu.py`
+#   - `gpu_unit_tests.yml`, run pytest on all scripts with file without the `on_cpu.py` suffix.
+#   - Since cpu/gpu unit tests by default runs all tests under `tests`, please make sure tests are manually excluded in them when
+#     - new workflow yaml is added to `.github/workflows`
+#     - new tests are added to workflow mentioned in 2.
+
+
+
+name: checkpoint_converter
+# latest version: Megatron-LM core_r0.11.0 https://github.com/NVIDIA/Megatron-LM/tree/core_r0.11.0
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  push:
+    branches:
+      - main
+      - v0.*
+  pull_request:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Recipes
+      - "!recipe/**"
+      # FSDP
+      - "!verl/workers/**/*dp_*.py"
+      # Entrypoints
+      - ".github/workflows/checkpoint_converter.yml"
+      - ".github/workflows/e2e_ppo_trainer_megatron.yml"
+      - "examples/data_preprocess/gsm8k.py"
+      - "tests/special_e2e/run_ppo_trainer_megatron.sh"
+      - "verl/trainer/main_ppo.py"
+      - "verl/trainer/config/ppo_megatron_trainer.yaml"
+
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+# Declare permissions just read content.
+permissions: 
+  contents: read
+
+jobs:
+  checkpoint_converter:
+    runs-on: [L20x8]
+    timeout-minutes: 20 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+            fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install -e .[test]
+      - name: Running Huggingface to Megatron dist_ckpt converter (Qwen/Qwen2.5-0.5B)
+        run: |
+          ray stop --force
+          python scripts/converter_hf_to_mcore.py --hf_model_path=${HOME}/models/Qwen/Qwen2.5-0.5B --output_path checkpoints/Qwen/Qwen2.5-0.5B --test
+      - name: Running Huggingface to Megatron dist_ckpt converter (deepseek-ai/deepseek-coder-1.3b-instruct)
+        run: |
+          ray stop --force
+          python scripts/converter_hf_to_mcore.py --hf_model_path=${HOME}/models/deepseek-ai/deepseek-coder-1.3b-instruct --output_path checkpoints/deepseek-ai/deepseek-coder-1.3b-instruct --test
+      - name: Clean up
+        run: |
+          rm -rf checkpoints
+  checkpoint_converter_large_moe_models:
+    runs-on: [L20x8]
+    timeout-minutes: 30 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+      HF_ENDPOINT: "https://hf-mirror.com"
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+            fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install -e .[test]
+      - name: Download Model to Use
+        run: |
+          huggingface-cli download Qwen/Qwen1.5-MoE-A2.7B-Chat --local-dir ${HOME}/models/Qwen/Qwen1.5-MoE-A2.7B-Chat
+          export HF_HUB_OFFLINE=1
+      - name: Running Huggingface to Megatron dist_ckpt CPU converter (Qwen/Qwen1.5-MoE-A2.7B-Chat)
+        run: |
+          ray stop --force
+          python scripts/converter_hf_to_mcore.py --hf_model_path=${HOME}/models/Qwen/Qwen1.5-MoE-A2.7B-Chat --output_path checkpoints/Qwen/Qwen1.5-MoE-A2.7B-Chat --use_cpu_initialization
+      - name: clean up
+        run: |
+          rm -rf checkpoints

.github/workflows/cpu_unit_tests.yml

@@ -0,0 +1,83 @@
+# # Tests layout
+
+# Each folder under tests/ corresponds to a test category for a sub-namespace in verl. For instance:
+# - `tests/trainer` for testing functionality related to `verl/trainer`
+# - `tests/models` for testing functionality related to `verl/models`
+# - ...
+
+# There are a few folders with `special_` prefix, created for special purposes:
+# - `special_distributed`: unit tests that must run with multiple GPUs
+# - `special_e2e`: end-to-end tests with training/generation scripts
+# - `special_npu`: tests for NPUs
+# - `special_sanity`: a suite of quick sanity tests
+# - `special_standalone`: a set of test that are designed to run in dedicated environments
+
+# Accelerators for tests 
+# - By default tests are run with GPU available, except for the ones under `special_npu`, and any test script whose name ends with `on_cpu.py`.
+# - For test scripts with `on_cpu.py` name suffix would be tested on CPU resources in linux environment.
+
+# # Workflow layout
+
+# All CI tests are configured by yaml files in `.github/workflows/`. Here's an overview of all test configs:
+# 1. A list of always triggered CPU sanity tests: `check-pr-title.yml`, `secrets_scan.yml`, `check-pr-title,yml`, `pre-commit.yml`, `doc.yml`
+# 2. Some heavy multi-GPU unit tests, such as `model.yml`, `vllm.yml`, `sgl.yml`
+# 3. End-to-end tests: `e2e_*.yml`
+# 4. Unit tests
+#   - `cpu_unit_tests.yml`, run pytest on all scripts with file name pattern `tests/**/test_*_on_cpu.py`
+#   - `gpu_unit_tests.yml`, run pytest on all scripts with file without the `on_cpu.py` suffix.
+#   - Since cpu/gpu unit tests by default runs all tests under `tests`, please make sure tests are manually excluded in them when
+#     - new workflow yaml is added to `.github/workflows`
+#     - new tests are added to workflow mentioned in 2.
+
+
+name: cpu_unit_tests
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  push:
+    branches:
+      - main
+      - v0.*
+  pull_request:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      - .github/workflows/cpu_unit_tests.yml
+      - "!recipe/**/*.py"
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+jobs:
+  cpu_unit_tests:
+    runs-on: ubuntu-latest
+    timeout-minutes: 10 # Increase this timeout value as needed
+    strategy:
+      matrix:
+        python-version: ["3.10"]
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Install the current repository
+        run: |
+          pip install -e .[test,prime,geo]
+          pip install --upgrade "ray>=2.40.0" pillow
+      - name: Running CPU unit tests
+        run: |
+          [ ! -d "$HOME/verl-data" ] && git clone --depth 1 https://github.com/eric-haibin-lin/verl-data ~/verl-data
+          python3 examples/data_preprocess/geo3k.py
+          echo '[pytest]' > pytest.ini
+          echo 'python_files = *_on_cpu.py' >> pytest.ini
+          pytest -s -x tests/

.github/workflows/disabled/e2e_prime.yml

@@ -0,0 +1,66 @@
+name: e2e_prime
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  push:
+    branches:
+      - disabled_ci
+  pull_request:
+    branches:
+      - disabled_ci
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Other recipes
+      - "!recipe/**"
+      # Megatron
+      - "!verl/workers/**/megatron_*.py"
+      # Home
+      - "recipe/prime"
+      # Entrypoints
+      - ".github/workflows/e2e_prime.yml"
+      - "examples/data_preprocess/gsm8k.py"
+      - "tests/special_e2e/run_prime.sh"
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+jobs:
+  e2e_prime:
+    runs-on: [L20x8]
+    timeout-minutes: 50 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install --no-deps -e .[test,gpu]
+      - name: Prepare gsm8k dataset
+        run: |
+          ray stop --force
+          python3 examples/data_preprocess/gsm8k.py
+      - name: Running GSM8K E2E with prime alg
+        run: |
+          ray stop --force
+          bash tests/special_e2e/run_prime.sh

.github/workflows/doc.yml

@@ -0,0 +1,92 @@
+# # Tests layout
+
+# Each folder under tests/ corresponds to a test category for a sub-namespace in verl. For instance:
+# - `tests/trainer` for testing functionality related to `verl/trainer`
+# - `tests/models` for testing functionality related to `verl/models`
+# - ...
+
+# There are a few folders with `special_` prefix, created for special purposes:
+# - `special_distributed`: unit tests that must run with multiple GPUs
+# - `special_e2e`: end-to-end tests with training/generation scripts
+# - `special_npu`: tests for NPUs
+# - `special_sanity`: a suite of quick sanity tests
+# - `special_standalone`: a set of test that are designed to run in dedicated environments
+
+# Accelerators for tests 
+# - By default tests are run with GPU available, except for the ones under `special_npu`, and any test script whose name ends with `on_cpu.py`.
+# - For test scripts with `on_cpu.py` name suffix would be tested on CPU resources in linux environment.
+
+# # Workflow layout
+
+# All CI tests are configured by yaml files in `.github/workflows/`. Here's an overview of all test configs:
+# 1. A list of always triggered CPU sanity tests: `check-pr-title.yml`, `secrets_scan.yml`, `check-pr-title,yml`, `pre-commit.yml`, `doc.yml`
+# 2. Some heavy multi-GPU unit tests, such as `model.yml`, `vllm.yml`, `sgl.yml`
+# 3. End-to-end tests: `e2e_*.yml`
+# 4. Unit tests
+#   - `cpu_unit_tests.yml`, run pytest on all scripts with file name pattern `tests/**/test_*_on_cpu.py`
+#   - `gpu_unit_tests.yml`, run pytest on all scripts with file without the `on_cpu.py` suffix.
+#   - Since cpu/gpu unit tests by default runs all tests under `tests`, please make sure tests are manually excluded in them when
+#     - new workflow yaml is added to `.github/workflows`
+#     - new tests are added to workflow mentioned in 2.
+
+
+name: doc_test
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  push:
+    branches:
+      - main
+      - v0.*
+  pull_request:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      - "docs/**"
+      - .github/workflows/doc.yml
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+# Declare permissions just read content.
+permissions:
+  contents: read      # for checkout
+  pages: write        # for deploy-pages
+  id-token: write     # for deploy-pages
+
+jobs:
+  doc_test:
+    runs-on: ubuntu-latest
+    timeout-minutes: 5 # Increase this timeout value as needed
+    strategy:
+      matrix:
+        python-version: ["3.10"]
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Install the current repository
+        run: |
+          pip install -e .[test]
+          pip install -r docs/requirements-docs.txt
+
+      - name: Run doc make html
+        run: |
+          cd docs 
+          make clean
+          make html SPHINXOPTS="--keep-going -w _build/sphinx.log"
+          if grep -q ": ERROR:" _build/sphinx.log; then
+            echo "🚨 Sphinx doc build contained ERRORs - see _build/sphinx.log"
+            exit 1
+          fi
+          if grep -q "WARNING: document isn't included in any toctree" _build/sphinx.log; then
+            echo "🚨 Sphinx doc build contained WARNING. Please include newly added docs in index.rst. See _build/sphinx.log for details"
+            exit 1
+          fi

.github/workflows/e2e_ascend.yml

@@ -0,0 +1,142 @@
+# # Tests layout
+
+# Each folder under tests/ corresponds to a test category for a sub-namespace in verl. For instance:
+# - `tests/trainer` for testing functionality related to `verl/trainer`
+# - `tests/models` for testing functionality related to `verl/models`
+# - ...
+
+# There are a few folders with `special_` prefix, created for special purposes:
+# - `special_distributed`: unit tests that must run with multiple GPUs
+# - `special_e2e`: end-to-end tests with training/generation scripts
+# - `special_npu`: tests for NPUs
+# - `special_sanity`: a suite of quick sanity tests
+# - `special_standalone`: a set of test that are designed to run in dedicated environments
+
+# Accelerators for tests 
+# - By default tests are run with GPU available, except for the ones under `special_npu`, and any test script whose name ends with `on_cpu.py`.
+# - For test scripts with `on_cpu.py` name suffix would be tested on CPU resources in linux environment.
+
+# # Workflow layout
+
+# All CI tests are configured by yaml files in `.github/workflows/`. Here's an overview of all test configs:
+# 1. A list of always triggered CPU sanity tests: `check-pr-title.yml`, `secrets_scan.yml`, `check-pr-title,yml`, `pre-commit.yml`, `doc.yml`
+# 2. Some heavy multi-GPU unit tests, such as `model.yml`, `vllm.yml`, `sgl.yml`
+# 3. End-to-end tests: `e2e_*.yml`
+# 4. Unit tests
+#   - `cpu_unit_tests.yml`, run pytest on all scripts with file name pattern `tests/**/test_*_on_cpu.py`
+#   - `gpu_unit_tests.yml`, run pytest on all scripts with file without the `on_cpu.py` suffix.
+#   - Since cpu/gpu unit tests by default runs all tests under `tests`, please make sure tests are manually excluded in them when
+#     - new workflow yaml is added to `.github/workflows`
+#     - new tests are added to workflow mentioned in 2.
+
+
+name: e2e_ascend
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  push:
+    branches:
+      - main
+      - v0.*
+  pull_request:
+    branches:
+      - main
+    paths:
+      - "**/*.py"
+      - "requirements-npu.txt"
+      # Other entrypoints
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Recipes
+      - "!recipe/**"
+      # Entrypoints
+      - ".github/workflows/e2e_ascend.yml"
+      - "examples/data_preprocess/gsm8k.py"
+      - "examples/data_preprocess/geo3k.py"
+      - "tests/special_e2e/ppo_trainer"
+      - "verl/trainer/main_ppo.py"
+      - "verl/trainer/config/ppo_trainer.yaml"
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+permissions:
+  contents: read
+
+jobs:
+  test:
+    name: verl Ascend test (self-host)
+    runs-on: [self-hosted, npu-0]
+    timeout-minutes: 30 # Increase this timeout value as needed
+    container:
+      image: crispig/verl_npu:cann8.1rc1-py3.10-torch2.5.1-vllm-ascend0.7.3.post1-250616
+      volumes:
+        - /usr/local/dcmi:/usr/local/dcmi
+        - /usr/local/bin/npu-smi:/usr/local/bin/npu-smi
+        - /usr/local/Ascend/driver/lib64/:/usr/local/Ascend/driver/lib64/
+        - /usr/local/Ascend/driver/version.info:/usr/local/Ascend/driver/version.info
+        - /etc/ascend_install.info:/etc/ascend_install.info
+        # Use self-host cache speed up pip and model download
+        # - /home/action/actions-runner/_work/cache:/github/home/.cache/
+      options: >-
+        --device /dev/davinci0
+        --device /dev/davinci_manager
+        --device /dev/devmm_svm
+        --device /dev/hisi_hdc
+        --network host
+        --privileged
+        --shm-size 16g
+    env: 
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    steps:
+      - name: Check npu and CANN info
+        run: |
+          cat /usr/local/Ascend/ascend-toolkit/latest/"$(uname -i)"-linux/ascend_toolkit_install.info
+          npu-smi info
+      - name: Checkout volcengine/verl repo
+        uses: actions/checkout@v4
+      - name: Install the current repository
+        run: |
+          pip3 install hf_transfer peft
+          pip3 install -r requirements-npu.txt
+          pip install -e .
+      - name: Install torchviison
+        run: |
+          pip install torchvision==0.20.1+cpu --index-url https://download.pytorch.org/whl/cpu
+      - name: Prepare gsm8k dataset
+        run: |
+          ray stop --force
+          python3 examples/data_preprocess/gsm8k.py
+      - name: Prepare geo3k dataset
+        run: |
+          ray stop --force
+          python3 examples/data_preprocess/geo3k.py
+      - name: Running gsm8k e2e training tests with LoRA on ASCEND NPU
+        run: |
+          ray stop --force
+          bash tests/special_e2e/sft/run_sft.sh
+          rm -rf $HOME/ckpts
+      - name: Running gsm8k e2e training tests with GRPO on ASCEND NPU
+        run: |
+          ray stop --force
+          bash tests/special_npu/run_qwen2_5_05b_grpo.sh
+          rm -rf $HOME/ckpts
+      - name: Running geo3k e2e training tests with GRPO on ASCEND NPU
+        run: |
+          ray stop --force
+          bash tests/special_npu/run_qwen2_5_vl_3b_npu.sh
+          rm -rf $HOME/ckpts
+      - name: Running gsm8k e2e training tests with DAPO on ASCEND NPU
+        run: |
+          ray stop --force
+          bash tests/special_npu/run_qwen2_5_05b_dapo.sh
+          rm -rf $HOME/ckpts

.github/workflows/e2e_dapo.yml

@@ -0,0 +1,110 @@
+# # Tests layout
+
+# Each folder under tests/ corresponds to a test category for a sub-namespace in verl. For instance:
+# - `tests/trainer` for testing functionality related to `verl/trainer`
+# - `tests/models` for testing functionality related to `verl/models`
+# - ...
+
+# There are a few folders with `special_` prefix, created for special purposes:
+# - `special_distributed`: unit tests that must run with multiple GPUs
+# - `special_e2e`: end-to-end tests with training/generation scripts
+# - `special_npu`: tests for NPUs
+# - `special_sanity`: a suite of quick sanity tests
+# - `special_standalone`: a set of test that are designed to run in dedicated environments
+
+# Accelerators for tests 
+# - By default tests are run with GPU available, except for the ones under `special_npu`, and any test script whose name ends with `on_cpu.py`.
+# - For test scripts with `on_cpu.py` name suffix would be tested on CPU resources in linux environment.
+
+# # Workflow layout
+
+# All CI tests are configured by yaml files in `.github/workflows/`. Here's an overview of all test configs:
+# 1. A list of always triggered CPU sanity tests: `check-pr-title.yml`, `secrets_scan.yml`, `check-pr-title,yml`, `pre-commit.yml`, `doc.yml`
+# 2. Some heavy multi-GPU unit tests, such as `model.yml`, `vllm.yml`, `sgl.yml`
+# 3. End-to-end tests: `e2e_*.yml`
+# 4. Unit tests
+#   - `cpu_unit_tests.yml`, run pytest on all scripts with file name pattern `tests/**/test_*_on_cpu.py`
+#   - `gpu_unit_tests.yml`, run pytest on all scripts with file without the `on_cpu.py` suffix.
+#   - Since cpu/gpu unit tests by default runs all tests under `tests`, please make sure tests are manually excluded in them when
+#     - new workflow yaml is added to `.github/workflows`
+#     - new tests are added to workflow mentioned in 2.
+
+
+name: e2e_dapo
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  # For push, for now only anti-patterns are specified so it is more conservative
+  # and achieves higher coverage.
+  push:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!examples/*trainer*"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Megatron
+      - "!verl/workers/**/megatron_*.py"
+  pull_request:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Other recipes
+      - "!recipe/**"
+      # Megatron
+      - "!verl/workers/**/megatron_*.py"
+      # Home
+      - "recipe/dapo"
+      # Entrypoints
+      - ".github/workflows/e2e_dapo.yml"
+      - "examples/data_preprocess/gsm8k.py"
+      - "tests/special_e2e/run_dapo.sh"
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+jobs:
+  e2e_dapo:
+    runs-on: [L20x8]
+    timeout-minutes: 40 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install --no-deps -e .[test,gpu]
+      - name: Prepare GSM8K dataset
+        run: |
+          python3 examples/data_preprocess/gsm8k.py
+      - name: Running the E2E test with the DAPO algorithm
+        run: |
+          ray stop --force
+          bash tests/special_e2e/run_dapo.sh

.github/workflows/e2e_eval_aime24.yml

@@ -0,0 +1,116 @@
+# # Tests layout
+
+# Each folder under tests/ corresponds to a test category for a sub-namespace in verl. For instance:
+# - `tests/trainer` for testing functionality related to `verl/trainer`
+# - `tests/models` for testing functionality related to `verl/models`
+# - ...
+
+# There are a few folders with `special_` prefix, created for special purposes:
+# - `special_distributed`: unit tests that must run with multiple GPUs
+# - `special_e2e`: end-to-end tests with training/generation scripts
+# - `special_npu`: tests for NPUs
+# - `special_sanity`: a suite of quick sanity tests
+# - `special_standalone`: a set of test that are designed to run in dedicated environments
+
+# Accelerators for tests 
+# - By default tests are run with GPU available, except for the ones under `special_npu`, and any test script whose name ends with `on_cpu.py`.
+# - For test scripts with `on_cpu.py` name suffix would be tested on CPU resources in linux environment.
+
+# # Workflow layout
+
+# All CI tests are configured by yaml files in `.github/workflows/`. Here's an overview of all test configs:
+# 1. A list of always triggered CPU sanity tests: `check-pr-title.yml`, `secrets_scan.yml`, `check-pr-title,yml`, `pre-commit.yml`, `doc.yml`
+# 2. Some heavy multi-GPU unit tests, such as `model.yml`, `vllm.yml`, `sgl.yml`
+# 3. End-to-end tests: `e2e_*.yml`
+# 4. Unit tests
+#   - `cpu_unit_tests.yml`, run pytest on all scripts with file name pattern `tests/**/test_*_on_cpu.py`
+#   - `gpu_unit_tests.yml`, run pytest on all scripts with file without the `on_cpu.py` suffix.
+#   - Since cpu/gpu unit tests by default runs all tests under `tests`, please make sure tests are manually excluded in them when
+#     - new workflow yaml is added to `.github/workflows`
+#     - new tests are added to workflow mentioned in 2.
+
+
+name: e2e_eval_aime24
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  # For push, for now only anti-patterns are specified so it is more conservative
+  # and achieves higher coverage.
+  push:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!*.md"
+      - "!docker/**"
+      - "!docs/**"
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      - "!recipe/r1/README.md"
+  pull_request:
+    branches:
+      - main
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!*.md"
+      - "!docker/**"
+      - "!docs/**"
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Home
+      - "recipe/r1"
+      - "!recipe/r1/README.md"
+      # Other recipes
+      - "!recipe/**"
+      # Entrypoints
+      - ".github/workflows/e2e_eval_aime24.yml"
+      - "tests/special_e2e/run_r1_distill_qwen_aime24_eval.sh"
+      - "verl/trainer/main_generation.py"
+      - "verl/trainer/config/generation.yaml"
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+jobs:
+  e2e_eval_aime24:
+    runs-on: [L20x8]
+    timeout-minutes: 40 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install --no-deps -e .[test,gpu,math]
+          pip3 install math-verify
+      - name: Prepare aime24 dataset
+        run: |
+          ray stop --force
+          python3 recipe/r1/data_process.py --task aime2024
+      - name: Running generation and evaluation in AIME 2024
+        run: |
+          ray stop --force
+          bash tests/special_e2e/run_r1_distill_qwen_aime24_eval.sh

.github/workflows/e2e_ppo_trainer.yml

@@ -0,0 +1,407 @@
+name: e2e_ppo_trainer
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  # For push, for now only anti-patterns are specified so it is more conservative
+  # and achieves higher coverage.
+  push:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Recipes
+      - "!recipe/**"
+      # Megatron
+      - "!verl/workers/**/megatron_*.py"
+
+  pull_request:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!**/*.md"
+      - "!docker/**"
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Docs
+      - "!docs/**"
+      # Recipes
+      - "!recipe/**"
+      # Megatron
+      - "!verl/workers/**/megatron_*.py"
+      # Entrypoints
+      - ".github/workflows/e2e_ppo_trainer.yml"
+      - "examples/data_preprocess/gsm8k.py"
+      - "examples/data_preprocess/geo3k.py"
+      - "tests/special_e2e/ppo_trainer"
+      - "verl/trainer/main_ppo.py"
+      - "verl/trainer/config/ppo_trainer.yaml"
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+jobs:
+  pre_commit_for_ppo:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version: ["3.12"]
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Set ruff --output-format=github
+        run: |
+          sed -i 's/--output-format=full/--output-format=github/' .pre-commit-config.yaml
+          git add .pre-commit-config.yaml
+      - uses: pre-commit/[email protected]
+        with:
+          extra_args: "" # Overriding default "--all-files"
+
+  e2e_ppo_trainer_vllm:
+    runs-on: [L20x8]
+    timeout-minutes: 60 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install --no-deps -e .[test,vllm]
+      - name: Prepare GSM8K dataset
+        run: |
+          ray stop --force
+          python3 examples/data_preprocess/gsm8k.py
+      # Function RM
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with rmpad using function rm with validation and saving (FSDP_SIZE=8)
+        run: |
+          ray stop --force
+          VAL_BEFORE_TRAIN=True TEST_FREQ=1 SAVE_FREQ=1 SAVE_HF_MODEL=True VERL_EXP_NAME="qwen2.5-0.5b-function-reward-minimal-fsdp-size8" bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with rmpad using function rm after resuming
+        run: |
+          ray stop --force
+          RESUME_MODE=auto VERL_EXP_NAME="qwen2.5-0.5b-function-reward-minimal-fsdp-size8" bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Test merging FSDP checkpoints (Qwen Actor) 
+        run: |
+          exp_name="qwen2.5-0.5b-function-reward-minimal-fsdp-size8"
+          python -m verl.model_merger test --backend fsdp --local_dir checkpoints/verl-test/${exp_name}/global_step_1/actor --test_hf_dir checkpoints/verl-test/${exp_name}/global_step_1/actor/huggingface
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with rmpad using function rm with validation and saving (DDP_SIZE=2, FSDP_SIZE=4)
+        run: |
+          ray stop --force
+          VAL_BEFORE_TRAIN=True TEST_FREQ=1 SAVE_FREQ=1 SAVE_HF_MODEL=True FSDP_SIZE=4 VERL_EXP_NAME="qwen2.5-0.5b-function-reward-minimal-ddp-size2-fsdp-size4" bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Test merging DDP+FSDP checkpoints (Qwen Actor) 
+        run: |
+          exp_name="qwen2.5-0.5b-function-reward-minimal-ddp-size2-fsdp-size4"
+          python -m verl.model_merger test --backend fsdp --local_dir checkpoints/verl-test/${exp_name}/global_step_1/actor --test_hf_dir checkpoints/verl-test/${exp_name}/global_step_1/actor/huggingface
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with rmpad using function rm with validation and saving (FSDP2)
+        run: |
+          ray stop --force
+          VAL_BEFORE_TRAIN=True TEST_FREQ=1 SAVE_FREQ=1 SAVE_HF_MODEL=True VERL_EXP_NAME="qwen2.5-0.5b-function-reward-minimal-fsdp2-size8" STRATEGY=fsdp2 bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Test merging FSDP2 checkpoints (Qwen Actor) 
+        run: |
+          exp_name="qwen2.5-0.5b-function-reward-minimal-fsdp2-size8"
+          python -m verl.model_merger test --backend fsdp --local_dir checkpoints/verl-test/${exp_name}/global_step_1/actor --test_hf_dir checkpoints/verl-test/${exp_name}/global_step_1/actor/huggingface
+      - name: Running GSM8K E2E without rmpad using function rm
+        run: |
+          ray stop --force
+          RM_PAD=False bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with rmpad using function rm (GRPO)
+        run: |
+          ray stop --force
+          ADV_ESTIMATOR=grpo USE_KL=True bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with rmpad using function rm (ReMax)
+        run: |
+          ray stop --force
+          ADV_ESTIMATOR=remax USE_KL=True bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with rmpad using customized reward function
+        run: |
+          ray stop --force
+          CUSTOM_REWARD_FN=True bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with rmpad using function rm with in-reward kl and kl loss
+        run: |
+          ray stop --force
+          USE_KL=True bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      # LoRA tests
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with grpo lora using function rm with use_shm
+        run: |
+          ray stop --force
+          ADV_ESTIMATOR=grpo USE_SHM=True LORA_RANK=32 LOAD_FORMAT=safetensors bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with grpo lora using function rm with use_shm and layered_summon
+        run: |
+          ray stop --force
+          ADV_ESTIMATOR=grpo USE_SHM=True LORA_RANK=32 LOAD_FORMAT=safetensors LAYERED_SUMMON=True TOTAL_TRAIN_STEPS=1 SAVE_FREQ=1 FSDP_SIZE=4 VERL_EXP_NAME="qwen2.5-0.5b-function-reward-minimal" bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Test GRPO LoRA checkpoints merging function
+        run: |
+          export EXP_NAME="qwen2.5-0.5b-function-reward-minimal"
+          ls checkpoints/verl-test/${EXP_NAME}/global_step_1/actor
+          cat checkpoints/verl-test/${EXP_NAME}/global_step_1/actor/huggingface/config.json
+          python3 -m verl.model_merger merge --backend fsdp --local_dir checkpoints/verl-test/${EXP_NAME}/global_step_1/actor/ --target_dir checkpoints/verl-test/${EXP_NAME}/global_step_1/actor/huggingface
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with grpo lora using function rm with use_shm and layered_summon with fsdp2
+        run: |
+          ray stop --force
+          ADV_ESTIMATOR=grpo USE_SHM=True LORA_RANK=32 LOAD_FORMAT=safetensors LAYERED_SUMMON=True STRATEGY=fsdp2 bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      # Model RM
+      - name: Running GRPO GSM8K E2E training tests with FSDP on 8 L20 GPUs (DeepSeek)
+        run: |
+          ray stop --force
+          MODEL_ID=deepseek-ai/deepseek-coder-1.3b-instruct bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Running GSM8K E2E with rmpad using model rm
+        run: |
+          ray stop --force
+          bash tests/special_e2e/ppo_trainer/run_model_reward.sh
+      - name: Running GSM8K E2E without rmpad using model rm
+        run: |
+          ray stop --force
+          RM_PAD=False bash tests/special_e2e/ppo_trainer/run_model_reward.sh
+      - name: Running GSM8K E2E with rmpad using model rm and ulysses sp=2
+        run: |
+          ray stop --force
+          SP_SIZE=2 bash tests/special_e2e/ppo_trainer/run_model_reward.sh
+      - name: Running GSM8K E2E with rmpad using model rm and dynamic batch size
+        run: |
+          ray stop --force
+          SEQ_BALANCE=True bash tests/special_e2e/ppo_trainer/run_model_reward.sh
+      - name: Running GSM8K E2E with rmpad using model rm with Liger Kernel enabled
+        run: |
+          ray stop --force
+          LIGER=True bash tests/special_e2e/ppo_trainer/run_model_reward.sh
+      - name: Running GSM8K E2E with rmpad using model rm with Fused Kernel enabled
+        run: |
+          ray stop --force
+          FUSED_KERNELS=True bash tests/special_e2e/ppo_trainer/run_model_reward.sh
+      - name: Running GSM8K E2E with rmpad using model rm with Fused Kernel enabled
+        run: |
+          ray stop --force
+          FUSED_KERNEL=True FUSED_KERNEL_BACKEND=triton bash tests/special_e2e/ppo_trainer/run_model_reward.sh
+
+  e2e_ppo_trainer_vllm_vlm:
+    runs-on: [L20x8]
+    needs: pre_commit_for_ppo
+    timeout-minutes: 40 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: hiyouga/verl:ngc-th2.6.0-cu126-vllm0.8.3-flashinfer0.2.2-cxx11abi0
+      options: --gpus all --shm-size=50g # Visual dataloader requires large memory
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install -e .[test,gpu,vllm,geo,trl]
+      # Geo3k
+      - name: Prepare GEO3K dataset
+        run: |
+          ray stop --force
+          python3 examples/data_preprocess/geo3k.py
+      - name: Running GEO3K VLM GRPO E2E training tests on 8 L20 GPUs with rmpad using function rm
+        run: |
+          ray stop --force
+          TRAIN_FILES=$HOME/data/geo3k/train.parquet VAL_FILES=$HOME/data/geo3k/test.parquet \
+            MAX_PROMPT_LEN=1536 MAX_RESPONSE_LEN=1536 \
+            MODEL_ID=Qwen/Qwen2-VL-2B-Instruct \
+            ADV_ESTIMATOR=grpo RM_PAD=True USE_KL=True ENABLE_CHUNKED_PREFILL=False \
+            SP_SIZE=2 \
+            bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+
+      - name: Running GEO3K VLM PPO E2E training tests on 8 L20 GPUs with rmpad using function rm
+        run: |
+          ray stop --force
+          TRAIN_FILES=$HOME/data/geo3k/train.parquet VAL_FILES=$HOME/data/geo3k/test.parquet \
+            MAX_PROMPT_LEN=1536 MAX_RESPONSE_LEN=1536 \
+            MODEL_ID=Qwen/Qwen2-VL-2B-Instruct \
+            ADV_ESTIMATOR=gae RM_PAD=True USE_KL=True ENABLE_CHUNKED_PREFILL=False \
+            SP_SIZE=2 \
+            bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+
+  e2e_ppo_trainer_sglang:
+    runs-on: [L20x8]
+    needs: pre_commit_for_ppo
+    timeout-minutes: 40 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install -e .[test,gpu,sglang] --no-deps
+      - name: Prepare gsm8k dataset
+        run: |
+          ray stop --force
+          python3 examples/data_preprocess/gsm8k.py
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with rmpad using function rm and save ckpt
+        run: |
+          ray stop --force
+          ENGINE=sglang bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Running GSM8K E2E training tests on sglang async
+        run: |
+          ray stop --force
+          ENGINE=sglang ROLLOUT_MODE=async bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Running GSM8K E2E training tests on vllm async
+        run: |
+          ray stop --force
+          export VLLM_USE_V1=1
+          ray start --head
+          ENGINE=vllm ROLLOUT_MODE=async bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+
+  e2e_ppo_trainer_sglang_multiturn_with_tool:
+    runs-on: [L20x8]
+    needs: pre_commit_for_ppo
+    timeout-minutes: 40 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install -e .[test,gpu,sglang] --no-deps
+      - name: Prepare gsm8k dataset with tool
+        run: |
+          ray stop --force
+          python3 examples/data_preprocess/gsm8k_multiturn_w_tool.py --local_dir $HOME/data/gsm8k_verl_sgl_multi_turn_preprocessed
+      - name: Running GSM8K with tool E2E training tests on 8 L20 GPUs with rmpad using function rm and save ckpt with sglang
+        run: |
+          ray stop --force
+          bash tests/special_e2e/run_gsm8k_fsdp_sgl_multiturn_w_tool.sh
+      - name: Running GSM8K with tool E2E training tests with FSDP2
+        run: |
+          ray stop --force
+          FSDP_STRATEGY=fsdp2 bash tests/special_e2e/run_gsm8k_fsdp_sgl_multiturn_w_tool.sh
+
+  e2e_ppo_trainer_sglang_vlm:
+    runs-on: [L20x8]
+    needs: pre_commit_for_ppo
+    timeout-minutes: 60 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=50g # Visual dataloader requires large memory
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install -e .[test,geo,gpu,sglang]
+      # Geo3k
+      - name: Prepare GEO3K dataset
+        run: |
+          ray stop --force
+          python3 examples/data_preprocess/geo3k.py
+      - name: Running GEO3K VLM E2E training tests on 8 L20 GPUs with rmpad using function rm
+        run: |
+          ray stop --force
+          TRAIN_FILES=$HOME/data/geo3k/train.parquet VAL_FILES=$HOME/data/geo3k/test.parquet \
+            MAX_PROMPT_LEN=1536 MAX_RESPONSE_LEN=1536 \
+            MODEL_ID=Qwen/Qwen2-VL-2B-Instruct \
+            ADV_ESTIMATOR=grpo RM_PAD=True USE_KL=True ENABLE_CHUNKED_PREFILL=False \
+            ENGINE=sglang GPU_MEMORY_UTILIZATION=0.6 ACTOR_FSDP_PARAM_OFFLOAD=True \
+            ACTOR_FSDP_OPTIMIZER_OFFLOAD=True REF_FSDP_PARAM_OFFLOAD=True \
+            bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Running GEO3K VLM E2E with rmpad using torch fused kernel (Qwen2.5-VL)
+        run: |
+          ray stop --force
+          FUSED_KERNELS=True TRAIN_FILES=$HOME/data/geo3k/train.parquet VAL_FILES=$HOME/data/geo3k/test.parquet \
+            MAX_PROMPT_LEN=1536 MAX_RESPONSE_LEN=1536 \
+            MODEL_ID=Qwen/Qwen2.5-VL-3B-Instruct \
+            ADV_ESTIMATOR=grpo RM_PAD=True USE_KL=True ENABLE_CHUNKED_PREFILL=False \
+            ENGINE=sglang GPU_MEMORY_UTILIZATION=0.6 ACTOR_FSDP_PARAM_OFFLOAD=True \
+            ACTOR_FSDP_OPTIMIZER_OFFLOAD=True REF_FSDP_PARAM_OFFLOAD=True \
+            bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+      - name: Running GEO3K VLM E2E with rmpad using triton fused kernel (Qwen2.5-VL)
+        run: |
+          ray stop --force
+          FUSED_KERNELS=True FUSED_KERNEL_BACKEND=triton \
+            TRAIN_FILES=$HOME/data/geo3k/train.parquet VAL_FILES=$HOME/data/geo3k/test.parquet \
+            MAX_PROMPT_LEN=1536 MAX_RESPONSE_LEN=1536 \
+            MODEL_ID=Qwen/Qwen2.5-VL-3B-Instruct \
+            ADV_ESTIMATOR=grpo RM_PAD=True USE_KL=True ENABLE_CHUNKED_PREFILL=False \
+            ENGINE=sglang GPU_MEMORY_UTILIZATION=0.6 ACTOR_FSDP_PARAM_OFFLOAD=True \
+            ACTOR_FSDP_OPTIMIZER_OFFLOAD=True REF_FSDP_PARAM_OFFLOAD=True \
+            bash tests/special_e2e/ppo_trainer/run_function_reward.sh
+
+  e2e_ppo_trainer_sglang_vlm_multiturn_with_tool:
+    runs-on: [L20x8]
+    needs: pre_commit_for_ppo
+    timeout-minutes: 40 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install -e .[test,geo,gpu,sglang]
+      - name: Prepare geo3k dataset with tool
+        run: |
+          ray stop --force
+          python3 examples/data_preprocess/geo3k_multiturn_w_tool.py --local_dir $HOME/data/geo3k_verl_sgl_multi_turn_preprocessed
+      - name: Running GEO3K with tool E2E training tests on 8 L20 GPUs with rmpad using function rm and save ckpt with sglang
+        run: |
+          ray stop --force
+          bash tests/special_e2e/run_geo3k_fsdp_sgl_multiturn_w_tool.sh
+      - name: Running GEO3K with tool E2E training tests with FSDP2
+        run: |
+          ray stop --force
+          FSDP_STRATEGY=fsdp2 bash tests/special_e2e/run_geo3k_fsdp_sgl_multiturn_w_tool.sh

.github/workflows/e2e_ppo_trainer_megatron.yml

@@ -0,0 +1,379 @@
+# # Tests layout
+
+# Each folder under tests/ corresponds to a test category for a sub-namespace in verl. For instance:
+# - `tests/trainer` for testing functionality related to `verl/trainer`
+# - `tests/models` for testing functionality related to `verl/models`
+# - ...
+
+# There are a few folders with `special_` prefix, created for special purposes:
+# - `special_distributed`: unit tests that must run with multiple GPUs
+# - `special_e2e`: end-to-end tests with training/generation scripts
+# - `special_npu`: tests for NPUs
+# - `special_sanity`: a suite of quick sanity tests
+# - `special_standalone`: a set of test that are designed to run in dedicated environments
+
+# Accelerators for tests 
+# - By default tests are run with GPU available, except for the ones under `special_npu`, and any test script whose name ends with `on_cpu.py`.
+# - For test scripts with `on_cpu.py` name suffix would be tested on CPU resources in linux environment.
+
+# # Workflow layout
+
+# All CI tests are configured by yaml files in `.github/workflows/`. Here's an overview of all test configs:
+# 1. A list of always triggered CPU sanity tests: `check-pr-title.yml`, `secrets_scan.yml`, `check-pr-title,yml`, `pre-commit.yml`, `doc.yml`
+# 2. Some heavy multi-GPU unit tests, such as `model.yml`, `vllm.yml`, `sgl.yml`
+# 3. End-to-end tests: `e2e_*.yml`
+# 4. Unit tests
+#   - `cpu_unit_tests.yml`, run pytest on all scripts with file name pattern `tests/**/test_*_on_cpu.py`
+#   - `gpu_unit_tests.yml`, run pytest on all scripts with file without the `on_cpu.py` suffix.
+#   - Since cpu/gpu unit tests by default runs all tests under `tests`, please make sure tests are manually excluded in them when
+#     - new workflow yaml is added to `.github/workflows`
+#     - new tests are added to workflow mentioned in 2.
+
+name: e2e_ppo_trainer_megatron
+# latest version: Megatron-LM core_r0.11.0 https://github.com/NVIDIA/Megatron-LM/tree/core_r0.11.0
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch.
+  # For push, for now only anti-patterns are specified so it is more conservative
+  # and achieves higher coverage.
+  push:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Recipes
+      - "!recipe/**"
+      # FSDP
+      - "!verl/workers/**/*dp_*.py"
+  pull_request:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!docker/**"
+      # Docs
+      - "!**/*.md"
+      - "!docs/**"
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Recipes
+      - "!recipe/**"
+      # FSDP
+      - "!verl/workers/**/*dp_*.py"
+      # Entrypoints
+      - ".github/workflows/e2e_ppo_trainer_megatron.yml"
+      - "examples/data_preprocess/gsm8k.py"
+      - "examples/data_preprocess/geo3k.py"
+      - "tests/special_e2e/run_ppo_trainer_megatron.sh"
+      - "verl/trainer/main_ppo.py"
+      - "verl/trainer/config/ppo_megatron_trainer.yaml"
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+env:
+  IMAGE: "verl-ci-cn-beijing.cr.volces.com/whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3"
+  DYNAMIC_RUNNER_ENDPOINT: "https://sd10g3clalm04ug7alq90.apigateway-cn-beijing.volceapi.com/runner"
+
+jobs:
+  setup:
+    runs-on: ubuntu-latest
+    outputs:
+      runner-label: ${{ steps.create-runner.outputs.runner_label }}
+      mlp-task-id: ${{ steps.create-runner.outputs.mlp_task_id }}
+    steps:
+      - name: create runner
+        id: create-runner
+        shell: bash
+        run: |
+          if [[ "${{ github.event.repository.full_name }}" != "volcengine/verl" ]]; then
+            echo "no need create runner, skip"
+            exit 0
+          fi
+          resp=$(curl -X POST "${{ env.DYNAMIC_RUNNER_ENDPOINT }}/create" \
+          -d '{"Image": "${{ env.IMAGE }}"}')
+          runner_label=$(echo $resp | jq -r '.runner_label')
+          if [[ -z $runner_label || $runner_label == "null" ]]; then
+            echo "create runner failed"
+            exit 1
+          fi
+          echo "runner_label=$runner_label" >> $GITHUB_OUTPUT   
+          mlp_task_id=$(echo $resp | jq -r '.task_id')
+          echo "mlp_task_id=$mlp_task_id" >> $GITHUB_OUTPUT 
+
+  e2e_ppo_trainer_megatron-deepseek:
+    needs: setup
+    runs-on: ["${{ needs.setup.outputs.runner-label || 'L20x8' }}"]
+    timeout-minutes: 60 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install --no-deps -e .[test]
+      - name: Prepare GSM8K dataset
+        run: |
+          python3 examples/data_preprocess/gsm8k.py
+      - name: Running GSM8K E2E training tests with 3D parallelism on 8 L20 GPUs with Megatron (DeepSeek)
+        run: |
+          ray stop --force
+          ALL_OFFLOAD=True SAVE_FREQ=1 MODEL_ID=deepseek-ai/deepseek-coder-1.3b-instruct bash tests/special_e2e/run_ppo_trainer_megatron.sh
+      - name: Running GSM8K E2E training tests with 3D parallelism on 8 L20 GPUs with Megatron (DeepSeek)
+        run: |
+          ray stop --force
+          export VLLM_USE_V1=1
+          ray start --head
+          MODE=async RESUME_MODE=auto MODEL_ID=deepseek-ai/deepseek-coder-1.3b-instruct TOTAL_TRAIN_STEPS=2 bash tests/special_e2e/run_ppo_trainer_megatron.sh
+      - name: Test Megatron checkpoints merging function (DeepSeek Actor and Critic)
+        run: |
+          exp_name="deepseek-coder-1.3b-instruct-megatron-gsm8k-minimal"
+          python -m verl.model_merger test --backend megatron --local_dir checkpoints/verl-test/${exp_name}/global_step_1/actor --test_hf_dir checkpoints/verl-test/${exp_name}/global_step_1/actor/huggingface
+          python -m verl.model_merger test --backend megatron --is-value-model --local_dir checkpoints/verl-test/${exp_name}/global_step_1/critic --test_hf_dir checkpoints/verl-test/${exp_name}/global_step_1/critic/huggingface
+      - name: Running GRPO GSM8K E2E training tests with 3D parallelism on 8 L20 GPUs with Megatron (Deepseek)
+        run: |
+          ray stop --force
+          ADV_ESTIMATOR=grpo USE_DYNAMIC_BSZ=False MODEL_ID=deepseek-ai/deepseek-coder-1.3b-instruct bash tests/special_e2e/run_ppo_trainer_megatron.sh
+      - name: clean up
+        run: |
+          rm -rf checkpoints
+  e2e_ppo_trainer_megatron-qwen3:
+    needs: setup
+    runs-on: ["${{ needs.setup.outputs.runner-label || 'L20x8' }}"]
+    timeout-minutes: 60 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install --no-deps -e .[test]
+      - name: Prepare GSM8K dataset
+        run: |
+          python3 examples/data_preprocess/gsm8k.py
+      - name: Running GSM8K E2E training tests with 3D parallelism on 8 L20 GPUs with Megatron (Qwen3) with validation and saving
+        run: |
+          ray stop --force
+          ALL_OFFLOAD=True VAL_BEFORE_TRAIN=True TEST_FREQ=1 SAVE_FREQ=1 MODEL_ID=Qwen/Qwen3-0.6B bash tests/special_e2e/run_ppo_trainer_megatron.sh
+      - name: Running GSM8K E2E training tests with 3D parallelism on 8 L20 GPUs with Megatron (Qwen3) testing learning rate scheduler
+        run: |
+          ray stop --force
+          LR_WARMUP_STEPS=1 TOTAL_TRAIN_STEPS=2 MODEL_ID=Qwen/Qwen3-0.6B bash tests/special_e2e/run_ppo_trainer_megatron.sh
+
+      - name: Test Megatron checkpoints merging function (Qwen3 Actor and Critic)
+        run: |
+          exp_name="qwen3-0.6b-megatron-gsm8k-minimal"
+          python -m verl.model_merger test --backend megatron --tie-word-embedding --local_dir checkpoints/verl-test/${exp_name}/global_step_1/actor --test_hf_dir checkpoints/verl-test/${exp_name}/global_step_1/actor/huggingface
+          python -m verl.model_merger test --backend megatron --is-value-model --local_dir checkpoints/verl-test/${exp_name}/global_step_1/critic --test_hf_dir checkpoints/verl-test/${exp_name}/global_step_1/critic/huggingface
+      - name: clean up
+        run: |
+          rm -rf checkpoints
+  e2e_ppo_trainer_megatron-different-train-infer-tp-qwen-tie-embedding:
+    needs: setup
+    runs-on: ["${{ needs.setup.outputs.runner-label || 'L20x8' }}"]
+    timeout-minutes: 60 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install --no-deps -e .[test]
+      - name: Prepare GSM8K dataset
+        run: |
+          python3 examples/data_preprocess/gsm8k.py
+      - name: Running GSM8K E2E training tests with 3D parallelism on 8 L20 GPUs with tie-embedding Megatron (Qwen) with train tp > infer tp
+        run: |
+          ray stop --force
+          VAL_BEFORE_TRAIN=True TEST_FREQ=1 SAVE_FREQ=1 TRAIN_TP=2 INFER_TP=1 MODEL_ID=Qwen/Qwen2.5-1.5B bash tests/special_e2e/run_ppo_trainer_megatron.sh
+      - name: Running GSM8K E2E training tests with 3D parallelism on 8 L20 GPUs with Megatron (Qwen) with  train tp < infer tp
+        run: |
+          ray stop --force
+          VAL_BEFORE_TRAIN=True TEST_FREQ=1 SAVE_FREQ=1 TRAIN_TP=1 INFER_TP=2 MODEL_ID=Qwen/Qwen2.5-1.5B bash tests/special_e2e/run_ppo_trainer_megatron.sh
+      - name: clean up
+        run: |
+          rm -rf checkpoints
+  e2e_ppo_trainer_megatron-qwen-override-transformer-config:
+    needs: setup
+    runs-on: ["${{ needs.setup.outputs.runner-label || 'L20x8' }}"]
+    timeout-minutes: 60 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install --no-deps -e .[test]
+      - name: Prepare GSM8K dataset
+        run: |
+          python3 examples/data_preprocess/gsm8k.py
+      - name: Prepare dist_ckpt of Qwen2.5-0.5B, uneven layer distribution only supports dist_ckpt
+        run: |
+          python3 scripts/converter_hf_to_mcore.py --hf_model_path ${HOME}/models/Qwen/Qwen2.5-0.5B --output_path checkpoints/verl-test/qwen2.5-0.5b-megatron
+      - name: Running GSM8K E2E training tests with 3D parallelism on 8 L20 GPUs with Megatron (Qwen)
+        run: |
+          ray stop --force
+          SAVE_FREQ=1 COMMON_PP=4 COMMON_VPP=null COMMON_CP=1 SKIP_SAVE_HF_MODEL=1 bash tests/special_e2e/run_ppo_trainer_megatron.sh +actor_rollout_ref.actor.megatron.override_transformer_config.num_layers_in_first_pipeline_stage=8 +actor_rollout_ref.actor.megatron.override_transformer_config.num_layers_in_last_pipeline_stage=4 actor_rollout_ref.actor.megatron.use_dist_checkpointing=true actor_rollout_ref.actor.megatron.dist_checkpointing_path=checkpoints/verl-test/qwen2.5-0.5b-megatron actor_rollout_ref.ref.megatron.use_dist_checkpointing=true actor_rollout_ref.ref.megatron.dist_checkpointing_path=checkpoints/verl-test/qwen2.5-0.5b-megatron critic.megatron.use_dist_checkpointing=true critic.megatron.dist_checkpointing_path=checkpoints/verl-test/qwen2.5-0.5b-megatron reward_model.megatron.use_dist_checkpointing=true reward_model.megatron.dist_checkpointing_path=checkpoints/verl-test/qwen2.5-0.5b-megatron
+          cp -r checkpoints checkpoints-dut
+          SAVE_FREQ=1 COMMON_PP=4 COMMON_VPP=null COMMON_CP=1 bash tests/special_e2e/run_ppo_trainer_megatron.sh
+      - name: Test Megatron checkpoints merging function (Qwen Actor and Critic)
+        run: |
+          exp_name="qwen2.5-0.5b-megatron-gsm8k-minimal"
+          python -m verl.model_merger test --backend megatron --tie-word-embedding --local_dir checkpoints-dut/verl-test/${exp_name}/global_step_1/actor --test_hf_dir checkpoints/verl-test/${exp_name}/global_step_1/actor/huggingface
+          python -m verl.model_merger test --backend megatron --is-value-model --local_dir checkpoints-dut/verl-test/${exp_name}/global_step_1/critic --test_hf_dir checkpoints/verl-test/${exp_name}/global_step_1/critic/huggingface
+      - name: clean up
+        run: |
+          rm -rf checkpoints
+  e2e_ppo_trainer_megatron-deepseek-override-transformer-config:
+    needs: setup
+    runs-on: ["${{ needs.setup.outputs.runner-label || 'L20x8' }}"]
+    timeout-minutes: 60 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install --no-deps -e .[test]
+      - name: Prepare GSM8K dataset
+        run: |
+          python3 examples/data_preprocess/gsm8k.py
+      - name: Running GSM8K E2E training tests with 3D parallelism on 8 L20 GPUs with Megatron (DeepSeek)
+        run: |
+          ray stop --force
+          SAVE_FREQ=1 MODEL_ID=deepseek-ai/deepseek-coder-1.3b-instruct COMMON_PP=2 COMMON_VPP=null bash tests/special_e2e/run_ppo_trainer_megatron.sh +actor_rollout_ref.actor.megatron.override_transformer_config.account_for_embedding_in_pipeline_split=true +actor_rollout_ref.actor.megatron.override_transformer_config.account_for_loss_in_pipeline_split=true
+      - name: Test Megatron checkpoints merging function (DeepSeek Actor and Critic)
+        run: |
+          exp_name="deepseek-coder-1.3b-instruct-megatron-gsm8k-minimal"
+          python -m verl.model_merger test --backend megatron --local_dir checkpoints/verl-test/${exp_name}/global_step_1/actor --test_hf_dir checkpoints/verl-test/${exp_name}/global_step_1/actor/huggingface
+          python -m verl.model_merger test --backend megatron --is-value-model --local_dir checkpoints/verl-test/${exp_name}/global_step_1/critic --test_hf_dir checkpoints/verl-test/${exp_name}/global_step_1/critic/huggingface
+      - name: clean up
+        run: |
+          rm -rf checkpoints
+  e2e_ppo_trainer_megatron-moe-expert-parallel:
+    needs: setup
+    runs-on: ["${{ needs.setup.outputs.runner-label || 'L20x8' }}"]
+    timeout-minutes: 60 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install --no-deps -e .[test]
+      - name: Prepare GSM8K dataset
+        run: |
+          python3 examples/data_preprocess/gsm8k.py
+      - name: Running GSM8K E2E training tests with 3D parallelism on 8 L20 GPUs with Megatron (DeepSeek)
+        run: |
+          ray stop --force
+          ADV_ESTIMATOR=grpo USE_DUMMY_MODEL=True DUMMY_MODEL_CONFIG_PATH=tests/special_e2e/ppo_trainer/expert_parallel/qwen2moe_minimal.json \
+          PPO_MAX_TOKEN_LEN=512 FWD_MAX_TOKEN_LEN=512 \
+          MAX_PROMPT_LENGTH=256 MAX_RESPONSE_LENGTH=256 \
+          MODEL_ID=Qwen/Qwen1.5-MoE-A2.7B-Chat \
+          COMMON_PP=2 COMMON_VPP=null COMMON_CP=1 COMMON_TP=4 COMMON_EP=4 COMMON_ETP=1 INFER_TP=8 \
+          USE_DIST_CKPT=True ALL_OFFLOAD=True SKIP_SAVE_HF_MODEL=1 bash tests/special_e2e/run_ppo_trainer_megatron.sh
+      - name: clean up
+        run: |
+          rm -rf checkpoints
+  e2e_ppo_trainer_megatron-qwen2_5vl-3b:
+    needs: setup
+    runs-on: ["${{ needs.setup.outputs.runner-label || 'L20x8' }}"]
+    timeout-minutes: 60 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install --no-deps -e .[test]
+      - name: Prepare Geo3k dataset
+        run: |
+          python3 examples/data_preprocess/geo3k.py
+      - name: Prepare dist_ckpt of Qwen2.5-VL-3B, only supports dist_ckpt
+        run: |
+          python3 scripts/converter_hf_to_mcore.py --hf_model_path ${HOME}/models/Qwen/Qwen2.5-VL-3B-Instruct --output_path checkpoints/verl-test/qwen2.5-vl-3b-megatron
+      - name: Running Geo3k E2E training tests with 3D parallelism on 8 L20 GPUs with Megatron (Qwen)
+        run: |
+          ray stop --force
+          TRAIN_FILES=${HOME}/data/geo3k/train.parquet VAL_FILES=${HOME}/data/geo3k/test.parquet MAX_PROMPT_LENGTH=1024 MAX_RESPONSE_LENGTH=2048  MODEL_ID=Qwen/Qwen2.5-VL-3B-Instruct ADV_ESTIMATOR=grpo USE_DYNAMIC_BSZ=False SKIP_SAVE_HF_MODEL=1 COMMON_PP=4 COMMON_VPP=null COMMON_CP=1 COMMON_TP=2 USE_DIST_CKPT=true DIST_CKPT_PATH=checkpoints/verl-test/qwen2.5-vl-3b-megatron bash tests/special_e2e/run_ppo_trainer_megatron.sh 
+      - name: clean up
+        run: |
+          rm -rf checkpoints
+
+  cleanup: 
+    runs-on: ubuntu-latest  
+    needs: [setup, 
+      e2e_ppo_trainer_megatron-deepseek, 
+      e2e_ppo_trainer_megatron-qwen3,
+      e2e_ppo_trainer_megatron-different-train-infer-tp-qwen-tie-embedding,
+      e2e_ppo_trainer_megatron-qwen-override-transformer-config,
+      e2e_ppo_trainer_megatron-deepseek-override-transformer-config,
+      e2e_ppo_trainer_megatron-moe-expert-parallel,
+      e2e_ppo_trainer_megatron-qwen2_5vl-3b]        
+    if: always()          
+    steps:
+      - name: remove runner
+        run: |
+          if [[ -z "${{ needs.setup.outputs.mlp-task-id }}" ]]; then
+            echo "no need remove runner, skip"
+            exit 0
+          fi
+          resp=$(curl -X POST "${{ env.DYNAMIC_RUNNER_ENDPOINT }}/delete" \
+          -d '{"TaskId": "${{ needs.setup.outputs.mlp-task-id }}"}')

.github/workflows/e2e_sft.yml

@@ -0,0 +1,121 @@
+# # Tests layout
+
+# Each folder under tests/ corresponds to a test category for a sub-namespace in verl. For instance:
+# - `tests/trainer` for testing functionality related to `verl/trainer`
+# - `tests/models` for testing functionality related to `verl/models`
+# - ...
+
+# There are a few folders with `special_` prefix, created for special purposes:
+# - `special_distributed`: unit tests that must run with multiple GPUs
+# - `special_e2e`: end-to-end tests with training/generation scripts
+# - `special_npu`: tests for NPUs
+# - `special_sanity`: a suite of quick sanity tests
+# - `special_standalone`: a set of test that are designed to run in dedicated environments
+
+# Accelerators for tests 
+# - By default tests are run with GPU available, except for the ones under `special_npu`, and any test script whose name ends with `on_cpu.py`.
+# - For test scripts with `on_cpu.py` name suffix would be tested on CPU resources in linux environment.
+
+# # Workflow layout
+
+# All CI tests are configured by yaml files in `.github/workflows/`. Here's an overview of all test configs:
+# 1. A list of always triggered CPU sanity tests: `check-pr-title.yml`, `secrets_scan.yml`, `check-pr-title,yml`, `pre-commit.yml`, `doc.yml`
+# 2. Some heavy multi-GPU unit tests, such as `model.yml`, `vllm.yml`, `sgl.yml`
+# 3. End-to-end tests: `e2e_*.yml`
+# 4. Unit tests
+#   - `cpu_unit_tests.yml`, run pytest on all scripts with file name pattern `tests/**/test_*_on_cpu.py`
+#   - `gpu_unit_tests.yml`, run pytest on all scripts with file without the `on_cpu.py` suffix.
+#   - Since cpu/gpu unit tests by default runs all tests under `tests`, please make sure tests are manually excluded in them when
+#     - new workflow yaml is added to `.github/workflows`
+#     - new tests are added to workflow mentioned in 2.
+
+name: e2e_sft
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  push:
+    branches:
+      - main
+      - v0.*
+  pull_request:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Recipes
+      - "!recipe/**"
+      # Megatron
+      - "!verl/workers/**/megatron_*.py"
+      # Entrypoints
+      - ".github/workflows/e2e_sft.yml"
+      - "examples/data_preprocess/gsm8k.py"
+      - "tests/special_e2e/sft"
+      - "verl/trainer/fsdp_sft_trainer.py"
+      - "verl/trainer/config/sft_trainer.yaml"
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+jobs:
+  e2e_sft:
+    runs-on: [L20x8]
+    timeout-minutes: 20 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install peft
+          pip3 install --no-deps -e .[test,gpu]
+      - name: Prepare gsm8k dataset
+        run: |
+          ray stop --force
+          python3 examples/data_preprocess/gsm8k.py
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with rmpad using function rm
+        run: |
+          ray stop --force
+          bash tests/special_e2e/sft/run_sft.sh
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs w/o rmpad using function rm
+        run: |
+          ray stop --force
+          RM_PAD=False bash tests/special_e2e/sft/run_sft.sh
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with sequence parallism
+        run: |
+          ray stop --force
+          SP_SIZE=2 bash tests/special_e2e/sft/run_sft.sh
+      - name: Check loss difference between sequence parallel vs. default implementation
+        run: |
+          ray stop --force
+          ENTRYPOINT="tests/special_e2e/sft/test_sp_loss_match.py" SP_SIZE=2 bash tests/special_e2e/sft/run_sft.sh
+      - name: Running GSM8K E2E training tests on 8 L20 GPUs with sequence parallism and liger
+        run: |
+          ray stop --force
+          SP_SIZE=2 LIGER=True bash tests/special_e2e/sft/run_sft.sh
+      - name: Running GSM8K E2E training tests with LoRA
+        run: |
+          ray stop --force
+          LORA_RANK=32 bash tests/special_e2e/sft/run_sft.sh
+      # TODO: multiturn

.github/workflows/e2e_spin.yml

@@ -0,0 +1,89 @@
+name: e2e_spin
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  push:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Other recipes
+      - "!recipe/**"
+      # Megatron
+      - "!verl/workers/**/megatron_*.py"
+      # Home
+      - "recipe/spin"
+      # Entrypoints
+      - ".github/workflows/e2e_spin.yml"
+      - "examples/data_preprocess/gsm8k.py"
+      - "tests/special_e2e/run_spin.sh"
+      - "!examples"
+  pull_request:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Other recipes
+      - "!recipe/**"
+      # Megatron
+      - "!verl/workers/**/megatron_*.py"
+      # Home
+      - "recipe/spin"
+      # Entrypoints
+      - ".github/workflows/e2e_spin.yml"
+      - "examples/data_preprocess/gsm8k.py"
+      - "tests/special_e2e/run_spin.sh"
+      - "!examples"
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+jobs:
+  e2e_spin:
+    runs-on: [L20x8]
+    timeout-minutes: 40 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: ocss884/verl-sglang:ngc-th2.6.0-cu126-sglang0.4.5.post3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install -e .[test,gpu,sglang] --no-deps
+      - name: Prepare gsm8k dataset
+        run: |
+          python3 examples/data_preprocess/gsm8k.py --local_dir ./data/gsm8k
+      - name: Prepare Model checkpoint
+        run: |
+          huggingface-cli download Qwen/Qwen2.5-0.5B-Instruct --local-dir ./models/Qwen2.5-0.5B-Instruct
+      - name: Running the E2E test with the spin algorithm
+        run: |
+          ray stop --force
+          bash tests/special_e2e/run_spin.sh

.github/workflows/e2e_sppo.yml

@@ -0,0 +1,87 @@
+name: e2e_sppo
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  push:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Other recipes
+      - "!recipe/**"
+      # Megatron
+      - "!verl/workers/**/megatron_*.py"
+      # Home
+      - "recipe/sppo"
+      # Entrypoints
+      - ".github/workflows/e2e_sppo.yml"
+      - "examples/data_preprocess/gsm8k.py"
+      - "tests/special_e2e/run_sppo.sh"
+  pull_request:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Other recipes
+      - "!recipe/**"
+      # Megatron
+      - "!verl/workers/**/megatron_*.py"
+      # Home
+      - "recipe/sppo"
+      # Entrypoints
+      - ".github/workflows/e2e_sppo.yml"
+      - "examples/data_preprocess/gsm8k.py"
+      - "tests/special_e2e/run_sppo.sh"
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+jobs:
+  e2e_sppo:
+    runs-on: [L20x8]
+    timeout-minutes: 40 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install -e .[test,gpu,sglang] --no-deps
+      - name: Prepare MATH dataset
+        run: |
+          python3 examples/data_preprocess/math_dataset.py --local_dir ./data/math
+      - name: Prepare Model checkpoint
+        run: |
+          huggingface-cli download Qwen/Qwen2.5-0.5B-Instruct --local-dir ./models/Qwen2.5-0.5B-Instruct
+      - name: Running the E2E test with the SPPO algorithm
+        run: |
+          ray stop --force
+          bash tests/special_e2e/run_sppo.sh

.github/workflows/gpu_unit_tests.yml

@@ -0,0 +1,100 @@
+# # Tests layout
+
+# Each folder under tests/ corresponds to a test category for a sub-namespace in verl. For instance:
+# - `tests/trainer` for testing functionality related to `verl/trainer`
+# - `tests/models` for testing functionality related to `verl/models`
+# - ...
+
+# There are a few folders with `special_` prefix, created for special purposes:
+# - `special_distributed`: unit tests that must run with multiple GPUs
+# - `special_e2e`: end-to-end tests with training/generation scripts
+# - `special_npu`: tests for NPUs
+# - `special_sanity`: a suite of quick sanity tests
+# - `special_standalone`: a set of test that are designed to run in dedicated environments
+
+# Accelerators for tests 
+# - By default tests are run with GPU available, except for the ones under `special_npu`, and any test script whose name ends with `on_cpu.py`.
+# - For test scripts with `on_cpu.py` name suffix would be tested on CPU resources in linux environment.
+
+# # Workflow layout
+
+# All CI tests are configured by yaml files in `.github/workflows/`. Here's an overview of all test configs:
+# 1. A list of always triggered CPU sanity tests: `check-pr-title.yml`, `secrets_scan.yml`, `check-pr-title,yml`, `pre-commit.yml`, `doc.yml`
+# 2. Some heavy multi-GPU unit tests, such as `model.yml`, `vllm.yml`, `sgl.yml`
+# 3. End-to-end tests: `e2e_*.yml`
+# 4. Unit tests
+#   - `cpu_unit_tests.yml`, run pytest on all scripts with file name pattern `tests/**/test_*_on_cpu.py`
+#   - `gpu_unit_tests.yml`, run pytest on all scripts with file without the `on_cpu.py` suffix.
+#   - Since cpu/gpu unit tests by default runs all tests under `tests`, please make sure tests are manually excluded in them when
+#     - new workflow yaml is added to `.github/workflows`
+#     - new tests are added to workflow mentioned in 2.
+
+name: GPU unit tests
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  push:
+    branches:
+      - main
+      - v0.4.x
+    paths:
+      - "**/*.py"
+      - .github/workflows/gpu_unit_tests.yml
+  pull_request:
+    branches:
+      - main
+      - v0.4.x
+    paths:
+      # The order that you define paths patterns matters:
+      # A matching negative pattern (prefixed with !) after a positive match will exclude the path.
+      # A matching positive pattern after a negative match will include the path again.
+      - "**/*.py"
+      # Other entrypoints
+      - "!examples/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      - "!recipe/**"
+      # Entrypoints
+      - .github/workflows/gpu_unit_tests.yml
+      - "tests/**test_*.py"
+      # Ignore CPU tests
+      - "!tests/*_on_cpu.py"
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+# Declare permissions just read content.
+permissions: 
+  contents: read
+
+jobs:
+  gpu_unit_tests:
+    runs-on: [L20x8]
+    timeout-minutes: 40 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1"
+      HF_HUB_ENABLE_HF_TRANSFER: 1
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+            fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install hf_transfer
+          pip3 install --no-deps -e .[test]
+          pip3 install --upgrade "ray>=2.40.0"
+          pip3 install cupy-cuda12x
+      - name: Run all GPU unit tests
+        run: |
+          pytest -s -x --ignore-glob="*test_linear_cross_entropy_tp.py" --ignore-glob='*on_cpu.py' --ignore-glob="*test_vllm*" --ignore-glob="*_sglang*" --ignore-glob="*_hf_rollout*" --ignore-glob="tests/models/" --ignore-glob='tests/special*' tests/
+      - name: Testing LinearCrossEntropyTP Correctness, Computation Time and Memory Consumption
+        run: |
+          LOW_MEMORY=True torchrun --standalone --nnodes=1 --nproc-per-node=8 tests/utils/test_linear_cross_entropy_tp.py

.github/workflows/model.yml

@@ -0,0 +1,144 @@
+# # Tests layout
+
+# Each folder under tests/ corresponds to a test category for a sub-namespace in verl. For instance:
+# - `tests/trainer` for testing functionality related to `verl/trainer`
+# - `tests/models` for testing functionality related to `verl/models`
+# - ...
+
+# There are a few folders with `special_` prefix, created for special purposes:
+# - `special_distributed`: unit tests that must run with multiple GPUs
+# - `special_e2e`: end-to-end tests with training/generation scripts
+# - `special_npu`: tests for NPUs
+# - `special_sanity`: a suite of quick sanity tests
+# - `special_standalone`: a set of test that are designed to run in dedicated environments
+
+# Accelerators for tests 
+# - By default tests are run with GPU available, except for the ones under `special_npu`, and any test script whose name ends with `on_cpu.py`.
+# - For test scripts with `on_cpu.py` name suffix would be tested on CPU resources in linux environment.
+
+# # Workflow layout
+
+# All CI tests are configured by yaml files in `.github/workflows/`. Here's an overview of all test configs:
+# 1. A list of always triggered CPU sanity tests: `check-pr-title.yml`, `secrets_scan.yml`, `check-pr-title,yml`, `pre-commit.yml`, `doc.yml`
+# 2. Some heavy multi-GPU unit tests, such as `model.yml`, `vllm.yml`, `sgl.yml`
+# 3. End-to-end tests: `e2e_*.yml`
+# 4. Unit tests
+#   - `cpu_unit_tests.yml`, run pytest on all scripts with file name pattern `tests/**/test_*_on_cpu.py`
+#   - `gpu_unit_tests.yml`, run pytest on all scripts with file without the `on_cpu.py` suffix.
+#   - Since cpu/gpu unit tests by default runs all tests under `tests`, please make sure tests are manually excluded in them when
+#     - new workflow yaml is added to `.github/workflows`
+#     - new tests are added to workflow mentioned in 2.
+# name: Check PR Title
+
+name: model_rmpad
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  push:
+    branches:
+      - main
+      - v0.*
+  pull_request:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "verl/**/*.py"
+      # Entrypoints
+      - ".github/workflows/model.yml"
+      - "tests/special_distributed/test_fsdp_ckpt.py"
+      - "tests/models/**"
+      - "tests/special_distributed/run_all.sh"
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+jobs:
+  model_rmpad:
+    runs-on: [L20x8]
+    timeout-minutes: 20 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository and upgrade to latest transformers/flash_attn
+        run: |
+          pip3 install --no-deps -e .[test]
+          pip3 install --upgrade transformers
+      - name: Running rmpad model tests on 8 L20 GPUs + flash_attn 2.5.8
+        run: |
+          pytest -s tests/models/test_transformer.py
+      - name: Running rmpad model tests on 8 L20 GPUs + latest flash_attn
+        run: |
+          pytest -s tests/models/test_transformer.py
+      - name: Running FSDP rmpad model tests on 8 L20 GPUs + latest flash_attn
+        run: |
+          STRATEGY=fsdp torchrun --nproc_per_node=8 tests/special_distributed/test_fsdp_ckpt.py
+      - name: Running transformers ulysses tests on 8 L20 GPUs + latest transformers
+        run: |
+          torchrun --nproc_per_node=8 -m pytest tests/models/test_transformers_ulysses.py
+      - name: Running transformers ulysses tests on 8 L20 GPUs + transformers 4.49.0
+        run: |
+          pip3 install transformers==4.49.0
+          torchrun --nproc_per_node=8 -m pytest tests/models/test_transformers_ulysses.py
+      - name: Running transformers ulysses tests on 8 L20 GPUs + transformers 4.48.0
+        run: |
+          pip3 install transformers==4.48.0
+          torchrun --nproc_per_node=8 -m pytest tests/models/test_transformers_ulysses.py
+      - name: Running transformers ulysses tests on 8 L20 GPUs + transformers 4.47.0
+        run: |
+          pip3 install transformers==4.47.0
+          torchrun --nproc_per_node=8 -m pytest tests/models/test_transformers_ulysses.py
+      - name: Running transformers ulysses tests on 8 L20 GPUs + transformers 4.46.0
+        run: |
+          pip3 install transformers==4.46.0
+          torchrun --nproc_per_node=8 -m pytest tests/models/test_transformers_ulysses.py
+      - name: Running transformers ulysses tests on 8 L20 GPUs + transformers 4.45.0
+        run: |
+          pip3 install transformers==4.45.0
+          torchrun --nproc_per_node=8 -m pytest tests/models/test_transformers_ulysses.py
+      - name: Run distributed test
+        run: |
+          bash tests/special_distributed/run_all.sh
+
+  # TODO: Move this back to model_rmpad once FSDP2 is stable.
+  # NOTE: List as an independent job to make rerun easier.
+  model_rmpad_fsdp2_unstable:
+    runs-on: [L20x8]
+    timeout-minutes: 20 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository and upgrade to latest transformers/flash_attn
+        run: |
+          pip3 install --no-deps -e .[test]
+          pip3 install --upgrade transformers
+      - name: Running FSDP2 rmpad model tests on 8 L20 GPUs + latest flash_attn
+        run: |
+          STRATEGY=fsdp2 torchrun --nproc_per_node=8 tests/special_distributed/test_fsdp_ckpt.py

.github/workflows/pre-commit-full.yml

@@ -0,0 +1,30 @@
+name: pre-commit-full
+
+# Run weekly on Sunday at 00:00 UTC
+on:
+  schedule:
+    - cron: "0 0 * * 0"
+  # Allow manual triggering
+  workflow_dispatch:
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+jobs:
+  pre-commit-full:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version: ["3.12"]
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Set ruff --output-format=github
+        run: |
+          sed -i 's/--output-format=full/--output-format=github/' .pre-commit-config.yaml
+          git add .pre-commit-config.yaml
+      - uses: pre-commit/[email protected]

.github/workflows/pre-commit.yml

@@ -0,0 +1,33 @@
+# c.f. https://github.com/pre-commit/action?tab=readme-ov-file#using-this-action
+name: pre-commit
+
+# No need to avoid / cancel lightweight pre-commit jobs
+on:
+  pull_request:
+  push:
+    branches:
+      - main
+      - v0.*
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+jobs:
+  pre-commit:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version: ["3.12"]
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Set ruff --output-format=github
+        run: |
+          sed -i 's/--output-format=full/--output-format=github/' .pre-commit-config.yaml
+          git add .pre-commit-config.yaml
+      # Check "--all-files" by default
+      - uses: pre-commit/[email protected]

.github/workflows/sanity.yml

@@ -0,0 +1,95 @@
+# # Tests layout
+
+# Each folder under tests/ corresponds to a test category for a sub-namespace in verl. For instance:
+# - `tests/trainer` for testing functionality related to `verl/trainer`
+# - `tests/models` for testing functionality related to `verl/models`
+# - ...
+
+# There are a few folders with `special_` prefix, created for special purposes:
+# - `special_distributed`: unit tests that must run with multiple GPUs
+# - `special_e2e`: end-to-end tests with training/generation scripts
+# - `special_npu`: tests for NPUs
+# - `special_sanity`: a suite of quick sanity tests
+# - `special_standalone`: a set of test that are designed to run in dedicated environments
+
+# Accelerators for tests 
+# - By default tests are run with GPU available, except for the ones under `special_npu`, and any test script whose name ends with `on_cpu.py`.
+# - For test scripts with `on_cpu.py` name suffix would be tested on CPU resources in linux environment.
+
+# # Workflow layout
+
+# All CI tests are configured by yaml files in `.github/workflows/`. Here's an overview of all test configs:
+# 1. A list of always triggered CPU sanity tests: `check-pr-title.yml`, `secrets_scan.yml`, `check-pr-title,yml`, `pre-commit.yml`, `doc.yml`
+# 2. Some heavy multi-GPU unit tests, such as `model.yml`, `vllm.yml`, `sgl.yml`
+# 3. End-to-end tests: `e2e_*.yml`
+# 4. Unit tests
+#   - `cpu_unit_tests.yml`, run pytest on all scripts with file name pattern `tests/**/test_*_on_cpu.py`
+#   - `gpu_unit_tests.yml`, run pytest on all scripts with file without the `on_cpu.py` suffix.
+#   - Since cpu/gpu unit tests by default runs all tests under `tests`, please make sure tests are manually excluded in them when
+#     - new workflow yaml is added to `.github/workflows`
+#     - new tests are added to workflow mentioned in 2.
+# name: Check PR Title
+
+name: sanity
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  push:
+    branches:
+      - main
+      - v0.*
+  pull_request:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      - .github/workflows/sanity.yml
+      - "tests/special_sanity/**"
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+jobs:
+  sanity:
+    runs-on: ubuntu-latest
+    timeout-minutes: 5 # Increase this timeout value as needed
+    strategy:
+      matrix:
+        python-version: ["3.10"]
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Install the current repository
+        run: |
+          pip install -e .[test]
+      - name: Run sanity test
+        run: |
+          pytest -s -x tests/special_sanity
+      - name: Run license test
+        run: |
+          python3 tests/special_sanity/check_license.py --directory .
+      - name: Assert naming convention
+        run: |
+          if grep -rIn --exclude-dir=.git --exclude-dir=.github --exclude-dir=venv --exclude-dir=__pycache__ 'veRL' .; then
+            echo "Please use verl instead of veRL in the codebase"
+            exit 1
+          fi
+      - name: Validate test folder structure
+        run: python3 tests/special_sanity/validate_structure.py
+      - name: Assert documentation requirement for functions
+        run: python3 tests/special_sanity/validate_imported_docs.py
+      - name: Assert device api usage in verl/recipe
+        run: python3 tests/special_sanity/check_device_api_usage.py --directory ./recipe
+      - name: Assert device api usage in verl/verl
+        run: python3 tests/special_sanity/check_device_api_usage.py --directory ./verl

.github/workflows/scorecard.yml

@@ -0,0 +1,66 @@
+# This workflow uses actions that are not certified by GitHub. They are provided
+# by a third-party and are governed by separate terms of service, privacy
+# policy, and support documentation.
+
+name: Scorecard supply-chain security
+on:
+  # For Branch-Protection check. Only the default branch is supported. See
+  # https://github.com/ossf/scorecard/blob/main/docs/checks.md#branch-protection
+  branch_protection_rule:
+  # To guarantee Maintained check is occasionally updated. See
+  # https://github.com/ossf/scorecard/blob/main/docs/checks.md#maintained
+  schedule:
+    - cron: "27 7 * * 1"
+  push:
+    branches:
+      - main
+      - v0.*
+
+# Declare default permissions as read only.
+permissions: read-all
+
+jobs:
+  analysis:
+    name: Scorecard analysis
+    runs-on: ubuntu-latest
+    permissions:
+      # Needed to upload the results to code-scanning dashboard.
+      security-events: write
+      # Needed to publish results and get a badge (see publish_results below).
+      id-token: write
+      # Uncomment the permissions below if installing in a private repository.
+      # contents: read
+      # actions: read
+
+    steps:
+      - name: "Checkout code"
+        uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
+        with:
+          persist-credentials: false
+
+      - name: "Run analysis"
+        uses: ossf/scorecard-action@0864cf19026789058feabb7e87baa5f140aac736 # v2.3.1
+        with:
+          results_file: results.sarif
+          results_format: sarif
+          # (Optional) "write" PAT token. Uncomment the `repo_token` line below if:
+          # - you want to enable the Branch-Protection check on a *public* repository, or
+          # - you are installing Scorecard on a *private* repository
+          # To create the PAT, follow the steps in https://github.com/ossf/scorecard-action?tab=readme-ov-file#authentication-with-fine-grained-pat-optional.
+          # repo_token: ${{ secrets.SCORECARD_TOKEN }}
+
+          # Public repositories:
+          #   - Publish results to OpenSSF REST API for easy access by consumers
+          #   - Allows the repository to include the Scorecard badge.
+          #   - See https://github.com/ossf/scorecard-action#publishing-results.
+          # For private repositories:
+          #   - `publish_results` will always be set to `false`, regardless
+          #     of the value entered here.
+          publish_results: true
+
+      # Upload the results to GitHub's code scanning dashboard (optional).
+      # Commenting out will disable upload of results to your repo's Code Scanning dashboard
+      - name: "Upload to code-scanning"
+        uses: github/codeql-action/upload-sarif@9e8d0789d4a0fa9ceb6b1738f7e269594bdd67f0 #v3.28.9
+        with:
+          sarif_file: results.sarif

.github/workflows/secrets_scan.yml

@@ -0,0 +1,22 @@
+on:
+  push:
+    branches:
+      - main
+      - v0.*
+  pull_request:
+
+permissions:
+  contents: read
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
+        with:
+          fetch-depth: 0
+      - name: Secret Scanning
+        uses: trufflesecurity/trufflehog@7dc056a193116ba8d82154bf0549381c8fb8545c # v3.88.14
+        with:
+          extra_args: --results=verified,unknown

.github/workflows/sgl.yml

@@ -0,0 +1,124 @@
+# # Tests layout
+
+# Each folder under tests/ corresponds to a test category for a sub-namespace in verl. For instance:
+# - `tests/trainer` for testing functionality related to `verl/trainer`
+# - `tests/models` for testing functionality related to `verl/models`
+# - ...
+
+# There are a few folders with `special_` prefix, created for special purposes:
+# - `special_distributed`: unit tests that must run with multiple GPUs
+# - `special_e2e`: end-to-end tests with training/generation scripts
+# - `special_npu`: tests for NPUs
+# - `special_sanity`: a suite of quick sanity tests
+# - `special_standalone`: a set of test that are designed to run in dedicated environments
+
+# Accelerators for tests 
+# - By default tests are run with GPU available, except for the ones under `special_npu`, and any test script whose name ends with `on_cpu.py`.
+# - For test scripts with `on_cpu.py` name suffix would be tested on CPU resources in linux environment.
+
+# # Workflow layout
+
+# All CI tests are configured by yaml files in `.github/workflows/`. Here's an overview of all test configs:
+# 1. A list of always triggered CPU sanity tests: `check-pr-title.yml`, `secrets_scan.yml`, `check-pr-title,yml`, `pre-commit.yml`, `doc.yml`
+# 2. Some heavy multi-GPU unit tests, such as `model.yml`, `vllm.yml`, `sgl.yml`
+# 3. End-to-end tests: `e2e_*.yml`
+# 4. Unit tests
+#   - `cpu_unit_tests.yml`, run pytest on all scripts with file name pattern `tests/**/test_*_on_cpu.py`
+#   - `gpu_unit_tests.yml`, run pytest on all scripts with file without the `on_cpu.py` suffix.
+#   - Since cpu/gpu unit tests by default runs all tests under `tests`, please make sure tests are manually excluded in them when
+#     - new workflow yaml is added to `.github/workflows`
+#     - new tests are added to workflow mentioned in 2.
+
+name: sgl
+
+on:
+  workflow_dispatch: # Manual
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  push:
+    branches:
+      - main
+      - v0.2.x
+    paths:
+      - "**/*.py"
+      - .github/workflows/vllm.yml
+  pull_request:
+    branches:
+      - main
+      - v0.2.x
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # FSDP
+      - "!verl/workers/**/*dp_*.py"
+      # Megatron
+      - "!verl/workers/**/megatron_*.py"
+      # vLLM
+      - "!**/*vllm*"
+      # Recipes
+      - "!recipe/**"
+      # Entrypoints
+      - ".github/workflows/sgl.yml"
+      - "tests/rollout/*sglang*"
+      - "tests/rollout/async_rollout_utils.py"
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+jobs:
+  sgl:
+    runs-on: [L20x8]
+    timeout-minutes: 20 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: 1
+      SGL_DISABLE_TP_MEMORY_INBALANCE_CHECK: "True"
+    container:
+      image: whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6.post5-mcore0.12.0-te2.3
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install hf_transfer fastmcp
+          pip3 install -e .[test,gpu,sglang] --no-deps
+      - name: Download Model to Use
+        run: |
+          huggingface-cli download 'Qwen/Qwen2-7B-Instruct'
+          export HF_HUB_OFFLINE=1
+      - name: Test the latest SGLang
+        run: |
+          cd tests/workers/rollout
+          torchrun --nnodes=1 --nproc_per_node=2 $(which pytest) -s test_sglang_spmd.py
+      - name: Test the latest SGLang Rollout async with tool
+        run: |
+          cd tests/workers/rollout
+          torchrun --nnodes=1 --nproc_per_node=2 $(which pytest) -s test_sglang_async_rollout_w_tools.py
+      - name: Test the latest SGLang Rollout async with sandbox fusion tool
+        run: |
+          cd tests/workers/rollout
+          pytest -s test_sglang_async_rollout_sf_tools.py
+      - name: Test the latest SGLang Rollout async with search tool
+        run: |
+          cd tests/workers/rollout
+          pytest -s test_sglang_async_rollout_search_tools.py
+      - name: Test the latest SGLang Rollout async with mcp search tool
+        run: |
+          cd tests/workers/rollout
+          pytest -s test_sglang_async_rollout_mcp_tools.py
+      # Note(haibin.lin): for any new test, please update gpu_unit_tests.yaml to avoid repeated tests

.github/workflows/type-coverage-check.yml

@@ -0,0 +1,29 @@
+name: Type Annotation and Docstring Coverage
+
+on:
+  pull_request:
+    paths:
+      - '**/*.py'
+
+jobs:
+  type-coverage-check:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0  # 🚨 Important: fetch full history so `origin/main` is available
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.10'
+
+      - name: Install dependencies
+        run: |
+          pip install gitpython
+          pip install -e .[sglang]
+      - name: Run type annotation coverage check
+        run: |
+          python3 tests/special_sanity/type_coverage_check.py
+      - name: Run docstring coverage check
+        run: |
+          python3 tests/special_sanity/check_api_docs.py verl

.github/workflows/vllm.yml

@@ -0,0 +1,131 @@
+# # Tests layout
+
+# Each folder under tests/ corresponds to a test category for a sub-namespace in verl. For instance:
+# - `tests/trainer` for testing functionality related to `verl/trainer`
+# - `tests/models` for testing functionality related to `verl/models`
+# - ...
+
+# There are a few folders with `special_` prefix, created for special purposes:
+# - `special_distributed`: unit tests that must run with multiple GPUs
+# - `special_e2e`: end-to-end tests with training/generation scripts
+# - `special_npu`: tests for NPUs
+# - `special_sanity`: a suite of quick sanity tests
+# - `special_standalone`: a set of test that are designed to run in dedicated environments
+
+# Accelerators for tests 
+# - By default tests are run with GPU available, except for the ones under `special_npu`, and any test script whose name ends with `on_cpu.py`.
+# - For test scripts with `on_cpu.py` name suffix would be tested on CPU resources in linux environment.
+
+# # Workflow layout
+
+# All CI tests are configured by yaml files in `.github/workflows/`. Here's an overview of all test configs:
+# 1. A list of always triggered CPU sanity tests: `check-pr-title.yml`, `secrets_scan.yml`, `check-pr-title,yml`, `pre-commit.yml`, `doc.yml`
+# 2. Some heavy multi-GPU unit tests, such as `model.yml`, `vllm.yml`, `sgl.yml`
+# 3. End-to-end tests: `e2e_*.yml`
+# 4. Unit tests
+#   - `cpu_unit_tests.yml`, run pytest on all scripts with file name pattern `tests/**/test_*_on_cpu.py`
+#   - `gpu_unit_tests.yml`, run pytest on all scripts with file without the `on_cpu.py` suffix.
+#   - Since cpu/gpu unit tests by default runs all tests under `tests`, please make sure tests are manually excluded in them when
+#     - new workflow yaml is added to `.github/workflows`
+#     - new tests are added to workflow mentioned in 2.
+
+
+name: vllm
+
+on:
+  # Trigger the workflow on push or pull request,
+  # but only for the main branch
+  push:
+    branches:
+      - main
+      - v0.*
+  pull_request:
+    branches:
+      - main
+      - v0.*
+    paths:
+      - "**/*.py"
+      # Other entrypoints
+      - "!examples/**"
+      - "!tests/**"
+      - "!verl/trainer/main_*.py"
+      - "!verl/trainer/fsdp_sft_trainer.py"
+      # Recipes
+      - "!recipe/**"
+      # FSDP
+      - "!verl/workers/**/*dp_*.py"
+      # Megatron
+      - "!verl/workers/**/megatron_*.py"
+      # SGLang
+      - "!**/*sglang*"
+      # Entrypoints
+      - ".github/workflows/vllm.yml"
+      - "tests/special_e2e/generation"
+      - "tests/workers/rollout"
+      - "verl/trainer/main_generation.py"
+      - "verl/trainer/config/generation.yaml"
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+# Declare permissions just read content.
+permissions:
+  contents: read
+
+jobs:
+  vllm:
+    runs-on: [L20x8]
+    timeout-minutes: 60 # Increase this timeout value as needed
+    env:
+      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
+      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
+      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
+      HF_ENDPOINT: "https://hf-mirror.com"
+      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
+    container:
+      image: whatcanyousee/verl:vemlp-th2.4.0-cu124-vllm0.6.3-ray2.10-te2.0-megatron0.11.0-v0.0.6
+      options: --gpus all --shm-size=10g
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+      - name: Install the current repository
+        run: |
+          pip3 install -e .[test]
+          pip3 install vllm==0.5.4
+      - name: Download Model to Use
+        run: |
+          huggingface-cli download Qwen/Qwen2.5-0.5B-Instruct
+          huggingface-cli download Qwen/Qwen2.5-1.5B-Instruct
+          huggingface-cli download 'Qwen/Qwen2-7B-Instruct'
+          huggingface-cli download 'deepseek-ai/deepseek-llm-7b-chat'
+          export HF_HUB_OFFLINE=1
+        # Disable requests to avoid network errors
+      - name: Running vllm tests on 8 L20 GPUs
+        run: |
+          cd tests/workers/rollout/rollout_vllm
+          torchrun --standalone --nnodes=1 --nproc_per_node=8 $(which pytest) -s test_vllm_hf_loader.py
+      - name: Test the latest vLLM
+        run: |
+          pip3 install --upgrade vllm==0.7.3
+          cd tests/workers/rollout/rollout_vllm
+          torchrun --standalone --nnodes=1 --nproc_per_node=4 $(which pytest) -s test_vllm_spmd.py
+      - name: Run Qwen 0.5B generation test
+        run: |
+          cd tests/special_e2e/generation
+          export OUTPUT_PATH="${HOME}/data/gen/qwen_05_gen_test.parquet"
+          MODEL_ID=Qwen/Qwen2.5-0.5B-Instruct NGPUS_PER_NODE=4 GEN_TP=2 bash ./run_gen_qwen05.sh
+          rm -rf "${OUTPUT_PATH}"
+      - name: Run Qwen 0.5B generation test when world_size == 1
+        run: |
+          cd tests/special_e2e/generation
+          export OUTPUT_PATH="${HOME}/data/gen/qwen_05_gen_test.parquet"
+          MODEL_ID=Qwen/Qwen2.5-0.5B-Instruct NGPUS_PER_NODE=1 GEN_TP=1 bash ./run_gen_qwen05.sh
+          rm -rf "${OUTPUT_PATH}"
+      - name: Running multi-turn rollout tests on 8 L20 GPUs
+        run: |
+          pip3 install --upgrade vllm==0.8.3 tensordict==0.7.2
+          pytest -svvv tests/workers/rollout/rollout_vllm/test_vllm_chat_scheduler.py
+      # Note(haibin.lin): for any new test, please update gpu_unit_tests.yaml to avoid repeated tests

.gitignore

@@ -0,0 +1,126 @@
+
+**/*.pt
+**/checkpoints
+**/wget-log
+**/_build/
+**/*.ckpt
+**/outputs
+**/*.tar.gz
+**/playground
+**/wandb
+
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+dataset/*
+tensorflow/my_graph/*
+.idea/
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+env/
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+tmp/
+*.egg-info/
+.installed.cfg
+*.egg
+
+# 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/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# IPython Notebook
+.ipynb_checkpoints
+
+# pyenv
+.python-version
+
+# celery beat schedule file
+celerybeat-schedule
+
+# dotenv
+.env
+
+# virtualenv
+venv/
+.venv/
+ENV/
+
+# Spyder project settings
+.spyderproject
+
+# Rope project settings
+.ropeproject
+
+# vscode
+.vscode
+
+# Mac
+.DS_Store
+
+# vim
+*.swp
+
+# ckpt
+*.lock
+
+# data
+*.parquet
+
+
+# local logs
+logs
+log
+outputs
+.history

.pre-commit-config.yaml

@@ -0,0 +1,8 @@
+repos:
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: "v0.11.4"
+    hooks:
+      - id: ruff
+        args: ["--fix", "--show-fixes", "--output-format=full"]
+        exclude: ^.*\.(ipynb)$
+      - id: ruff-format

.readthedocs.yaml

@@ -0,0 +1,19 @@
+# Read the Docs configuration file
+# See https://docs.readthedocs.io/en/stable/config-file/v2.html for details
+
+version: 2
+
+build:
+  os: ubuntu-22.04
+  tools:
+    python: "3.11"
+    rust: "1.70"
+
+sphinx:
+  configuration: docs/conf.py
+
+python:
+  install:
+    - requirements: docs/requirements-docs.txt
+    - method: pip
+      path: .

LICENSE

@@ -0,0 +1,202 @@
+
+                                 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.

Notice.txt

@@ -0,0 +1 @@
+Copyright 2023-2024 Bytedance Ltd. and/or its affiliates 

README.md

@@ -0,0 +1,269 @@
+<div align="center">
+ 👋 Hi, everyone! 
+    verl is a RL training library initiated by <b>ByteDance Seed team</b> and maintained by the verl community.
+    <br>
+    <br>
+</div>
+
+<div align="center">
+
+[<img src="https://devin.ai/assets/deepwiki-badge.png" alt="Ask DeepWiki.com" height="20"/>](https://deepwiki.com/volcengine/verl)
+[![GitHub Repo stars](https://img.shields.io/github/stars/volcengine/verl)](https://github.com/volcengine/verl/stargazers)
+[![Twitter](https://img.shields.io/twitter/follow/verl_project)](https://twitter.com/verl_project)
+<a href="https://join.slack.com/t/verlgroup/shared_invite/zt-2w5p9o4c3-yy0x2Q56s_VlGLsJ93A6vA"><img src="https://img.shields.io/badge/Slack-verl-blueviolet?logo=slack&amp"></a>
+<a href="https://arxiv.org/pdf/2409.19256"><img src="https://img.shields.io/static/v1?label=EuroSys&message=Paper&color=red"></a>
+[![Documentation](https://img.shields.io/badge/documentation-blue)](https://verl.readthedocs.io/en/latest/)
+<a href="https://raw.githubusercontent.com/eric-haibin-lin/verl-community/refs/heads/main/WeChat.JPG"><img src="https://img.shields.io/badge/微信-green?logo=wechat&amp"></a>
+
+</div>
+
+![seed logo](https://github.com/user-attachments/assets/c42e675e-497c-4508-8bb9-093ad4d1f216)
+
+<h1 style="text-align: center;">verl: Volcano Engine Reinforcement Learning for LLMs</h1>
+
+verl is a flexible, efficient and production-ready RL training library for large language models (LLMs).
+
+verl is the open-source version of **[HybridFlow: A Flexible and Efficient RLHF Framework](https://arxiv.org/abs/2409.19256v2)** paper.
+
+verl is flexible and easy to use with:
+
+- **Easy extension of diverse RL algorithms**: The hybrid-controller programming model enables flexible representation and efficient execution of complex post-training dataflows. Build RL dataflows such as GRPO, PPO in a few lines of code.
+
+- **Seamless integration of existing LLM infra with modular APIs**: Decouples computation and data dependencies, enabling seamless integration with existing LLM frameworks, such as FSDP, Megatron-LM, vLLM, SGLang, etc
+
+- **Flexible device mapping**: Supports various placement of models onto different sets of GPUs for efficient resource utilization and scalability across different cluster sizes.
+
+- Ready integration with popular HuggingFace models
+
+verl is fast with:
+
+- **State-of-the-art throughput**: SOTA LLM training and inference engine integrations and SOTA RL throughput.
+
+- **Efficient actor model resharding with 3D-HybridEngine**: Eliminates memory redundancy and significantly reduces communication overhead during transitions between training and generation phases.
+
+</p>
+
+## News
+
+- [2025/06] verl with Megatron backend enables large MoE models such as [DeepSeek-671b and Qwen3-236b](https://verl.readthedocs.io/en/latest/perf/dpsk.html).
+- [2025/06] verl team will provide latest project updates at [PyTorch Day China](https://www.lfasiallc.com/pytorch-day-china/) on June 7th. Meet our dev team in Beijing!
+- [2025/05] [PF-PPO](https://arxiv.org/abs/2409.06957), accepted to ICML 2025, is now supported in verl! PF-PPO enhances policy learning efficiency and robustness by filtering potentially noisy reward signals and reusing high-quality experiences via a replay buffer.
+- [2025/04] We will give a tutorial about latest post-training techniques and programming guide for verl at [ICLR 2025 Expo](https://iclr.cc/virtual/2025/calendar?filter_events=Expo+Talk+Panel&filter_rooms=), [SCI-FM workshop](https://open-foundation-model.github.io/) and [LMSys afterparty](https://lu.ma/d23nyynm). Talk materials available [here](https://github.com/eric-haibin-lin/verl-community/tree/main/iclr25).
+- [2025/04] [Seed-Thinking-v1.5](https://github.com/ByteDance-Seed/Seed-Thinking-v1.5/blob/main/seed-thinking-v1.5.pdf) tech report is released! Trained with verl, Seed-Thinking-v1.5 achieves 86.7 on AIME 2024, 55.0 on Codeforces and 77.3 on GPQA, demonstrating excellent reasoning abilities in STEM and coding. Beyond reasoning tasks, the method demonstrates notable generalization across diverse domains.
+- [2025/04] [VAPO](https://arxiv.org/pdf/2504.05118) (value-based augmented PPO) paper covers our latest RL method for reasoning models. Trained from Qwen-32B-base model, VAPO achieves 60.4 on AIME 2024, outperforming DAPO-32B.
+- [2025/03] verl v0.3.0.post1 is released! See [release note](https://github.com/volcengine/verl/releases/) for details. It achieves [~1.4x speedup](https://tongyx361.github.io/blogs/posts/verl-intro/#/verl-flexible-and-efficient-rl-for-llms) compared to prev versions.
+- [2025/03] [DAPO](https://dapo-sia.github.io/) is the open-sourced SOTA RL algorithm that achieves 50 points on AIME 2024 based on the Qwen2.5-32B pre-trained model, surpassing the previous SOTA achieved by DeepSeek's GRPO (DeepSeek-R1-Zero-Qwen-32B). DAPO's training is fully powered by verl and the reproduction code is available in `recipe/dapo` now.
+<details><summary> more... </summary>
+<ul>
+  <li>[2025/05] verl will be presented at [A2M Shanghai](https://a2m.msup.com.cn/home/?aid=4488&city=shanghai) on 5/16 - 5/17.</li>
+  <li>[2025/05] verl will be presented at [GOSIM x PyTorch Day 2025](https://paris2025.gosim.org/). See you in Paris! </li>
+  <li>[2025/03] We introduced the programming model of verl at the [vLLM Beijing Meetup](https://mp.weixin.qq.com/s/n77GibL2corAtQHtVEAzfg) and [verl intro and updates](https://github.com/eric-haibin-lin/verl-community/blob/main/slides/verl-lmsys-meetup.pdf) at the [SGLang-LMSYS Org Meetup](https://lu.ma/ntjrr7ig) in Sunnyvale mid-March.</li>
+  <li>[2025/03] We will present verl(HybridFlow) at EuroSys 2025. See you in Rotterdam!</li>
+  <li>[2025/02] verl v0.2.0.post2 is released!</li>
+  <li>[2025/02] We presented verl in the <a href="https://lu.ma/ji7atxux">Bytedance/NVIDIA/Anyscale Ray Meetup</a>. See you in San Jose!</li>
+  <li>[2025/01] [Doubao-1.5-pro](https://team.doubao.com/zh/special/doubao_1_5_pro) is released with SOTA-level performance on LLM & VLM. The RL scaling preview model is trained using verl, reaching OpenAI O1-level performance on math benchmarks (70.0 pass@1 on AIME).</li>
+  <li>[2024/12] verl is presented at Ray Forward 2024. Slides available <a href="https://github.com/eric-haibin-lin/verl-community/blob/main/slides/Ray_Forward_2024_%E5%B7%AB%E9%94%A1%E6%96%8C.pdf">here</a></li>
+  <li>[2024/12] The team presented <a href="https://neurips.cc/Expo/Conferences/2024/workshop/100677">Post-training LLMs: From Algorithms to Infrastructure</a> at NeurIPS 2024. <a href="https://github.com/eric-haibin-lin/verl-data/tree/neurips">Slides</a> and <a href="https://neurips.cc/Expo/Conferences/2024/workshop/100677">video</a> available.</li>
+  <li>[2024/10] verl is presented at Ray Summit. <a href="https://www.youtube.com/watch?v=MrhMcXkXvJU&list=PLzTswPQNepXntmT8jr9WaNfqQ60QwW7-U&index=37">Youtube video</a> available.</li>
+  <li>[2024/08] HybridFlow (verl) is accepted to EuroSys 2025.</li>
+</ul>   
+</details>
+
+## Key Features
+
+- **FSDP**, **FSDP2** and **Megatron-LM** for training.
+- **vLLM**, **SGLang** and **HF Transformers** for rollout generation.
+- Compatible with Hugging Face Transformers and Modelscope Hub: [Qwen-3](https://github.com/volcengine/verl/blob/main/examples/grpo_trainer/run_qwen3-8b.sh), Qwen-2.5, Llama3.1, Gemma2, DeepSeek-LLM, etc
+- Supervised fine-tuning.
+- Reinforcement learning with [PPO](examples/ppo_trainer/), [GRPO](examples/grpo_trainer/), [ReMax](examples/remax_trainer/), [REINFORCE++](https://verl.readthedocs.io/en/latest/examples/config.html#algorithm), [RLOO](examples/rloo_trainer/), [PRIME](recipe/prime/), [DAPO](recipe/dapo/), [DrGRPO](recipe/drgrpo), [KL_Cov & Clip_Cov](recipe/entropy) etc.
+  - Support model-based reward and function-based reward (verifiable reward) for math, [coding](https://github.com/volcengine/verl/tree/main/recipe/dapo), etc
+  - Support vision-language models (VLMs) and [multi-modal RL](examples/grpo_trainer/run_qwen2_5_vl-7b.sh) with Qwen2.5-vl, Kimi-VL
+  - [Multi-turn with tool calling](https://github.com/volcengine/verl/tree/main/examples/sglang_multiturn)
+- LLM alignment recipes such as [Self-play preference optimization (SPPO)](https://github.com/volcengine/verl/tree/main/recipe/sppo)
+- Flash attention 2, [sequence packing](examples/ppo_trainer/run_qwen2-7b_seq_balance.sh), [sequence parallelism](examples/ppo_trainer/run_deepseek7b_llm_sp2.sh) support via DeepSpeed Ulysses, [LoRA](examples/sft/gsm8k/run_qwen_05_peft.sh), [Liger-kernel](examples/sft/gsm8k/run_qwen_05_sp2_liger.sh).
+- Scales up to 671B models and hundreds of GPUs with [expert parallelism](https://github.com/volcengine/verl/pull/1467)
+- Multi-gpu [LoRA RL](https://verl.readthedocs.io/en/latest/advance/ppo_lora.html) support to save memory.
+- Experiment tracking with wandb, swanlab, mlflow and tensorboard.
+
+## Upcoming Features and Changes
+
+- Roadmap https://github.com/volcengine/verl/issues/710
+- DeepSeek 671b optimizations with Megatron v0.11 https://github.com/volcengine/verl/issues/708
+- Multi-turn rollout and tools using optimizations https://github.com/volcengine/verl/issues/1882
+- Environment interactions https://github.com/volcengine/verl/issues/1172
+- List of breaking changes since v0.3 https://github.com/volcengine/verl/discussions/943, entropy_coeff defaults to 0
+- Lora for RL https://github.com/volcengine/verl/pull/1127 
+
+## Getting Started
+
+<a href="https://verl.readthedocs.io/en/latest/index.html"><b>Documentation</b></a>
+
+**Quickstart:**
+
+- [Installation](https://verl.readthedocs.io/en/latest/start/install.html)
+- [Quickstart](https://verl.readthedocs.io/en/latest/start/quickstart.html)
+- [Programming Guide](https://verl.readthedocs.io/en/latest/hybrid_flow.html)
+- [PPO in verl](https://verl.readthedocs.io/en/latest/algo/ppo.html)
+- [GRPO in verl](https://verl.readthedocs.io/en/latest/algo/grpo.html)
+
+**Running a PPO example step-by-step:**
+
+
+- [Prepare Data for Post-Training](https://verl.readthedocs.io/en/latest/preparation/prepare_data.html)
+- [Implement Reward Function for Dataset](https://verl.readthedocs.io/en/latest/preparation/reward_function.html)
+- [PPO Example Architecture](https://verl.readthedocs.io/en/latest/examples/ppo_code_architecture.html)
+- [Config Explanation](https://verl.readthedocs.io/en/latest/examples/config.html)
+
+**Reproducible algorithm baselines:**
+
+- [RL performance on coding, math](https://verl.readthedocs.io/en/latest/algo/baseline.html)
+
+**For code explanation and advance usage (extension):**
+
+- PPO Trainer and Workers
+  - [PPO Ray Trainer](https://verl.readthedocs.io/en/latest/workers/ray_trainer.html)
+  - [PyTorch FSDP Backend](https://verl.readthedocs.io/en/latest/workers/fsdp_workers.html)
+  - [Megatron-LM Backend](https://verl.readthedocs.io/en/latest/index.html)
+
+- Advanced Usage and Extension
+  - [Add Models with the FSDP Backend](https://verl.readthedocs.io/en/latest/advance/fsdp_extension.html)
+  - [Add Models with the Megatron-LM Backend](https://verl.readthedocs.io/en/latest/advance/megatron_extension.html)
+  - [Multi-turn Rollout Support](https://verl.readthedocs.io/en/latest/sglang_multiturn/multiturn.html)
+  - [Search Tool Integration](https://verl.readthedocs.io/en/latest/sglang_multiturn/search_tool_example.html)
+  - [Sandbox Fusion Integration](https://verl.readthedocs.io/en/latest/examples/sandbox_fusion_example.html)
+  - [Deployment using Separate GPU Resources](https://github.com/volcengine/verl/tree/main/examples/split_placement)
+  - [Extend to Other RL(HF) algorithms](https://verl.readthedocs.io/en/latest/advance/dpo_extension.html)
+  - [Ray API design tutorial](https://verl.readthedocs.io/en/latest/advance/placement.html)
+
+**Blogs from the community**
+
+- [SGLang, verl, OpenBMB and Tsinghua University: Pioneering End-to-End Multi-Turn RLHF](https://github.com/zhaochenyang20/Awesome-ML-SYS-Tutorial/blob/main/rlhf/verl/multi-turn/verl-multiturn-rollout-Release.md)
+- [Reinforcement Learning from Human Feedback on AMD GPUs with verl and ROCm Integration](https://rocm.blogs.amd.com/artificial-intelligence/verl-large-scale/README.html)
+- [veMLP x verl ：玩转强化学习训练](https://mp.weixin.qq.com/s/7nbqxk4knMGd-hQE9ls2tA)
+- [使用 verl 进行 GRPO 分布式强化学习训练最佳实践](https://www.volcengine.com/docs/6459/1463942)
+- [HybridFlow verl 原文浅析](https://github.com/zhaochenyang20/Awesome-ML-SYS-Tutorial/blob/main/rlhf/verl/readme.md)
+- [最高提升 20 倍吞吐量！豆包大模型团队发布全新 RLHF 框架，现已开源！](https://team.doubao.com/en/blog/%E6%9C%80%E9%AB%98%E6%8F%90%E5%8D%8720%E5%80%8D%E5%90%9E%E5%90%90%E9%87%8F-%E8%B1%86%E5%8C%85%E5%A4%A7%E6%A8%A1%E5%9E%8B%E5%9B%A2%E9%98%9F%E5%8F%91%E5%B8%83%E5%85%A8%E6%96%B0-rlhf-%E6%A1%86%E6%9E%B6-%E7%8E%B0%E5%B7%B2%E5%BC%80%E6%BA%90)
+
+## Performance Tuning Guide
+
+The performance is essential for on-policy RL algorithm. We have written a detailed [performance tuning guide](https://verl.readthedocs.io/en/latest/perf/perf_tuning.html) to help you optimize performance.
+
+## Upgrade to vLLM >= v0.8.2
+
+verl now supports vLLM>=0.8.2 when using FSDP as the training backend. Please refer to [this document](https://github.com/volcengine/verl/blob/main/docs/README_vllm0.8.md) for the installation guide and more information. Please avoid vllm 0.7.x, which contains bugs that may lead to OOMs and unexpected errors.
+
+## Use Latest SGLang
+
+SGLang is fully supported with verl, and SGLang RL Group is working extensively on building unique features, including multi-turn agentic RL, VLM RLHF, server-based RL, and partial rollout. Please refer to [this document](https://verl.readthedocs.io/en/latest/workers/sglang_worker.html) for the installation guide and more information.
+
+## Upgrade to FSDP2
+
+verl is fully embracing FSDP2! FSDP2 is recommended by torch distributed team, providing better throughput and memory usage, and is composible with other features (e.g. torch.compile). To enable FSDP2, simply use verl main and set the following options:
+```
+actor_rollout_ref.ref.strategy=fsdp2
+actor_rollout_ref.actor.strategy=fsdp2
+critic.strategy=fsdp2 
+reward_model.strategy=fsdp2 
+```
+Furthermore, FSDP2 cpu offloading is compatible with gradient accumulation. You can turn it on to save memory with `actor_rollout_ref.actor.fsdp_config.offload_policy=True`. For more details, see https://github.com/volcengine/verl/pull/1026
+
+## AMD Support (ROCm Kernel)
+
+verl now supports FSDP as the training engine (Megatron support coming soon) and both integrates with vLLM and SGLang as inference engines. Please refer to [this document](https://github.com/volcengine/verl/blob/main/docs/amd_tutorial/amd_build_dockerfile_page.rst) for the installation guide and more information, and [this document](https://github.com/volcengine/verl/blob/main/docs/amd_tutorial/amd_vllm_page.rst) for the vLLM performance tuning for ROCm.
+
+
+## Citation and acknowledgement
+
+If you find the project helpful, please cite:
+
+- [HybridFlow: A Flexible and Efficient RLHF Framework](https://arxiv.org/abs/2409.19256v2)
+- [A Framework for Training Large Language Models for Code Generation via Proximal Policy Optimization](https://i.cs.hku.hk/~cwu/papers/gmsheng-NL2Code24.pdf)
+
+```bibtex
+@article{sheng2024hybridflow,
+  title   = {HybridFlow: A Flexible and Efficient RLHF Framework},
+  author  = {Guangming Sheng and Chi Zhang and Zilingfeng Ye and Xibin Wu and Wang Zhang and Ru Zhang and Yanghua Peng and Haibin Lin and Chuan Wu},
+  year    = {2024},
+  journal = {arXiv preprint arXiv: 2409.19256}
+}
+```
+
+verl is inspired by the design of Nemo-Aligner, Deepspeed-chat and OpenRLHF. The project is adopted and contributed by Bytedance, Anyscale, LMSys.org, [Alibaba Qwen team](https://github.com/QwenLM/), Shanghai AI Lab, Tsinghua University, UC Berkeley, UCLA, UIUC, University of Hong Kong, ke.com, [All Hands AI](https://www.all-hands.dev/), [ModelBest](http://modelbest.cn/), OpenPipe, JD AI Lab, Microsoft Research, [StepFun](https://www.stepfun.com/), Amazon, LinkedIn, Meituan, [Camel-AI](https://www.camel-ai.org/), [OpenManus](https://github.com/OpenManus), Xiaomi, Prime Intellect, NVIDIA research, [Baichuan](https://www.baichuan-ai.com/home), [RedNote](https://www.xiaohongshu.com/), [SwissAI](https://www.swiss-ai.org/), [Moonshot AI (Kimi)](https://www.moonshot-ai.com/), Baidu, Snowflake, [IceSword Lab](https://www.iceswordlab.com), and many more.
+
+## Awesome work using verl
+
+- [TinyZero](https://github.com/Jiayi-Pan/TinyZero): a reproduction of **DeepSeek R1 Zero** recipe for reasoning tasks ![GitHub Repo stars](https://img.shields.io/github/stars/Jiayi-Pan/TinyZero)
+- [SkyThought](https://github.com/NovaSky-AI/SkyThought): RL training for Sky-T1-7B by NovaSky AI team. ![GitHub Repo stars](https://img.shields.io/github/stars/NovaSky-AI/SkyThought)
+- [simpleRL-reason](https://github.com/hkust-nlp/simpleRL-reason): SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wild ![GitHub Repo stars](https://img.shields.io/github/stars/hkust-nlp/simpleRL-reason)
+- [Easy-R1](https://github.com/hiyouga/EasyR1): **Multi-modal** RL training framework ![GitHub Repo stars](https://img.shields.io/github/stars/hiyouga/EasyR1)
+- [OpenManus-RL](https://github.com/OpenManus/OpenManus-RL): LLM Agents RL tunning framework for multiple agent environments. ![GitHub Repo stars](https://img.shields.io/github/stars/OpenManus/OpenManus-RL)
+- [rllm](https://github.com/agentica-project/rllm): async RL training with [verl-pipeline](https://github.com/agentica-project/verl-pipeline) ![GitHub Repo stars](https://img.shields.io/github/stars/agentica-project/rllm)
+- [PRIME](https://github.com/PRIME-RL/PRIME): Process reinforcement through implicit rewards ![GitHub Repo stars](https://img.shields.io/github/stars/PRIME-RL/PRIME)
+- [RAGEN](https://github.com/ZihanWang314/ragen): a general-purpose reasoning **agent** training framework ![GitHub Repo stars](https://img.shields.io/github/stars/ZihanWang314/ragen)
+- [Search-R1](https://github.com/PeterGriffinJin/Search-R1): RL with reasoning and **searching (tool-call)** interleaved LLMs ![GitHub Repo stars](https://img.shields.io/github/stars/PeterGriffinJin/Search-R1)
+- [DeepRetrieval](https://github.com/pat-jj/DeepRetrieval): RL Training of **Search Agent** with **Search/Retrieval Outcome** ![GitHub Repo stars](https://img.shields.io/github/stars/pat-jj/DeepRetrieval)
+- [ReSearch](https://github.com/Agent-RL/ReSearch): Learning to **Re**ason with **Search** for LLMs via Reinforcement Learning ![GitHub Repo stars](https://img.shields.io/github/stars/Agent-RL/ReSearch)
+- [Code-R1](https://github.com/ganler/code-r1): Reproducing R1 for **Code** with Reliable Rewards ![GitHub Repo stars](https://img.shields.io/github/stars/ganler/code-r1)
+- [Skywork-OR1](https://github.com/SkyworkAI/Skywork-OR1): Skywork open reaonser series ![GitHub Repo stars](https://img.shields.io/github/stars/SkyworkAI/Skywork-OR1)
+- [ToRL](https://github.com/GAIR-NLP/ToRL): Scaling tool-integrated RL ![GitHub Repo stars](https://img.shields.io/github/stars/GAIR-NLP/ToRL)
+- [verl-agent](https://github.com/langfengQ/verl-agent): A scalable training framework for **long-horizon LLM/VLM agents**, along with a new algorithm **GiGPO** ![GitHub Repo stars](https://img.shields.io/github/stars/langfengQ/verl-agent)
+- [PF-PPO](https://arxiv.org/abs/2409.06957): Policy Filtration for PPO based on the reliability of reward signals for more efficient and robust RLHF.
+- [GUI-R1](https://github.com/ritzz-ai/GUI-R1): **GUI-R1**: A Generalist R1-style Vision-Language Action Model For **GUI Agents** ![GitHub Repo stars](https://img.shields.io/github/stars/ritzz-ai/GUI-R1)
+- [DeepResearcher](https://github.com/GAIR-NLP/DeepResearcher): Scaling deep research via reinforcement learning in real-world environments ![GitHub Repo stars](https://img.shields.io/github/stars/GAIR-NLP/DeepResearcher)
+- [VAGEN](https://github.com/RAGEN-AI/VAGEN): Training VLM agents with multi-turn reinforcement learning ![GitHub Repo stars](https://img.shields.io/github/stars/RAGEN-AI/VAGEN)
+- [ReTool](https://retool-rl.github.io/): ReTool: reinforcement learning for strategic tool use in LLMs. Code release is in progress...
+- [RM-R1](https://arxiv.org/abs/2505.02387): RL training of reasoning reward models ![GitHub Repo stars](https://img.shields.io/github/stars/RM-R1-UIUC/RM-R1)
+- [Absolute Zero Reasoner](https://arxiv.org/abs/2505.03335): A no human curated data self-play framework for reasoning![GitHub Repo stars](https://img.shields.io/github/stars/LeapLabTHU/Absolute-Zero-Reasoner)
+- [LUFFY](https://arxiv.org/pdf/2504.14945): Learning to Reason under Off-Policy Guidance![GitHub Repo stars](https://img.shields.io/github/stars/ElliottYan/LUFFY)
+- [verl-tool](https://github.com/TIGER-AI-Lab/verl-tool): An unified and easy-to-extend tool-agent training framework based on verl![GitHub Repo stars](https://img.shields.io/github/stars/TIGER-AI-Lab/verl-tool)
+- [DeepMath](https://github.com/zwhe99/DeepMath): DeepMath-103K data and series models for math reasoning![GitHub Repo stars](https://img.shields.io/github/stars/zwhe99/DeepMath)
+- [Entropy Mechanism of RL](https://github.com/PRIME-RL/Entropy-Mechanism-of-RL): The Entropy Mechanism of Reinforcement Learning for Large Language Model Reasoning![GitHub Repo stars](https://img.shields.io/github/stars/PRIME-RL/Entropy-Mechanism-of-RL)
+- [LLaSA-TTS-GRPO](https://github.com/channel-io/ch-tts-llasa-rl-grpo): TTS fine-tuning with GRPO optimization based on LLASA models ![GitHub Repo stars](https://img.shields.io/github/stars/channel-io/ch-tts-llasa-rl-grpo)
+- [RL-Factory](https://github.com/Simple-Efficient/RL-Factory): An easy and efficient RL post-training framework for Agentic Learning ![GitHub Repo stars](https://img.shields.io/github/stars/Simple-Efficient/RL-Factory)
+- [RACRO](https://github.com/gyhdog99/RACRO2): Build multi-modal reasoning models via decoupling it into query-conditioned captioning and text-only reasoning ![GitHub Repo stars](https://img.shields.io/github/stars/gyhdog99/RACRO2)
+
+and many more awesome work listed in [recipe](recipe/README.md).
+## Contribution Guide
+
+Contributions from the community are welcome! Please check out our [project roadmap](https://github.com/volcengine/verl/issues/710) and [good first issues](https://github.com/volcengine/verl/issues?q=is%3Aissue%20state%3Aopen%20label%3A%22good%20first%20issue%22) to see where you can contribute.
+
+### Code Linting and Formatting
+
+We use pre-commit to help improve code quality. To initialize pre-commit, run:
+
+```bash
+pip install pre-commit
+pre-commit install
+```
+
+To resolve CI errors locally, you can manually run pre-commit by:
+
+```bash
+pre-commit run
+```
+
+### Adding CI tests
+
+If possible, please add CI test(s) for your new feature:
+
+1. Find the most relevant workflow yml file, which usually corresponds to a `hydra` default config (e.g. `ppo_trainer`, `ppo_megatron_trainer`, `sft_trainer`, etc).
+2. Add related path patterns to the `paths` section if not already included.
+3. Minimize the workload of the test script(s) (see existing scripts for examples).
+
+## About [ByteDance Seed Team](https://team.doubao.com/)
+
+Founded in 2023, ByteDance Seed Team is dedicated to crafting the industry's most advanced AI foundation models. The team aspires to become a world-class research team and make significant contributions to the advancement of science and society. You can get to know Bytedance Seed better through the following channels👇
+<div>
+  <a href="https://team.doubao.com/">
+    <img src="https://img.shields.io/badge/Website-%231e37ff?style=for-the-badge&logo=bytedance&logoColor=white"></a>
+  <a href="https://github.com/user-attachments/assets/469535a8-42f2-4797-acdf-4f7a1d4a0c3e">
+    <img src="https://img.shields.io/badge/WeChat-07C160?style=for-the-badge&logo=wechat&logoColor=white"></a>
+ <a href="https://www.xiaohongshu.com/user/profile/668e7e15000000000303157d?xsec_token=ABl2-aqekpytY6A8TuxjrwnZskU-6BsMRE_ufQQaSAvjc%3D&xsec_source=pc_search">
+    <img src="https://img.shields.io/badge/Xiaohongshu-%23FF2442?style=for-the-badge&logo=xiaohongshu&logoColor=white"></a>
+  <a href="https://www.zhihu.com/org/dou-bao-da-mo-xing-tuan-dui/">
+    <img src="https://img.shields.io/badge/zhihu-%230084FF?style=for-the-badge&logo=zhihu&logoColor=white"></a>
+
+</div>
+---
+
+We are HIRING! Send us an [email](mailto:[email protected]) if you are interested in internship/FTE opportunities in RL for agents.

docker/Apptainerfile.rocm

@@ -0,0 +1,57 @@
+Bootstrap: docker
+
+# Support - Traing: fsdp; Inference: vllm
+# FROM: rocm/vllm:rocm6.2_mi300_ubuntu20.04_py3.9_vllm_0.6.4
+# Support - Traing: fsdp; Inference: vllm, sglang
+FROM lmsysorg/sglang:v0.4.5-rocm630
+
+%environment
+    export PYTORCH_ROCM_ARCH="gfx90a;gfx942"
+
+    export HIPCC_COMPILE_FLAGS_APPEND="--amdgpu-target=gfx90a;gfx942 -D__HIP_PLATFORM_AMD__"
+    export CFLAGS="-D__HIP_PLATFORM_AMD__"
+    export CXXFLAGS="-D__HIP_PLATFORM_AMD__"
+
+%post
+    # Create source directory
+    mkdir -p /opt/src
+
+    # Uninstall and reinstall vllm
+    pip uninstall -y vllm
+    cd /opt/src
+    git clone -b v0.6.3 https://github.com/vllm-project/vllm.git
+    cd vllm
+    MAX_JOBS=$(nproc) python3 setup.py install
+    cd /opt
+    rm -rf /opt/src/vllm
+
+    # Install dependencies
+    pip install "tensordict<0.6" --no-deps
+    pip install accelerate \
+        codetiming \
+        datasets \
+        dill \
+        hydra-core \
+        liger-kernel \
+        numpy \
+        pandas \
+        peft \
+        "pyarrow>=15.0.0" \
+        pylatexenc \
+        "ray[data,train,tune,serve]" \
+        torchdata \
+        transformers \
+        wandb \
+        orjson \
+        pybind11
+
+    # Clone and install verl from GitHub
+    cd /opt
+    git clone https://github.com/volcengine/verl.git
+    cd verl
+    # Uncomment to use a specific version
+    # git checkout v0.3.0.post0
+    pip install -e . --no-deps
+
+    # Install torch_memory_saver
+    pip install git+https://github.com/ExtremeViscent/torch_memory_saver.git --no-deps

docker/Dockerfile.awsefa

@@ -0,0 +1,53 @@
+FROM whatcanyousee/verl:ngc-cu124-vllm0.8.5-sglang0.4.6-mcore0.12.0-te2.3
+
+# For aws instances with EFA net interface (Sagemaker AI Pod)
+#     install EFA driver:
+######## AWS EFA ############
+ENV NCCL_VERSION=2.25.1-1
+ENV DEBIAN_FRONTEND=noninteractive
+ENV EFA_INSTALLER_VERSION=1.40.0
+ENV AWS_OFI_NCCL_VERSION=1.14.2
+ENV FI_EFA_SET_CUDA_SYNC_MEMOPS=0
+ENV FI_PROVIDER=efa
+
+RUN apt update && apt install -y linux-image-generic libhwloc-dev
+
+RUN cd /tmp && \
+    curl -O https://efa-installer.amazonaws.com/aws-efa-installer-${EFA_INSTALLER_VERSION}.tar.gz  && \
+    tar -xf aws-efa-installer-${EFA_INSTALLER_VERSION}.tar.gz && \
+    cd aws-efa-installer && \
+    ./efa_installer.sh -y -g --skip-kmod --skip-limit-conf --no-verify && \
+    ldconfig && \
+    rm -rf /tmp/aws-efa-installer /var/lib/apt/lists/*
+
+# NCCL EFA Plugin
+RUN cd /tmp && \
+    curl -LO https://github.com/aws/aws-ofi-nccl/archive/refs/tags/v${AWS_OFI_NCCL_VERSION}.tar.gz && \
+    tar -xzf /tmp/v${AWS_OFI_NCCL_VERSION}.tar.gz && \
+    rm /tmp/v${AWS_OFI_NCCL_VERSION}.tar.gz && \
+    mv aws-ofi-nccl-${AWS_OFI_NCCL_VERSION} aws-ofi-nccl && \
+    cd /tmp/aws-ofi-nccl && \
+    ./autogen.sh && \
+    ./configure --prefix=/opt/amazon/efa \
+    --with-libfabric=/opt/amazon/efa \
+    --with-cuda=/usr/local/cuda \
+    --enable-platform-aws \
+    --with-mpi=/opt/amazon/openmpi && \
+    make -j$(nproc) install && \
+    rm -rf /tmp/aws-ofi/nccl
+
+# NCCL
+RUN echo "/usr/local/lib"      >> /etc/ld.so.conf.d/local.conf && \
+    echo "/opt/amazon/openmpi/lib" >> /etc/ld.so.conf.d/efa.conf && \
+    ldconfig
+
+ENV OMPI_MCA_pml=^cm,ucx            \
+    OMPI_MCA_btl=tcp,self           \
+    OMPI_MCA_btl_tcp_if_exclude=lo,docker0,veth_def_agent \
+    OPAL_PREFIX=/opt/amazon/openmpi \
+    NCCL_SOCKET_IFNAME=^docker,lo,veth_def_agent  \
+    FI_EFA_USE_HUGE_PAGE=0
+
+# docker build -t whatcanyousee/verl:awsefa --label "commit=$(git rev-parse --short HEAD)" .
+# on aws:
+# docker run --ipc=host --privileged --name verldev --gpus all --network=host --shm-size=1800gb -itd whatcanyousee/verl:awsefa

docker/Dockerfile.ngc.vllm

@@ -0,0 +1,47 @@
+# docker buildx build --platform linux/x86_64 -t "verlai/verl:ngc-th2.4.0-cu124-vllm0.6.3-ray2.4-te1.7-v0.0.6" -f docker/Dockerfile.ngc.vllm . --builder cloud-verlai-verl-builder --progress=plain --push
+FROM nvcr.io/nvidia/pytorch:24.05-py3
+
+# uninstall nv-pytorch fork
+RUN pip3 uninstall pytorch-quantization \
+    pytorch-triton \
+    torch \
+    torch-tensorrt \
+    torchvision \
+    xgboost transformer_engine flash_attn \
+    apex megatron-core -y
+
+RUN pip3 install torch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 --index-url https://download.pytorch.org/whl/cu124
+
+# =============== Megatron dependencies (optional) =================
+# install apex, set MAX_JOBS to avoid OOMs
+RUN MAX_JOBS=4 pip3 install -v --disable-pip-version-check --no-cache-dir --no-build-isolation \
+    --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" \
+    git+https://github.com/NVIDIA/apex
+# =============== End of Megatron dependencies (optional) =================
+
+RUN pip3 install --no-cache-dir \
+    accelerate \
+    codetiming \
+    datasets \
+    dill \
+    hydra-core \
+    numpy \
+    'pandas' \
+    'peft' \
+    'pyarrow>=15.0.0' \
+    'pybind11' \
+    'pylatexenc' \
+    'ray>=2.10' \
+    'tensordict<0.6' \
+    'transformers' \
+    'vllm==0.6.3.post1' \
+    'wandb'
+
+# full dependencies
+RUN pip3 install pytest pre-commit py-spy pyext liger-kernel
+
+# =============== Megatron dependencies (optional) =================
+# install Transformer Engine, which requires FA 2.5.8. Do it in a separate step for docker cache
+RUN MAX_JOBS=4 NINJA_FLAGS="-j4" pip3 install flash-attn==2.5.8 --no-cache-dir --no-build-isolation
+RUN MAX_JOBS=1 NINJA_FLAGS="-j1" TE_BUILD_WITH_NINJA=0 pip3 install git+https://github.com/eric-haibin-lin/[email protected]
+# =============== End of Megatron dependencies (optional) =================

docker/Dockerfile.ngc.vllm0.8

@@ -0,0 +1,75 @@
+# Start from the NVIDIA official image (ubuntu-22.04 + cuda-12.6 + python-3.10)
+# https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes/rel-24-08.html
+FROM nvcr.io/nvidia/pytorch:24.08-py3
+
+# Define environments
+ENV MAX_JOBS=32
+ENV VLLM_WORKER_MULTIPROC_METHOD=spawn
+ENV DEBIAN_FRONTEND=noninteractive
+ENV NODE_OPTIONS=""
+ENV PIP_ROOT_USER_ACTION=ignore
+ENV HF_HUB_ENABLE_HF_TRANSFER="1"
+
+# Define installation arguments
+ARG APT_SOURCE=https://mirrors.tuna.tsinghua.edu.cn/ubuntu/
+ARG PIP_INDEX=https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple
+
+# Set apt source
+RUN cp /etc/apt/sources.list /etc/apt/sources.list.bak && \
+    { \
+    echo "deb ${APT_SOURCE} jammy main restricted universe multiverse"; \
+    echo "deb ${APT_SOURCE} jammy-updates main restricted universe multiverse"; \
+    echo "deb ${APT_SOURCE} jammy-backports main restricted universe multiverse"; \
+    echo "deb ${APT_SOURCE} jammy-security main restricted universe multiverse"; \
+    } > /etc/apt/sources.list
+
+# Install systemctl
+RUN apt-get update && \
+    apt-get install -y -o Dpkg::Options::="--force-confdef" systemd && \
+    apt-get clean
+
+# Install tini
+RUN apt-get update && \
+    apt-get install -y tini && \
+    apt-get clean
+
+# Change pip source
+RUN pip config set global.index-url "${PIP_INDEX}" && \
+    pip config set global.extra-index-url "${PIP_INDEX}" && \
+    python -m pip install --upgrade pip
+
+# Uninstall nv-pytorch fork
+RUN pip uninstall -y torch torchvision torchaudio \
+    pytorch-quantization pytorch-triton torch-tensorrt \
+    xgboost transformer_engine flash_attn apex megatron-core grpcio
+
+# Install torch-2.6.0+cu124 + vllm-0.8.3
+# torch-2.6.0+cu124: cxx11abi=False
+# torch-2.6.0+cu126: cxx11abi=True
+# see https://github.com/flashinfer-ai/flashinfer/issues/911
+RUN pip install --no-cache-dir "vllm==0.8.3" "torch==2.6.0" "torchvision==0.21.0" "torchaudio==2.6.0" "tensordict==0.6.2" torchdata \
+    "transformers[hf_xet]>=4.51.0" accelerate datasets peft hf-transfer \
+    "numpy<2.0.0" "pyarrow>=15.0.0" pandas \
+    ray[default] codetiming hydra-core pylatexenc qwen-vl-utils wandb dill pybind11 liger-kernel mathruler \
+    pytest py-spy pyext pre-commit ruff
+
+# Install flash-attn-2.7.4.post1 (cxx11abi=False)
+RUN wget -nv https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.4.post1/flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl && \
+    pip install --no-cache-dir flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
+
+# Install flashinfer-0.2.2.post1+cu124 (cxx11abi=False)
+# vllm-0.8.3 does not support flashinfer>=0.2.3
+# see https://github.com/vllm-project/vllm/pull/15777
+RUN wget -nv https://github.com/flashinfer-ai/flashinfer/releases/download/v0.2.2.post1/flashinfer_python-0.2.2.post1+cu124torch2.6-cp38-abi3-linux_x86_64.whl && \
+    pip install --no-cache-dir flashinfer_python-0.2.2.post1+cu124torch2.6-cp38-abi3-linux_x86_64.whl
+
+# Fix packages
+RUN pip uninstall -y pynvml nvidia-ml-py && \
+    pip install --no-cache-dir --upgrade "nvidia-ml-py>=12.560.30" "fastapi[standard]>=0.115.0" "optree>=0.13.0" "pydantic>=2.9" "grpcio>=1.62.1"
+
+# Install verl
+RUN pip install --no-cache-dir verl[vllm] -U
+
+# Reset pip config
+RUN pip config unset global.index-url && \
+    pip config unset global.extra-index-url

docker/Dockerfile.ngc.vllm0.8.sagemaker

@@ -0,0 +1,46 @@
+# Using a pre-built image from AWS DLC which contains the current version of python (3.10) and supported cuda version (12.1)
+FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/huggingface-pytorch-training:2.1.0-transformers4.36.0-gpu-py310-cu121-ubuntu20.04
+
+# uninstall nv-pytorch fork
+RUN pip3 uninstall -y pytorch-quantization \
+    pytorch-triton torch torch-tensorrt torchvision \
+    xgboost transformer_engine flash_attn apex megatron-core
+
+# Define environments
+ENV MAX_JOBS=32
+ENV VLLM_WORKER_MULTIPROC_METHOD=spawn
+ENV DEBIAN_FRONTEND=noninteractive
+ENV NODE_OPTIONS=""
+ENV HF_HUB_ENABLE_HF_TRANSFER="1"
+
+# Install systemctl
+RUN apt-get update && \
+    apt-get install -y -o Dpkg::Options::="--force-confdef" systemd && \
+    apt-get clean
+
+# Install tini
+RUN apt-get update && \
+    apt-get install -y tini && \
+    apt-get clean
+
+# Install torch-2.6.0 + vllm-0.8.2
+RUN pip install --no-cache-dir vllm==0.8.2 torch==2.6.0 torchvision==0.21.0 torchaudio==2.6.0 tensordict torchdata==0.11.0 \
+    transformers>=4.49.0 accelerate datasets peft hf-transfer \
+    ray[default] codetiming hydra-core pandas pyarrow>=15.0.0 pylatexenc qwen-vl-utils wandb dill pybind11 liger-kernel mathruler \
+    pytest pre-commit py-spy pyext ruff
+
+# Install flash_attn-2.7.4.post1
+RUN pip uninstall -y transformer-engine flash-attn && \
+    pip install flash-attn==2.7.4.post1 --no-build-isolation
+
+# Fix cv2
+RUN pip uninstall -y pynvml nvidia-ml-py && \
+    pip install --no-cache-dir nvidia-ml-py>=12.560.30 opencv-python-headless==4.8.0.74 fastapi==0.115.6 && \
+    pip install --no-cache-dir --upgrade optree>=0.13.0
+
+# Install verl
+RUN pip install --no-cache-dir verl[vllm] -U
+
+# Reset pip config
+RUN pip config unset global.index-url && \
+    pip config unset global.extra-index-url

docker/Dockerfile.rocm

@@ -0,0 +1,55 @@
+#  Build the docker in the repo dir:
+# docker build -f docker/Dockerfile.rocm -t verl-rocm:03.04.2015 .
+# docker images # you can find your built docker
+
+
+# Support - Traing: fsdp; Inference: vllm
+# FROM rocm/vllm:rocm6.2_mi300_ubuntu20.04_py3.9_vllm_0.6.4
+# Support - Traing: fsdp; Inference: vllm, sglang
+FROM lmsysorg/sglang:v0.4.6.post5-rocm630
+
+# Set working directory
+# WORKDIR $PWD/app
+
+# Set environment variables
+ENV PYTORCH_ROCM_ARCH="gfx90a;gfx942"
+
+ENV HIPCC_COMPILE_FLAGS_APPEND="--amdgpu-target=gfx90a;gfx942 -D__HIP_PLATFORM_AMD__"
+ENV CFLAGS="-D__HIP_PLATFORM_AMD__"
+ENV CXXFLAGS="-D__HIP_PLATFORM_AMD__"
+
+# Install vllm
+RUN pip uninstall -y vllm && \
+    rm -rf vllm && \
+    git clone -b v0.6.3 https://github.com/vllm-project/vllm.git && \
+    cd vllm && \
+    MAX_JOBS=$(nproc) python3 setup.py install && \
+    cd .. && \
+    rm -rf vllm
+
+# Copy the entire project directory
+COPY . .
+
+# Install dependencies
+RUN pip install "tensordict==0.6.2" --no-deps && \
+    pip install accelerate \
+    codetiming \
+    datasets \
+    dill \
+    hydra-core \
+    liger-kernel \
+    numpy \
+    pandas \
+    peft \
+    "pyarrow>=15.0.0" \
+    pylatexenc \
+    "ray[data,train,tune,serve]<2.45.0" \
+    torchdata \
+    transformers \
+    wandb \
+    orjson \
+    pybind11 && \
+    pip install -e . --no-deps
+
+# Install torch_memory_saver
+RUN pip install git+https://github.com/ExtremeViscent/torch_memory_saver.git --no-deps

docker/Dockerfile.sglang

@@ -0,0 +1,55 @@
+# Start from the NVIDIA official image (ubuntu-22.04 + python-3.10)
+# https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes/rel-24-08.html
+FROM nvcr.io/nvidia/pytorch:24.08-py3
+
+# Define environments
+ENV MAX_JOBS=32
+ENV DEBIAN_FRONTEND=noninteractive
+ENV NODE_OPTIONS=""
+
+# Define installation arguments
+ARG APT_SOURCE=https://mirrors.ustc.edu.cn/ubuntu/
+
+# Set apt source
+RUN cp /etc/apt/sources.list /etc/apt/sources.list.bak && \
+    { \
+    echo "deb ${APT_SOURCE} jammy main restricted universe multiverse"; \
+    echo "deb ${APT_SOURCE} jammy-updates main restricted universe multiverse"; \
+    echo "deb ${APT_SOURCE} jammy-backports main restricted universe multiverse"; \
+    echo "deb ${APT_SOURCE} jammy-security main restricted universe multiverse"; \
+    } > /etc/apt/sources.list
+
+# Install systemctl
+RUN apt-get update && \
+    apt-get install -y -o Dpkg::Options::="--force-confdef" systemd && \
+    apt-get clean
+
+# Install tini
+RUN apt-get update && \
+    apt-get install -y tini && \
+    apt-get clean
+
+# Change pip source
+ARG PIP_INDEX=https://mirrors.aliyun.com/pypi/simple/
+
+RUN pip config set global.index-url "${PIP_INDEX}" && \
+    pip config set global.extra-index-url "${PIP_INDEX}" && \
+    python -m pip install --upgrade pip
+
+# Install sglang-0.4.6.post5 and torch-memory-saver
+RUN pip uninstall -y cuda-python && pip install "sglang[all]==0.4.6.post5" --no-cache-dir --find-links https://flashinfer.ai/whl/cu124/torch2.6/flashinfer-python && pip install torch-memory-saver --no-cache-dir
+
+# Install torch-2.6.0
+RUN pip install --no-cache-dir torch==2.6.0 torchvision==0.21.0 torchaudio==2.6.0 tensordict torchdata \
+    transformers>=4.49.0 accelerate datasets peft hf_transfer \
+    ray[default] codetiming hydra-core pandas pyarrow>=15.0.0 pylatexenc qwen-vl-utils wandb liger-kernel \
+    pytest pre-commit py-spy pyext
+
+# Install flash_attn-2.7.4.post1
+RUN pip uninstall -y transformer-engine flash-attn && \
+    wget -v https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.4.post1/flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl && \
+    pip install --no-cache-dir flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
+
+# Fix cv2
+RUN pip uninstall -y pynvml nvidia-ml-py && \
+    pip install --no-cache-dir nvidia-ml-py>=12.560.30 opencv-python-headless==4.8.0.74 fastapi==0.115.6

docker/Dockerfile.vemlp.vllm.te

@@ -0,0 +1,41 @@
+# docker buildx build --platform linux/x86_64 -t "verlai/verl:$TAG" -f docker/$FILE .
+
+# the one in docker.io is an alias for the one veturbo
+# FROM vemlp-cn-beijing.cr.volces.com/veturbo/pytorch:2.4-cu124
+FROM docker.io/haibinlin/verl:v0.0.5-th2.4.0-cu124-base
+
+# only config pip index with https://pypi.tuna.tsinghua.edu.cn/simple if needed
+# unset for now
+RUN pip3 config unset global.index-url
+
+# transformers 4.47.0 contains the following bug:
+# AttributeError: 'Gemma2Attention' object has no attribute '_flash_attn_uses_top_left_mask'
+RUN pip3 install --no-cache-dir \
+    torch==2.4.0 \
+    accelerate \
+    codetiming \
+    dill \
+    hydra-core \
+    numpy \
+    pybind11 \
+    tensordict \
+    "transformers <= 4.46.0"
+
+RUN pip3 install --no-cache-dir flash-attn==2.7.0.post2 --no-build-isolation
+
+# vllm depends on ray
+RUN pip3 install --no-cache-dir vllm==0.6.3 ray==2.10
+
+# install apex
+RUN MAX_JOBS=4 pip3 install -v --disable-pip-version-check --no-cache-dir --no-build-isolation \
+    --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" \
+    git+https://github.com/NVIDIA/apex
+
+# install Transformer Engine
+# - flash-attn pinned to 2.5.3 by TransformerEngine, switch to eric-haibin-lin/[email protected] to relax version req
+# - install with: MAX_JOBS=1 NINJA_FLAGS="-j1" TE_BUILD_WITH_NINJA=0 to avoid OOM
+# - cudnn is required by TransformerEngine
+# RUN CUDNN_PATH=/opt/conda/lib/python3.11/site-packages/nvidia/cudnn \
+#     pip3 install git+https://github.com/eric-haibin-lin/[email protected]
+RUN MAX_JOBS=1 NINJA_FLAGS="-j1" pip3 install flash-attn==2.5.3 --no-cache-dir --no-build-isolation
+RUN MAX_JOBS=1 NINJA_FLAGS="-j1" pip3 install git+https://github.com/NVIDIA/[email protected]

docker/Dockerfile.vllm.sglang.megatron

@@ -0,0 +1,124 @@
+# Start from the NVIDIA official image (ubuntu-22.04 + cuda-12.6 + python-3.10)
+# https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes/rel-24-08.html
+FROM nvcr.io/nvidia/pytorch:24.08-py3
+
+# Define environments
+ENV MAX_JOBS=32
+ENV VLLM_WORKER_MULTIPROC_METHOD=spawn
+ENV DEBIAN_FRONTEND=noninteractive
+ENV NODE_OPTIONS=""
+ENV PIP_ROOT_USER_ACTION=ignore
+ENV HF_HUB_ENABLE_HF_TRANSFER="1"
+
+# Define installation arguments
+ARG APT_SOURCE=https://mirrors.tuna.tsinghua.edu.cn/ubuntu/
+ARG PIP_INDEX=https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple
+
+# Set apt source
+RUN cp /etc/apt/sources.list /etc/apt/sources.list.bak && \
+    { \
+    echo "deb ${APT_SOURCE} jammy main restricted universe multiverse"; \
+    echo "deb ${APT_SOURCE} jammy-updates main restricted universe multiverse"; \
+    echo "deb ${APT_SOURCE} jammy-backports main restricted universe multiverse"; \
+    echo "deb ${APT_SOURCE} jammy-security main restricted universe multiverse"; \
+    } > /etc/apt/sources.list
+
+# Install systemctl
+RUN apt-get update && \
+    apt-get install -y -o Dpkg::Options::="--force-confdef" systemd && \
+    apt-get clean
+
+# Install tini
+RUN apt-get update && \
+    apt-get install -y tini aria2 && \
+    apt-get clean
+
+# Change pip source
+RUN pip config set global.index-url "${PIP_INDEX}" && \
+    pip config set global.extra-index-url "${PIP_INDEX}" && \
+    python -m pip install --upgrade pip
+
+# Uninstall nv-pytorch fork
+RUN pip uninstall -y torch torchvision torchaudio \
+    pytorch-quantization pytorch-triton torch-tensorrt \
+    xgboost transformer_engine flash_attn apex megatron-core grpcio
+
+# Reinstall CUDA 12.4
+RUN aria2c https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-ubuntu2204.pin && \
+    mv cuda-ubuntu2204.pin /etc/apt/preferences.d/cuda-repository-pin-600
+
+RUN aria2c --always-resume=true --max-tries=99999 https://developer.download.nvidia.com/compute/cuda/12.4.1/local_installers/cuda-repo-ubuntu2204-12-4-local_12.4.1-550.54.15-1_amd64.deb && \
+    dpkg -i cuda-repo-ubuntu2204-12-4-local_12.4.1-550.54.15-1_amd64.deb && \
+    cp /var/cuda-repo-ubuntu2204-12-4-local/cuda-*-keyring.gpg /usr/share/keyrings/ && \
+    apt-get update && \
+    apt-get -y install cuda-toolkit-12-4 && \
+    rm cuda-repo-ubuntu2204-12-4-local_12.4.1-550.54.15-1_amd64.deb && \
+    update-alternatives --set cuda /usr/local/cuda-12.4 && \
+    rm -rf /usr/local/cuda-12.6
+
+# Install torch-2.6.0+cu124 + vllm-0.8.5.post1 + sglang-0.4.6.post5
+# torch-2.6.0+cu124: cxx11abi=False
+# torch-2.6.0+cu126: cxx11abi=True
+# see https://github.com/flashinfer-ai/flashinfer/issues/911
+# Install sglang-0.4.6.post1 and torch-memory-saver
+RUN pip install "sglang[all]==0.4.6.post5" --no-cache-dir --find-links https://flashinfer.ai/whl/cu124/torch2.6/flashinfer-python && pip install torch-memory-saver --no-cache-dir
+
+RUN pip install --no-cache-dir "vllm==0.8.5.post1" "torch==2.6.0" "torchvision==0.21.0" "torchaudio==2.6.0" "tensordict==0.6.2" torchdata
+
+RUN pip install --no-cache-dir "transformers[hf_xet]>=4.51.0" accelerate datasets peft hf-transfer \
+    "numpy<2.0.0" "pyarrow>=15.0.0" pandas \
+    ray[default] codetiming hydra-core pylatexenc qwen-vl-utils wandb dill pybind11 liger-kernel mathruler blobfile \
+    pytest py-spy pyext pre-commit ruff
+
+# Install flash-attn-2.7.4.post1 (cxx11abi=False)
+RUN wget -nv https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.4.post1/flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl && \
+    pip install --no-cache-dir flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
+
+# Fix packages
+RUN pip uninstall -y pynvml nvidia-ml-py && \
+    pip install --no-cache-dir --upgrade "nvidia-ml-py>=12.560.30" "fastapi[standard]>=0.115.0" "optree>=0.13.0" "pydantic>=2.9" "grpcio>=1.62.1"
+
+# Install cudnn
+RUN aria2c --max-tries=9999 https://developer.download.nvidia.com/compute/cudnn/9.8.0/local_installers/cudnn-local-repo-ubuntu2204-9.8.0_1.0-1_amd64.deb && \
+    dpkg -i cudnn-local-repo-ubuntu2204-9.8.0_1.0-1_amd64.deb && \
+    cp /var/cudnn-local-repo-ubuntu2204-9.8.0/cudnn-*-keyring.gpg /usr/share/keyrings/ && \
+    apt-get update && \
+    apt-get -y install cudnn-cuda-12 && \
+    rm cudnn-local-repo-ubuntu2204-9.8.0_1.0-1_amd64.deb
+
+RUN pip install --no-cache-dir nvidia-cudnn-cu12==9.8.0.87
+
+# Install Apex
+RUN git clone https://github.com/NVIDIA/apex.git && \
+    cd apex && \
+    pip install -v --disable-pip-version-check --no-cache-dir --no-build-isolation --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" ./
+
+# Install TransformerEngine
+RUN export NVTE_FRAMEWORK=pytorch && pip3 install --no-deps --no-cache-dir git+https://github.com/NVIDIA/[email protected]
+
+# Install Megatron-LM
+RUN pip3 install --no-deps --no-cache-dir git+https://github.com/NVIDIA/Megatron-LM.git@core_v0.12.0
+
+# Fix opencv
+RUN pip install opencv-python
+
+RUN pip install opencv-fixer && \
+    python -c "from opencv_fixer import AutoFix; AutoFix()"
+
+# Install verl
+
+# Reset pip config
+RUN pip config unset global.index-url && \
+    pip config unset global.extra-index-url
+
+    RUN apt-get update && \
+    apt-get install -y aria2 libfreeimage3 libfreeimage-dev zlib1g
+
+RUN aria2c --always-resume=true --max-tries=99999 https://developer.nvidia.com/downloads/assets/tools/secure/nsight-systems/2025_3/nsight-systems-2025.3.1_2025.3.1.90-1_amd64.deb && \
+apt-get update && apt-get install -y libxcb-cursor0 && \
+dpkg -i ./nsight-systems-2025.3.1_2025.3.1.90-1_amd64.deb && \
+rm -rf /usr/local/cuda/bin/nsys && \
+ln -s /opt/nvidia/nsight-systems/2025.3.1/target-linux-x64/nsys  /usr/local/cuda/bin/nsys && \
+rm -rf /usr/local/cuda/bin/nsys-ui && \
+ln -s /opt/nvidia/nsight-systems/2025.3.1/target-linux-x64/nsys-ui /usr/local/cuda/bin/nsys-ui && \
+rm nsight-systems-2025.3.1_2025.3.1.90-1_amd64.deb

docker/Dockerfile.vllm.sglang.megatron.deepseek

@@ -0,0 +1,115 @@
+# Start from the NVIDIA official image (ubuntu-22.04 + cuda-12.6 + python-3.10)
+# https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes/rel-24-08.html
+FROM nvcr.io/nvidia/pytorch:24.08-py3
+
+# Define environments
+ENV MAX_JOBS=32
+ENV VLLM_WORKER_MULTIPROC_METHOD=spawn
+ENV DEBIAN_FRONTEND=noninteractive
+ENV NODE_OPTIONS=""
+ENV PIP_ROOT_USER_ACTION=ignore
+ENV HF_HUB_ENABLE_HF_TRANSFER="1"
+
+# Define installation arguments
+ARG APT_SOURCE=https://mirrors.tuna.tsinghua.edu.cn/ubuntu/
+ARG PIP_INDEX=https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple
+
+# Set apt source
+RUN cp /etc/apt/sources.list /etc/apt/sources.list.bak && \
+    { \
+    echo "deb ${APT_SOURCE} jammy main restricted universe multiverse"; \
+    echo "deb ${APT_SOURCE} jammy-updates main restricted universe multiverse"; \
+    echo "deb ${APT_SOURCE} jammy-backports main restricted universe multiverse"; \
+    echo "deb ${APT_SOURCE} jammy-security main restricted universe multiverse"; \
+    } > /etc/apt/sources.list
+
+# Install systemctl
+RUN apt-get update && \
+    apt-get install -y -o Dpkg::Options::="--force-confdef" systemd && \
+    apt-get clean
+
+# Install tini
+RUN apt-get update && \
+    apt-get install -y tini aria2 && \
+    apt-get clean
+
+# Change pip source
+RUN pip config set global.index-url "${PIP_INDEX}" && \
+    pip config set global.extra-index-url "${PIP_INDEX}" && \
+    python -m pip install --upgrade pip
+
+# Uninstall nv-pytorch fork
+RUN pip uninstall -y torch torchvision torchaudio \
+    pytorch-quantization pytorch-triton torch-tensorrt \
+    xgboost transformer_engine flash_attn apex megatron-core grpcio
+
+# Reinstall CUDA 12.4
+RUN aria2c https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-ubuntu2204.pin && \
+    mv cuda-ubuntu2204.pin /etc/apt/preferences.d/cuda-repository-pin-600
+
+RUN aria2c --always-resume=true --max-tries=99999 https://developer.download.nvidia.com/compute/cuda/12.4.1/local_installers/cuda-repo-ubuntu2204-12-4-local_12.4.1-550.54.15-1_amd64.deb && \
+    dpkg -i cuda-repo-ubuntu2204-12-4-local_12.4.1-550.54.15-1_amd64.deb && \
+    cp /var/cuda-repo-ubuntu2204-12-4-local/cuda-*-keyring.gpg /usr/share/keyrings/ && \
+    apt-get update && \
+    apt-get -y install cuda-toolkit-12-4 && \
+    rm cuda-repo-ubuntu2204-12-4-local_12.4.1-550.54.15-1_amd64.deb && \
+    update-alternatives --set cuda /usr/local/cuda-12.4 && \
+    rm -rf /usr/local/cuda-12.6
+
+# Install torch-2.6.0+cu124 + vllm-0.8.5.post1 + sglang-0.4.6.post5
+# torch-2.6.0+cu124: cxx11abi=False
+# torch-2.6.0+cu126: cxx11abi=True
+# see https://github.com/flashinfer-ai/flashinfer/issues/911
+# Install sglang-0.4.6.post1 and torch-memory-saver
+RUN pip install --resume-retries 999 "sglang[all]==0.4.6.post5" --no-cache-dir --find-links https://flashinfer.ai/whl/cu124/torch2.6/flashinfer-python && pip install --resume-retries 999 torch-memory-saver --no-cache-dir
+
+RUN pip install --resume-retries 999 --no-cache-dir "vllm==0.8.5.post1" "torch==2.6.0" "torchvision==0.21.0" "torchaudio==2.6.0" "tensordict==0.6.2" torchdata
+
+RUN pip install --resume-retries 999 --no-cache-dir "transformers[hf_xet]>=4.51.0" accelerate datasets peft hf-transfer \
+    "numpy<2.0.0" "pyarrow>=15.0.0" pandas \
+    ray[default] codetiming hydra-core pylatexenc qwen-vl-utils wandb dill pybind11 liger-kernel mathruler blobfile \
+    pytest py-spy pyext pre-commit ruff
+
+# Install flash-attn-2.7.4.post1 (cxx11abi=False)
+RUN wget -nv https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.4.post1/flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl && \
+    pip install --no-cache-dir flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
+
+# Fix packages
+RUN pip uninstall -y pynvml nvidia-ml-py && \
+    pip install --resume-retries 999 --no-cache-dir --upgrade "nvidia-ml-py>=12.560.30" "fastapi[standard]>=0.115.0" "optree>=0.13.0" "pydantic>=2.9" "grpcio>=1.62.1"
+
+# Install cudnn
+RUN aria2c --max-tries=9999 https://developer.download.nvidia.com/compute/cudnn/9.8.0/local_installers/cudnn-local-repo-ubuntu2204-9.8.0_1.0-1_amd64.deb && \
+    dpkg -i cudnn-local-repo-ubuntu2204-9.8.0_1.0-1_amd64.deb && \
+    cp /var/cudnn-local-repo-ubuntu2204-9.8.0/cudnn-*-keyring.gpg /usr/share/keyrings/ && \
+    apt-get update && \
+    apt-get -y install cudnn-cuda-12 && \
+    rm cudnn-local-repo-ubuntu2204-9.8.0_1.0-1_amd64.deb
+
+RUN pip install --resume-retries 999 --no-cache-dir nvidia-cudnn-cu12==9.8.0.87
+
+# Install Apex
+RUN git clone https://github.com/NVIDIA/apex.git && \
+    cd apex && \
+    pip install -v --disable-pip-version-check --no-cache-dir --no-build-isolation --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" ./
+
+# Install TransformerEngine
+RUN export NVTE_FRAMEWORK=pytorch && pip3 install --no-deps --no-cache-dir git+https://github.com/NVIDIA/[email protected]
+
+# Install Megatron-LM
+RUN pip3 install --no-deps --no-cache-dir git+https://github.com/NVIDIA/Megatron-LM.git@core_v0.12.1
+
+# Fix opencv
+RUN pip install opencv-python
+
+RUN pip install opencv-fixer && \
+    python -c "from opencv_fixer import AutoFix; AutoFix()"
+
+# Install verl
+
+# Reset pip config
+RUN pip config unset global.index-url && \
+    pip config unset global.extra-index-url
+
+    RUN apt-get update && \
+    apt-get install -y aria2 libfreeimage3 libfreeimage-dev zlib1g

docs/Makefile

@@ -0,0 +1,20 @@
+# Minimal makefile for Sphinx documentation
+#
+
+# You can set these variables from the command line.
+SPHINXOPTS    =
+SPHINXBUILD   = sphinx-build
+SPHINXPROJ    = verl
+SOURCEDIR     = .
+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)

docs/README.md

@@ -0,0 +1,19 @@
+# verl documentations
+
+## Build the docs
+
+```bash
+# Install dependencies.
+pip install -r requirements-docs.txt
+
+# Build the docs.
+make clean
+make html
+```
+
+## Open the docs with your browser
+
+```bash
+python -m http.server -d _build/html/
+```
+Launch your browser and navigate to http://localhost:8000 to view the documentation.

docs/README_vllm0.7.md

@@ -0,0 +1,73 @@
+# Upgrading to vllm >= 0.7
+
+Note: verl+vllm 0.8.3 is now stable. Please see ``docs/README_vllm0.8.md`` for upgrade guide.
+
+## Installation
+
+Note: At time of writing, verl+vllm 0.7.x supports **FSDP** for training and **vLLM** for rollout.
+
+```
+# Create the conda environment
+conda create -n verl python==3.10
+conda activate verl
+
+# Install verl
+git clone https://github.com/volcengine/verl.git
+cd verl
+pip3 install -e .
+
+# Install the latest stable version of vLLM
+pip3 install vllm==0.7.3 
+
+# Install flash-attn
+pip3 install flash-attn --no-build-isolation
+
+```
+
+Note that if you are installing lower versions of vLLM (0.7.0, 0.7.1, 0.7.2), you need to make some tiny patches manually on vllm (/path/to/site-packages/vllm after installation) after the above steps:
+
+- vllm/distributed/parallel_state.py: Remove the assertion below:
+
+```
+if (world_size
+        != tensor_model_parallel_size * pipeline_model_parallel_size):
+    raise RuntimeError(
+        f"world_size ({world_size}) is not equal to "
+        f"tensor_model_parallel_size ({tensor_model_parallel_size}) x "
+        f"pipeline_model_parallel_size ({pipeline_model_parallel_size})")
+
+```
+
+- vllm/executor/uniproc_executor.py: change `local_rank = rank` to `local_rank = int(os.environ["LOCAL_RANK"])`
+- vllm/model_executor/model_loader/weight_utils.py: remove the `torch.cuda.empty_cache()` in `pt_weights_iterator`
+
+## Features
+
+### Use cuda graph
+
+After installation, examples using FSDP as training backends can be used. By default, the `enforce_eager` is set to True, which disables the cuda graph. To enjoy cuda graphs and the sleep mode of vLLM>=0.7, add the following lines to the bash script:
+
+```
+actor_rollout_ref.rollout.enforce_eager=False \
+actor_rollout_ref.rollout.free_cache_engine=False \
+
+```
+
+For a typical job like examples/ppo_trainer/run_qwen2-7b_seq_balance.sh, the rollout generation time is 115 seconds with vLLM0.6.3, while it is 85 seconds with vLLM0.7.0. By enabling the cudagraph, the generation duration is further reduced to 62 seconds.
+
+**Note:** Currently, if the `n` is greater than 1 in `SamplingParams` in vLLM>=0.7, there is a potential performance issue on the stability of rollout generation time (Some iterations would see generation time bursts) using vLLM's V0 Engine.
+
+### Use vLLM V1 Engine
+
+Using the vLLM V1 engine can avoid instability issues and achieve additional performance improvements. To use the V1 engine, you can first uninstall the previously installed vLLM and then follow the steps below to install the newer version.
+
+```
+git clone https://github.com/vllm-project/vllm.git
+cd vllm
+git checkout 2275784
+sed -i "903a\    data_parallel_size = world_size // pipeline_model_parallel_size // tensor_model_parallel_size" ./vllm/distributed/parallel_state.py
+VLLM_USE_PRECOMPILED=1 pip install --editable .
+```
+
+Then you can enable the V1 engine by setting `export VLLM_USE_V1=1`. In some benchmark tests, the V1 engine demonstrates a 1.5x speed improvement over the vLLM V0 engine.
+The stable support of the vLLM V1 engine is available on verl main.

docs/README_vllm0.8.md

@@ -0,0 +1,55 @@
+# Upgrading to vLLM >= 0.8
+
+## Installation
+
+Note: This version of verl+vLLM 0.8+ supports **FSDP** for training and **vLLM** for rollout.
+
+```bash
+# Create the conda environment
+conda create -n verl python==3.10
+conda activate verl
+
+# Install verl
+git clone https://github.com/volcengine/verl.git
+cd verl
+pip3 install -e .
+
+# Install the latest stable version of vLLM
+pip3 install vllm==0.8.3
+
+# Install flash-attn
+pip3 install flash-attn --no-build-isolation
+
+```
+
+We have a pre-built docker image for verl+vLLM 0.8.3. You can direct import it with the following command:
+
+```bash
+docker pull hiyouga/verl:ngc-th2.6.0-cu126-vllm0.8.3-flashinfer0.2.2-cxx11abi0
+```
+
+## Features
+
+vLLM 0.8+ supports cuda graph and V1 engine by default in verl. To enable these features, remember to add the following lines to the bash script:
+
+```bash
+actor_rollout_ref.rollout.enforce_eager=False \
+actor_rollout_ref.rollout.free_cache_engine=False \
+```
+
+and also **remove** the environment variable if it exists:
+
+```bash
+# If you are using vllm<=0.6.3, you might need to set the following environment variable to avoid bugs:
+# export VLLM_ATTENTION_BACKEND=XFORMERS
+```
+
+## Notes
+
+When you just directly upgrade vllm>=0.8, some dependency packages may undergo version changes. If you encounter the following problems:
+
+```bash
+in <module> from torch.multiprocessing.reductions import ForkingPickler ImportError: cannot import name 'ForkingPickler' from 'torch.multiprocessing.reductions' (/opt/conda/lib/python3.11/site-packages/torch/multiprocessing/reductions.py)
+```
+
+You need to upgrade `tensordict` to version 0.6.2 using the command `pip install tensordict==0.6.2`.

docs/_static/js/runllm-widget.js

@@ -0,0 +1,14 @@
+document.addEventListener("DOMContentLoaded", function () {
+    var script = document.createElement("script");
+    script.type = "module";
+    script.id = "runllm-widget-script";
+    script.src = "https://widget.runllm.com";
+    script.setAttribute("version", "stable");
+    script.setAttribute("crossorigin", "true");
+    script.setAttribute("runllm-keyboard-shortcut", "Mod+j");
+    script.setAttribute("runllm-name", "verl Chatbot");
+    script.setAttribute("runllm-position", "TOP_RIGHT");
+    script.setAttribute("runllm-assistant-id", "679");
+    script.async = true;
+    document.head.appendChild(script);
+  });

docs/advance/checkpoint.rst

@@ -0,0 +1,159 @@
+Using Checkpoints to Support Fault Tolerance Training
+=====================================================
+
+There could be training errors or machine failure during the whole RLHF training process, 
+so it is recommended to enable checkpoints to minimize your loss.
+
+The API Interface has already been listed in :ref:`config-explain-page`,
+and we will not repeat them. But there are still some technique details
+we hope to clarify.
+
+.. note:: 
+
+    Notice that the ``checkpoint.contents`` field has no effect to FSDP checkpoint except ``hf_model``, 
+    the other 3 fields are binded together to save and load. We recommend to include ``model``, ``optimizer`` and ``extra`` all.
+
+Checkpoint Saving Directory Structure
+-------------------------------------
+
+Commonly, we use the ``default_local_dir`` declared in ``ppo_trainer.yaml`` or ``ppo_megatron_trainer.yml``
+to work as preffix when saving checkpoints, which is ``checkpoints/${trainer.project_name}/${trainer.experiment_name}``.
+
+So the inner checkpoint structure of **FSDP** is like:
+
+.. code::
+
+    checkpoints/${trainer.project_name}/${trainer.experiment_name}
+    ├── global_steps_${i}
+    │   ├── actor
+    │   │   ├── huggingface     # default save config and tokenizer, save huggingface model if include ``hf_model`` in checkpoint.contents
+    │   │   ├── model_world_size_{self.world_size}_rank_{self.rank}.pt
+    │   │   ├── optim_world_size_{self.world_size}_rank_{self.rank}.pt
+    │   │   └── extra_state_world_size_{self.world_size}_rank_{self.rank}.pt
+    │   ├── critic
+    │   │   ├── huggingface
+    │   │   ├── model_world_size_{self.world_size}_rank_{self.rank}.pt
+    │   │   ├── optim_world_size_{self.world_size}_rank_{self.rank}.pt
+    │   │   └── extra_state_world_size_{self.world_size}_rank_{self.rank}.pt
+    └── latest_checkpointed_iteration.txt
+
+All model shards, optimizers and extra states are stored together, in a sharded and distributed way.
+
+While **Megatron** current checkpoint structure is:
+
+.. code::
+
+    checkpoints/${trainer.project_name}/${trainer.experiment_name}
+    ├── global_steps_${i}
+    │   ├── actor
+    │   │   ├── huggingface     # default save config and tokenizer, save huggingface model if include ``hf_mode`` in checkpoint.contents
+    │   │   └── dist_ckpt       # save sharded model/optimizer/rng_states, naming the same as Megatron
+    │   └── critic
+    │   │   ├── huggingface
+    │   │   └── dist_ckpt
+    └── latest_checkpointed_iteration.txt
+
+Convert FSDP and Megatron Checkpoints to HuggingFace Format Model
+-----------------------------------------------------------------
+
+We provide a tool to convert the FSDP and Megatron checkpoints to HuggingFace format model.
+The tool is located in ``verl/model_merger``.
+
+The script supports two main sub-commands: `merge` (to convert and save checkpoints) and `test` (to validate merged checkpoints against a reference model).
+The arguments for the `merge` sub-command are as follows:
+
+.. code:: bash
+
+    usage: python -m verl.model_merger merge [-h] --backend {fsdp,megatron} --local_dir LOCAL_DIR [--hf_model_path HF_MODEL_PATH]
+                                [--tie-word-embedding] [--is-value-model] [--target_dir TARGET_DIR]
+                                [--hf_upload_path HF_UPLOAD_PATH] [--private]
+
+    options:
+    -h, --help            show this help message and exit
+    --backend {fsdp,megatron}
+                            The backend of the model
+    --local_dir LOCAL_DIR
+                            Path to the saved model checkpoints
+    --hf_model_path HF_MODEL_PATH
+                            (Deprecated) Path to the original Hugging Face model for config.
+    --tie-word-embedding  Whether to tie word embedding weights (currently only Megatron supported)
+    --is-value-model      Whether the model is a value model (currently only Megatron supported)
+    --target_dir TARGET_DIR
+                            Directory to save the merged huggingface model
+    --hf_upload_path HF_UPLOAD_PATH
+                            Hugging Face repository ID to upload the model
+    --private             Whether to upload the model to a private Hugging Face repository
+
+Example usage for merging Megatron checkpoints:
+
+.. code:: bash
+
+    python -m verl.model_merger merge \
+        --backend megatron \
+        --tie-word-embedding \
+        --local_dir checkpoints/verl_megatron_gsm8k_examples/qwen2_5_0b5_megatron_saveload/global_step_1/actor \
+        --target_dir /path/to/merged_hf_model
+
+Example usage for merging FSDP checkpoints:
+
+.. code:: bash
+
+    python -m verl.model_merger merge \
+        --backend fsdp \
+        --local_dir checkpoints/verl_fsdp_gsm8k_examples/qwen2_5_0b5_fsdp_saveload/global_step_1/actor \
+        --target_dir /path/to/merged_hf_model
+
+
+Megatron Merger details
+-----------------------
+
+Current implement of decoder layers uses ``nn.ModuleList`` to store the layers, 
+and thus the model layers on every PP rank and VPP rank starts their index from 0.
+
+There are 3 ways to correct this behavior:
+
+1. Modify the decoder layer's state_dict, add ``offset`` to each layer's index, thus rewrite ``nn.ModuleList`` implementation.
+2. Modify the layer index when saving checkpoint and recover them when loading checkpoint.
+3. The Checkpoint merger do this work, calculate the actual ``offset`` from ``state_dict`` only, a little complex.
+
+Current implementation use solution 2.
+
+
+HuggingFace to Megatron DistCheckpoint details
+----------------------------------------------
+
+If your model is quite huge, we recommend you to use Megatron dist-checkpoint to load the model.
+Megatron dist-checkpoint supports loading with different kinds of model parallelism,
+and it is much faster than the original checkpoint loading.
+
+To convert original HuggingFace model to Megatron dist-checkpoint,
+you can use the ``scripts/converter_hf_to_mcore.py`` script. Large MoE models are temporarily supported with CPU initialization,
+which is a little slower. While we are working on a better solution to support large models.
+
+Example command to convert the model is as follows:
+
+.. code:: bash
+
+    python scripts/converter_hf_to_mcore.py \
+        --hf_model_path Qwen/Qwen1.5-MoE-A2.7B-Chat \
+        --output_path /mnt/disk/Qwen/Qwen1.5-MoE-A2.7B-Chat \
+        --use_cpu_initialization    # Only work for MoE models
+
+
+Original Checkpoint Utils
+-------------------------
+
+Original Checkpoint Utils refer to original checkpoint implementation in ``verl/models/[model]/megatron/checkpoint_utils``.
+
+We only need ``[model]_loader.py`` in original checkpoint utils now, since we get rid of storing ``hf_model`` every time (which is not recommended for large model training, try only saving sharded models if you can).
+
+.. note:: 
+
+    Note that ``[model]_loader`` only support environments where **storage clusters are able to connect with every calculation nodes**. 
+    Because it utilizes **sharded load way to minimize the loading checkpoint overhead**. 
+    Every rank loads its own data from ``state_dict`` which can be accessed by all of them.
+    While there is also no need to broadcast among DP ranks, since the saved state_dict is only produced by DP rank 0.
+
+    For users who can **only place the huggingface model on one device**, we keep the original costly implementation in ``[model]_loader_deprecated``. In this implementation, rank 0 broadcast all weights to each tp and pp rank, and then dp rank 0 broadcast to all dp ranks. There may be at risks of OOM.
+
+    To use deprecated loader, change the import package of ``load_state_dict_to_megatron_llama``.

docs/advance/dpo_extension.rst

@@ -0,0 +1,271 @@
+Extend to other RL(HF) algorithms
+=================================
+
+We already implemented the complete training pipeline of the PPO
+algorithms. To extend to other algorithms, we analyze the high-level
+principle to use verl and provide a tutorial to implement the DPO
+algorithm. Users can follow the similar paradigm to extend to other RL algorithms.
+
+.. note:: **Key ideas**: Single process drives multi-process computation and data communication.
+
+Overall Approach
+----------------
+
+Step 1: Consider what multi-machine multi-GPU computations are needed
+for each model, such as ``generate_sequence`` , ``compute_log_prob`` and
+``update_policy`` in the actor_rollout model. Implement distributed
+single-process-multiple-data (SPMD) computation and encapsulate them
+into APIs
+
+Step 2: Based on different distributed scenarios, including FSDP and 3D
+parallelism in Megatron-LM, implement single-process control of data
+interaction among multi-process computations.
+
+Step 3: Utilize the encapsulated APIs to implement the control flow
+
+Example: Online DPO
+-------------------
+
+We use verl to implement a simple online DPO algorithm. The algorithm
+flow of Online DPO is as follows:
+
+1. There is a prompt (rollout) generator which has the same weight as
+   the actor model. After a batch of prompts are fed into the generator,
+   it generates N responses for each prompt.
+2. Send all the prompts + responses to a verifier for scoring, which can
+   be reward model or a rule-based function. Then sort them in pairs to
+   form a training batch.
+3. Use this training batch to train the actor model using DPO. During
+   the process, a reference policy is needed.
+
+Step 1: What are the multi-machine multi-GPU computations
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+**Sample Generator**
+
+Implementation details:
+
+.. code:: python
+
+   from verl.single_controller.base import Worker
+   from verl.single_controller.ray import RayWorkerGroup, RayClassWithInitArgs, RayResourcePool
+   import ray
+
+   @ray.remote
+   class SampleGenerator(Worker):
+       def __init__(self, config):
+           super().__init__()
+           self.config = config
+           
+       def generate_sequences(self, data):
+           pass
+
+Here, ``SampleGenerator`` can be viewed as a multi-process pulled up by
+``torchrun``, with each process running the same code (SPMD).
+``SampleGenerator`` needs to implement a ``generate_sequences`` API for
+the control flow to call. The implementation details inside can use any
+inference engine including vllm, sglang and huggingface. Users can
+largely reuse the code in
+verl/verl/workers/rollout/vllm_rollout/vllm_rollout.py and we won't
+go into details here.
+
+**ReferencePolicy inference**
+
+API: compute reference log probability
+
+.. code:: python
+
+   from verl.single_controller.base import Worker
+   import ray
+
+   @ray.remote
+   class ReferencePolicy(Worker):
+       def __init__(self):
+           super().__init__()
+           self.model = Model()
+           
+       def infer(self, data):
+           return self.model(data)
+
+**Actor update**
+
+API: Update actor model parameters
+
+.. code:: python
+
+   from verl.single_controller.base import Worker
+   import ray
+
+   @ray.remote
+   class DPOActor(Worker):
+       def __init__(self):
+           super().__init__()
+           self.model = Model()
+           self.model = FSDP(self.model)  # or other distributed strategy
+           self.optimizer = optim.Adam(self.model.parameters(), lr=1e-3)
+           self.loss_fn = xxx
+           
+       def update(self, data):
+           self.optimizer.zero_grad()
+           logits = self.model(data)
+           loss = self.loss_fn(logits)
+           loss.backward()
+           self.optimizer.step()
+
+**Notes: How to distinguish between control processes and distributed computation processes**
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+- Control processes are generally functions directly decorated with
+  ``@ray.remote``
+- Computation processes are all wrapped into a ``RayWorkerGroup``.
+
+Users can reuse most of the distribtued computation logics implemented
+in PPO algorithm, including FSDP and Megatron-LM backend in
+verl/verl/trainer/ppo.
+
+Step 2: Based on different distributed scenarios, implement single-process control of multi-process data interaction
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+**The core problem to solve here is how a single process sends data to
+multiple processes, drives multi-process computation, and how the
+control process obtains the results of multi-process computation.**
+First, we initialize the multi-process ``WorkerGroup`` in the control
+process.
+
+.. code:: python
+
+   @ray.remote(num_cpus=1)
+   def main_task(config):
+       # construct SampleGenerator
+       resource_pool = RayResourcePool(process_on_nodes=[8] * 2)  # 16 GPUs
+       ray_cls = RayClassWithInitArgs(SampleGenerator, config=config)
+       # put SampleGenerator onto resource pool
+       worker_group = RayWorkerGroup(resource_pool, ray_cls)
+       
+       # construct reference policy
+
+As we can see, in the control process, multiple processes are wrapped
+into a ``RayWorkerGroup``. Inside this ``WorkerGroup``, there is a
+``self._workers`` member, where each worker is a RayActor
+(https://docs.ray.io/en/latest/ray-core/actors.html) of SampleGenerator.
+ray_trainer.md also provide an implementation of
+``MegatronRayWorkerGroup``.
+
+Assuming the model is distributed using FSDP, and there is a batch of
+data on the control process, for data parallelism, the underlying
+calling process is:
+
+.. code:: python
+
+   data = xxx
+   data_list = data.chunk(dp_size)
+
+   output = []
+   for d in data_list:
+       # worker_group._workers[i] is a SampleGenerator
+       output.append(worker_group._workers[i].generate_sequences.remote(d))
+
+   output = ray.get(output)
+   output = torch.cat(output)
+
+Single process calling multiple processes involves the following 3
+steps:
+
+1. Split the data into DP parts on the control process.
+2. Send the data to remote, call the remote computation through RPC, and
+   utilize multi-process computation.
+3. Obtain the computation results of each worker on the control process
+   and merge them.
+
+Frequently calling these 3 steps on the controller process greatly hurts
+code readability. **In verl, we have abstracted and encapsulated these 3
+steps, so that the worker's method + dispatch + collect can be
+registered into the worker_group**
+
+.. code:: python
+
+   from verl.single_controller.base.decorator import register
+
+   def dispatch_data(worker_group, data):
+       return data.chunk(worker_group.world_size)
+       
+   def collect_data(worker_group, data):
+       return torch.cat(data)
+
+   dispatch_mode = {
+       'dispatch_fn': dispatch_data,
+       'collect_fn': collect_data
+   }
+
+   @register(dispatch_mode=dispatch_mode)
+   def generate_sequences(self, data):
+       pass
+
+In this way, we can directly call the method inside the worker through
+the ``worker_group`` on the control (driver) process (which is a single
+process):
+
+.. code:: python
+
+   output = worker_group.generate_sequences(data)
+
+This single line includes data splitting, data distribution and
+computation, and data collection.
+
+Furthermore, the model parallelism size of each model is usually fixed,
+including dp, tp, pp. So for these common distributed scenarios, we have
+pre-implemented specific dispatch and collect methods,in `decorator.py <https://github.com/volcengine/verl/blob/main/verl/single_controller/base/decorator.py>`_, which can be directly used to wrap the computations.
+
+.. code:: python
+
+   from verl.single_controller.base.decorator import register, Dispatch
+
+   @register(dispatch_mode=Dispatch.DP_COMPUTE_PROTO)
+   def generate_sequences(self, data: DataProto) -> DataProto:
+       pass
+
+Here it requires the data interface to be ``DataProto``. Definition of
+``DataProto`` is in `protocol.py <https://github.com/volcengine/verl/blob/main/verl/protocol.py>`_.
+
+Step 3: Main training loop
+~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+With the above training flows, we can implement the algorithm's control
+flow. It is recommended that ``main_task`` is also a ray remote process.
+
+.. code:: python
+
+   @ray.remote(num_cpus=1)
+   def main_task(config):
+       # construct SampleGenerator
+       resource_pool = RayResourcePool(process_on_nodes=[8] * 2)  # 16 GPUs
+       ray_cls = RayClassWithInitArgs(SampleGenerator, config=config) 
+       # put SampleGenerator onto resource pool
+       sample_gen = RayWorkerGroup(resource_pool, ray_cls)
+       
+       # construct reference policy
+       ray_cls = RayClassWithInitArgs(ReferencePolicy)
+       ref_policy = RayWorkerGroup(resource_pool, ray_cls)
+       
+       # construct actor
+       ray_cls = RayClassWithInitArgs(DPOActor)  
+       dpo_policy = RayWorkerGroup(resource_pool, ray_cls)
+       
+       dataloader = DataLoader()
+       
+       for data in dataloader:
+           # generate data
+           data = sample_gen.generate_sequences(data)
+           # generate scores for each data 
+           data = generate_scores(data)
+           # generate pairwise data using scores
+           data = generate_pairwise_data(data)
+           # generate ref_log_prob
+           data.batch['ref_log_prob'] = ref_policy.infer(data)
+           # update using dpo
+           dpo_policy.update(data)
+           # logging
+
+Here, different ``WorkerGroups`` can be placed in the same resource pool or
+in different resource pools using ``create_colocated_worker_cls``
+similar as in `ray_trainer.py <https://github.com/volcengine/verl/blob/main/verl/trainer/ppo/ray_trainer.py>`_.