e2e_ppo_grpo_trainer_trtllm.yml

# # 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_trtllm

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
      - "verl/workers/rollout/trtllm_rollout/**"
      - "tests/workers/rollout/rollout_trtllm/**"
      - ".github/workflows/e2e_ppo_grpo_trainer_trtllm.yml"
      - "examples/data_preprocess/gsm8k.py"
      - "examples/data_preprocess/geo3k.py"
      - "examples/data_preprocess/dapo_multiturn_w_tool.py"
      - "examples/data_preprocess/aime2024_multiturn_w_tool.py"
      - "examples/grpo_trainer/run_qwen2-7b_math_trtllm.sh"
      - "examples/grpo_trainer/run_qwen2-7b_math_megatron_trtllm.sh"
      - "examples/grpo_trainer/run_qwen3-30b_dapo_megatron_fp8_trtllm.sh"
      # add back when ppo flow is ready
      # - "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/verlai/verl:trtllm1.3.0rc4"
  DYNAMIC_RUNNER_ENDPOINT: "https://sd10g3clalm04ug7alq90.apigateway-cn-beijing.volceapi.com/runner"

jobs:
  setup:
    if: github.repository_owner == 'verl-project'
    runs-on: ubuntu-latest
    outputs:
      runner-label: ${{ steps.create-runner.outputs.runner-label }}
      mlp-task-id: ${{ steps.create-runner.outputs.mlp-task-id }}
    steps:
      - uses: actions/checkout@v4
      - id: create-runner
        uses: volcengine/vemlp-github-runner@v1
        with:
          mode: "create"
          faas-url: "${{ env.DYNAMIC_RUNNER_ENDPOINT }}"
          mlp-image: "${{ env.IMAGE }}"

  trtllm_unit_tests:
    needs: setup
    runs-on: ["${{ needs.setup.outputs.runner-label || 'L20x8' }}"]
    timeout-minutes: 30 # Increase this timeout value as needed
    env:
      HTTP_PROXY: ${{ secrets.PROXY_HTTP }}
      HTTPS_PROXY: ${{ secrets.PROXY_HTTPS }}
      NO_PROXY: "localhost,127.0.0.1,hf-mirror.com"
      HF_ENDPOINT: "https://hf-mirror.com"
      HF_HUB_ENABLE_HF_TRANSFER: "0" # This is more stable
    steps:
      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
        with:
          fetch-depth: 0
      - name: Install the current repository
        run: |
          pip3 install pytest-asyncio
          pip3 install -r requirements-test.txt
          pip3 install --no-deps -e .
      - name: Run TRTLLM unit tests
        run: |
          export TRTLLM_TEST_MODEL_PATH_ROOT="${HOME}/models"
          ray stop --force
          pytest -v -s \
            tests/workers/rollout/rollout_trtllm/test_adapter.py \
            tests/workers/rollout/rollout_trtllm/test_async_server.py \
            tests/workers/rollout/rollout_trtllm/test_trtllm_rollout_utils.py

  e2e_grpo_trainer_fsdp-qwen2:
    needs: setup
    runs-on: ["${{ needs.setup.outputs.runner-label || '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-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 -r requirements-test.txt
          pip3 install --no-deps -e .
      - name: Prepare GSM8K dataset
        run: |
          python3 examples/data_preprocess/gsm8k.py --local_dataset_path ${HOME}/models/hf_data/gsm8k --local_save_dir ${PWD}/data/gsm8k
      - name: Running GSM8K E2E training tests with FSDP on 8 L20 GPUs (Qwen)
        run: |
          ray stop --force
          DATADIR=${HOME}/data \
            bash examples/grpo_trainer/run_qwen2-7b_math_trtllm.sh 2 \
            trainer.total_training_steps=1 \
            data.train_files="['${PWD}/data/gsm8k/train.parquet']" \
            data.val_files="['${PWD}/data/gsm8k/test.parquet']" \
            trainer.logger='["console"]' \
            actor_rollout_ref.model.path="${HOME}/models/Qwen/Qwen2.5-0.5B-Instruct"
      - name: clean up
        run: |
          rm -rf checkpoints

  e2e_grpo_trainer_megatron-qwen2:
    needs: setup
    runs-on: ["${{ needs.setup.outputs.runner-label || '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-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 -r requirements-test.txt
          pip3 install --no-deps -e .
      - name: Prepare GSM8K dataset
        run: |
          python3 examples/data_preprocess/gsm8k.py --local_dataset_path ${HOME}/models/hf_data/gsm8k --local_save_dir ${PWD}/data/gsm8k
      - name: Running GSM8K E2E training tests with 3D parallelism on 8 L20 GPUs with Megatron (Qwen)
        run: |
          ray stop --force
          DATADIR=${HOME}/data \
          ACTOR_TP=2 \
            bash examples/grpo_trainer/run_qwen2-7b_math_megatron_trtllm.sh 2 \
            trainer.total_training_steps=1 \
            data.train_files="['${PWD}/data/gsm8k/train.parquet']" \
            data.val_files="['${PWD}/data/gsm8k/test.parquet']" \
            trainer.logger='["console"]' \
            actor_rollout_ref.model.path="${HOME}/models/Qwen/Qwen2.5-0.5B-Instruct"
      - name: clean up
        run: |
          rm -rf checkpoints
  e2e_grpo_trainer_fsdp-vlm:
    needs: setup
    runs-on: ["${{ needs.setup.outputs.runner-label || '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-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 -r requirements-test.txt
          pip3 install --no-deps -e .
      - name: Prepare GEO3K dataset
        run: |
          python3 examples/data_preprocess/geo3k.py --local_dataset_path ${HOME}/models/hf_data/hiyouga/geometry3k --local_save_dir ${PWD}/data/geo3k
      - name: Running GEO3K E2E training tests with FSDP on 8 L20 GPUs (VLM)
        run: |
          ray stop --force
          DATADIR=${HOME}/data \
            bash examples/grpo_trainer/run_qwen2_5_vl_3b_trtllm.sh 2 \
            trainer.total_training_steps=1 \
            data.train_files="['${PWD}/data/geo3k/train.parquet']" \
            data.val_files="['${PWD}/data/geo3k/test.parquet']" \
            trainer.logger='["console"]' \
            actor_rollout_ref.model.path="${HOME}/models/Qwen/Qwen3-VL-2B-Instruct"
      - name: clean up
        run: |
          rm -rf checkpoints
      - name: Prepare DAPO-Math-17k and AIME-2024 datasets (data_preprocess)
        run: |
          python3 examples/data_preprocess/dapo_multiturn_w_tool.py --local_save_dir ${PWD}/data/dapo-math-17k
          python3 examples/data_preprocess/aime2024_multiturn_w_tool.py --local_save_dir ${PWD}/data/aime-2024
      - name: Running DAPO E2E with FP8 TRT-LLM rollout (Qwen3-0.6B)
        run: |
          ray stop --force
          export INFER_TP=2 ACTOR_TP=2 ACTOR_PP=2 ACTOR_VPP=2 ACTOR_EP=1 ACTOR_CP=2 REF_TP=2 REF_PP=2 REF_VPP=2 REF_EP=1 REF_CP=2 GEN_MOE_TP=null GEN_MOE_EP=null
          export NNODES=1 GPUS_PER_NODE=8 TRTLLM_MOE_BACKEND=CUTLASS
          export DATA_DIR=${PWD} DAPO_MATH_TRAIN=${PWD}/data/dapo-math-17k/train.parquet AIME_VAL=${PWD}/data/aime-2024/train.parquet MODEL_PATH=${HOME}/models/Qwen/Qwen3-0.6B
          bash examples/grpo_trainer/run_qwen3-30b_dapo_megatron_fp8_trtllm.sh \
            reward_model.reward_kwargs.overlong_buffer_cfg.len=258 \
            reward_model.reward_kwargs.max_resp_len=512 \
            data.max_prompt_length=512 \
            data.max_response_length=512 \
            data.train_batch_size=32 \
            actor_rollout_ref.rollout.n=4 \
            actor_rollout_ref.rollout.max_num_seqs=16 \
            actor_rollout_ref.rollout.max_num_batched_tokens=1024 \
            actor_rollout_ref.rollout.max_model_len=1024 \
            actor_rollout_ref.actor.megatron.override_transformer_config.moe_grouped_gemm=False \
            actor_rollout_ref.actor.megatron.override_transformer_config.moe_permute_fusion=False \
            trainer.total_training_steps=1 \
            trainer.logger='["console"]'
      - name: clean up
        run: |
          rm -rf checkpoints

  cleanup:
    runs-on: ubuntu-latest
    needs: [setup, trtllm_unit_tests, e2e_grpo_trainer_fsdp-qwen2, e2e_grpo_trainer_megatron-qwen2, e2e_grpo_trainer_fsdp-vlm]
    if: always()
    steps:
      - id: destroy-runner
        uses: volcengine/vemlp-github-runner@v1
        with:
          mode: "destroy"
          faas-url: "${{ env.DYNAMIC_RUNNER_ENDPOINT }}"
          mlp-task-id: "${{ needs.setup.outputs.mlp-task-id }}"