Compare commits
6 Commits
main
...
fix/ci-badge
| Author | SHA1 | Date | |
|---|---|---|---|
 Yuge Zhang
|
a9c05b0b9a | ||
|
Yuge Zhang
|
54902ace02 | ||
|
Yuge Zhang
|
3edbcb0d02 | ||
|
Yuge Zhang
|
abb8b8bb90 | ||
|
Yuge Zhang
|
0b276f7d53 | ||
|
Yuge Zhang
|
3d61860b53 |
@@ -0,0 +1,29 @@
|
||||
name: Badge - Compatibility
|
||||
|
||||
on:
|
||||
workflow_run:
|
||||
workflows:
|
||||
- Examples - Backward Compatibility
|
||||
types: [completed]
|
||||
|
||||
workflow_dispatch:
|
||||
|
||||
permissions:
|
||||
actions: read
|
||||
contents: read
|
||||
|
||||
jobs:
|
||||
badge:
|
||||
if: ${{ github.event_name == 'workflow_dispatch' || (github.event_name == 'workflow_run' && github.event.workflow_run.head_branch == 'main') }}
|
||||
runs-on: ubuntu-latest
|
||||
steps:
|
||||
- uses: actions/checkout@v4
|
||||
- uses: actions/github-script@v8
|
||||
with:
|
||||
github-token: ${{ secrets.GITHUB_TOKEN }}
|
||||
script: |
|
||||
const badgeAggregation = require('./scripts/badge_aggregation.js');
|
||||
const dependencies = [
|
||||
{ workflow: 'examples-compat.yml', label: 'examples-compat', variants: ['legacy', 'stable'] },
|
||||
];
|
||||
await badgeAggregation({ github, context, core, dependencies });
|
||||
@@ -0,0 +1,31 @@
|
||||
name: Badge - Unit Test
|
||||
|
||||
on:
|
||||
workflow_run:
|
||||
workflows:
|
||||
- CPU Test
|
||||
- GPU Test
|
||||
types: [completed]
|
||||
|
||||
workflow_dispatch:
|
||||
|
||||
permissions:
|
||||
actions: read
|
||||
contents: read
|
||||
|
||||
jobs:
|
||||
badge:
|
||||
if: ${{ github.event_name == 'workflow_dispatch' || (github.event_name == 'workflow_run' && github.event.workflow_run.head_branch == 'main') }}
|
||||
runs-on: ubuntu-latest
|
||||
steps:
|
||||
- uses: actions/checkout@v4
|
||||
- uses: actions/github-script@v8
|
||||
with:
|
||||
github-token: ${{ secrets.GITHUB_TOKEN }}
|
||||
script: |
|
||||
const badgeAggregation = require('./scripts/badge_aggregation.js');
|
||||
const dependencies = [
|
||||
{ workflow: 'tests-full.yml', label: 'tests-full', variants: ['legacy', 'stable'] },
|
||||
{ workflow: 'tests.yml', label: 'tests', variants: ['legacy', 'stable', 'Lint', 'documentation', 'JavaScript'] },
|
||||
];
|
||||
await badgeAggregation({ github, context, core, dependencies });
|
||||
@@ -22,12 +22,12 @@ run-name: >-
|
||||
|| format('Calc-X - {0}', github.event_name) }}
|
||||
|
||||
jobs:
|
||||
calc-x:
|
||||
calc-x-perf:
|
||||
if: >
|
||||
github.event_name != 'repository_dispatch' ||
|
||||
github.event.action == 'ci-calc-x' ||
|
||||
github.event.action == 'ci-all'
|
||||
name: Calc-X (Python ${{ matrix.python-version }}, ${{ matrix.setup-script }})
|
||||
name: Calc-X Performance (Python ${{ matrix.python-version }}, ${{ matrix.setup-script }})
|
||||
runs-on: [self-hosted, 1ES.Pool=agl-runner-gpu]
|
||||
timeout-minutes: 90
|
||||
strategy:
|
||||
@@ -74,7 +74,7 @@ jobs:
|
||||
- name: Upload dependencies artifact
|
||||
uses: actions/upload-artifact@v4
|
||||
with:
|
||||
name: dependencies-calc-x-${{ matrix.python-version }}-${{ matrix.setup-script }}
|
||||
name: dependencies-calc-x-performance-${{ matrix.python-version }}-${{ matrix.setup-script }}
|
||||
path: requirements-freeze.txt
|
||||
compression-level: 0
|
||||
|
||||
@@ -116,13 +116,11 @@ jobs:
|
||||
# Don't ask why. Don't touch this.
|
||||
- name: Calc-X training
|
||||
run: |
|
||||
set -ex
|
||||
source .venv/bin/activate
|
||||
cd examples/calc_x
|
||||
../../scripts/restart_ray.sh
|
||||
sleep 5
|
||||
PYTHONUNBUFFERED=1 python train_calc_agent.py --val-file data/test_mini.parquet --ci
|
||||
sleep 10
|
||||
python train_calc_agent.py --val-file data/test_mini.parquet --ci
|
||||
shell: bash
|
||||
env:
|
||||
WANDB_BASE_URL: ${{ secrets.MSR_WANDB_BASE_URL }}
|
||||
@@ -137,7 +135,95 @@ jobs:
|
||||
WANDB_BASE_URL: ${{ secrets.MSR_WANDB_BASE_URL }}
|
||||
WANDB_API_KEY: ${{ secrets.MSR_WANDB_API_KEY }}
|
||||
|
||||
- name: Calc-X training with local model
|
||||
calc-x-variants:
|
||||
if: >
|
||||
github.event_name != 'repository_dispatch' ||
|
||||
github.event.action == 'ci-calc-x' ||
|
||||
github.event.action == 'ci-all'
|
||||
name: Calc-X Variants (Python ${{ matrix.python-version }}, ${{ matrix.setup-script }})
|
||||
runs-on: [self-hosted, 1ES.Pool=agl-runner-gpu]
|
||||
timeout-minutes: 90
|
||||
strategy:
|
||||
matrix:
|
||||
include:
|
||||
- python-version: '3.10'
|
||||
setup-script: 'legacy'
|
||||
- python-version: '3.12'
|
||||
setup-script: 'stable'
|
||||
- python-version: '3.13'
|
||||
setup-script: 'latest'
|
||||
fail-fast: false
|
||||
steps:
|
||||
- name: Check GPU status
|
||||
run: nvidia-smi
|
||||
- name: Check disk space
|
||||
run: df -h
|
||||
- uses: actions/checkout@v4
|
||||
with:
|
||||
ref: ${{ github.event_name == 'repository_dispatch' && github.event.client_payload.pr_ref || (github.event.pull_request.number && format('refs/pull/{0}/merge', github.event.pull_request.number)) || github.ref }}
|
||||
- uses: astral-sh/setup-uv@v7
|
||||
with:
|
||||
enable-cache: true
|
||||
python-version: ${{ matrix.python-version }}
|
||||
- name: Upgrade dependencies (latest)
|
||||
run: uv lock --upgrade
|
||||
if: matrix.setup-script == 'latest'
|
||||
- name: Sync dependencies (latest)
|
||||
run: |
|
||||
uv sync --frozen --no-default-groups --extra verl \
|
||||
--group dev --group experiment --group agents --group torch-gpu-stable
|
||||
if: matrix.setup-script == 'latest'
|
||||
- name: Sync dependencies (stable & legacy)
|
||||
run: |
|
||||
uv sync --frozen --no-default-groups --extra verl \
|
||||
--group dev --group experiment --group agents --group torch-gpu-${{ matrix.setup-script }}
|
||||
if: matrix.setup-script != 'latest'
|
||||
- name: Freeze dependencies
|
||||
run: |
|
||||
set -ex
|
||||
uv pip freeze | tee requirements-freeze.txt
|
||||
echo "UV_LOCKED=1" >> $GITHUB_ENV
|
||||
echo "UV_NO_SYNC=1" >> $GITHUB_ENV
|
||||
- name: Upload dependencies artifact
|
||||
uses: actions/upload-artifact@v4
|
||||
with:
|
||||
name: dependencies-calc-x-variants-${{ matrix.python-version }}-${{ matrix.setup-script }}
|
||||
path: requirements-freeze.txt
|
||||
compression-level: 0
|
||||
|
||||
- name: Launch LiteLLM Proxy
|
||||
run: |
|
||||
./scripts/litellm_run.sh
|
||||
env:
|
||||
AZURE_API_BASE: ${{ secrets.AZURE_GROUP_SUBSCRIPTION_API_BASE }}
|
||||
AZURE_API_KEY: ${{ secrets.AZURE_GROUP_SUBSCRIPTION_API_KEY }}
|
||||
|
||||
- name: Prepare Calc-X dataset
|
||||
run: |
|
||||
set -ex
|
||||
cd examples/calc_x
|
||||
uv run gdown --fuzzy https://drive.google.com/file/d/1FQMyKLLd6hP9dw9rfZn1EZOWNvKaDsqw/view
|
||||
unzip calc-x-data.zip -d data
|
||||
rm calc-x-data.zip
|
||||
|
||||
- name: Calc-X MCP sanity check
|
||||
run: |
|
||||
set -ex
|
||||
cd examples/calc_x
|
||||
uv run tests/test_mcp_calculator.py
|
||||
env:
|
||||
OPENAI_API_BASE: http://localhost:12306/
|
||||
OPENAI_API_KEY: dummy
|
||||
- name: Calc-X sanity check
|
||||
run: |
|
||||
set -ex
|
||||
cd examples/calc_x
|
||||
uv run legacy_calc_agent_debug.py
|
||||
env:
|
||||
OPENAI_BASE_URL: http://localhost:12306/
|
||||
OPENAI_API_KEY: dummy
|
||||
|
||||
- name: Training with local model
|
||||
run: |
|
||||
set -ex
|
||||
source .venv/bin/activate
|
||||
@@ -145,7 +231,7 @@ jobs:
|
||||
../../scripts/restart_ray.sh
|
||||
sleep 5
|
||||
hf download Qwen/Qwen2.5-0.5B-Instruct --local-dir data/qwen_model
|
||||
PYTHONUNBUFFERED=1 python train_calc_agent.py --val-file data/test_mini.parquet --ci --model $(realpath data/qwen_model)
|
||||
PYTHONUNBUFFERED=1 python train_calc_agent.py --val-file data/test_mini.parquet --ci-fast --model $(realpath data/qwen_model)
|
||||
sleep 10
|
||||
shell: bash
|
||||
env:
|
||||
@@ -153,14 +239,14 @@ jobs:
|
||||
WANDB_API_KEY: ${{ secrets.MSR_WANDB_API_KEY }}
|
||||
id: calc_x_train_local_model
|
||||
|
||||
- name: Calc-X training LLM Proxy
|
||||
- name: Training with LLM Proxy
|
||||
run: |
|
||||
set -ex
|
||||
source .venv/bin/activate
|
||||
cd examples/calc_x
|
||||
../../scripts/restart_ray.sh
|
||||
sleep 5
|
||||
PYTHONUNBUFFERED=1 python train_calc_agent.py --val-file data/test_mini.parquet --ci --llm-proxy
|
||||
PYTHONUNBUFFERED=1 python train_calc_agent.py --val-file data/test_mini.parquet --ci-fast --llm-proxy
|
||||
sleep 10
|
||||
shell: bash
|
||||
env:
|
||||
@@ -168,7 +254,7 @@ jobs:
|
||||
WANDB_API_KEY: ${{ secrets.MSR_WANDB_API_KEY }}
|
||||
id: calc_x_train_llm_proxy
|
||||
|
||||
- name: Calc-X training with external store
|
||||
- name: Training with external store
|
||||
run: |
|
||||
set -euo pipefail
|
||||
source .venv/bin/activate
|
||||
@@ -198,7 +284,7 @@ jobs:
|
||||
WANDB_API_KEY: ${{ secrets.MSR_WANDB_API_KEY }}
|
||||
id: calc_x_train_external_store
|
||||
|
||||
- name: Calc-X training with role-based environment variables
|
||||
- name: Training with role-based environment variables
|
||||
run: |
|
||||
set -euo pipefail
|
||||
source .venv/bin/activate
|
||||
|
||||
@@ -4,7 +4,7 @@
|
||||
|
||||
# Agent Lightning⚡
|
||||
|
||||
[](https://github.com/microsoft/agent-lightning/actions/workflows/tests-full.yml)
|
||||
[](https://github.com/microsoft/agent-lightning/actions/workflows/badge-unit.yml)
|
||||
[](https://microsoft.github.io/agent-lightning/)
|
||||
[](https://badge.fury.io/py/agentlightning)
|
||||
[](LICENSE)
|
||||
@@ -75,10 +75,11 @@ No rewrites, no lock-in, just a clear path from first rollout to steady improvem
|
||||
| Workflow | Status |
|
||||
|----------|--------|
|
||||
| CPU Tests | [](https://github.com/microsoft/agent-lightning/actions/workflows/tests.yml) |
|
||||
| GPU Tests | [](https://github.com/microsoft/agent-lightning/actions/workflows/tests-full.yml) |
|
||||
| Full Tests | [](https://github.com/microsoft/agent-lightning/actions/workflows/badge-unit.yml) |
|
||||
| UI Tests | [](https://github.com/microsoft/agent-lightning/actions/workflows/dashboard.yml) |
|
||||
| Examples Integration | [](https://github.com/microsoft/agent-lightning/actions/workflows/badge-examples.yml) |
|
||||
| Latest Dependency Compatibility | [](https://github.com/microsoft/agent-lightning/actions/workflows/badge-latest.yml) |
|
||||
| Legacy Examples Compatibility | [](https://github.com/microsoft/agent-lightning/actions/workflows/examples-compat.yml) |
|
||||
| Legacy Examples Compatibility | [](https://github.com/microsoft/agent-lightning/actions/workflows/badge-compat.yml) |
|
||||
|
||||
## ⚡ Citation
|
||||
|
||||
|
||||
@@ -162,16 +162,17 @@ def train(
|
||||
print(f"EXPERIMENT_NAME={EXPERIMENT_NAME}")
|
||||
|
||||
# Keep it tiny/light without adding new knobs
|
||||
config["actor_rollout_ref"]["rollout"]["gpu_memory_utilization"] = 0.6
|
||||
config["actor_rollout_ref"]["rollout"]["gpu_memory_utilization"] = 0.8
|
||||
config["trainer"]["total_epochs"] = 1
|
||||
config["trainer"]["total_training_steps"] = 6
|
||||
config["trainer"]["test_freq"] = 6
|
||||
config["trainer"]["total_training_steps"] = 20
|
||||
config["trainer"]["test_freq"] = 20
|
||||
config["trainer"]["experiment_name"] = EXPERIMENT_NAME
|
||||
config["trainer"]["project_name"] = PROJECT_NAME
|
||||
config["trainer"].pop("save_freq", None)
|
||||
|
||||
if ci_fast:
|
||||
# Extra fast CI toggle for testing purposes.
|
||||
config["actor_rollout_ref"]["rollout"]["gpu_memory_utilization"] = 0.6
|
||||
config["trainer"]["total_training_steps"] = 1
|
||||
config["trainer"]["test_freq"] = 1
|
||||
|
||||
|
||||
@@ -65,10 +65,14 @@ module.exports = async function badgeAggregation({ github, context, core, depend
|
||||
workflow_id: dep.workflow,
|
||||
branch: 'main', // Always check the main branch status no matter what
|
||||
status: 'completed', // only completed runs
|
||||
per_page: 1, // latest only
|
||||
per_page: 50, // retrieve latest 50 so we can filter
|
||||
sort: 'created',
|
||||
direction: 'desc',
|
||||
});
|
||||
|
||||
const run = runsData?.workflow_runs?.[0];
|
||||
const filteredRuns = runsData?.workflow_runs?.filter(run => ['schedule', 'workflow_dispatch'].includes(run.event));
|
||||
|
||||
const run = filteredRuns?.[0];
|
||||
if (!run) {
|
||||
failures.push(`No completed run found for ${dep.label} on branch "${branch}"`);
|
||||
continue;
|
||||
|
||||
Reference in New Issue
Block a user