Merge remote-tracking branch 'origin/main' into elastic

Author: wanglei19991004

.github/workflows/all-tests-metax.yml

@@ -27,7 +27,7 @@ jobs:
       - name: Set Environment Variable
         id: set-env # Assign an ID to this step
         run: |
-          echo "ci_image=localhost:5000/flagscale:maca.ai2.33.0.13-torch2.6-py310-ubuntu22.04-amd64-time2508051515" >> $GITHUB_OUTPUT  # Set output variable
+          echo "ci_image=localhost:5000/flagscale:maca.ai3.0.0.5-torch2.6-py310-ubuntu22.04-amd64-time2508201730" >> $GITHUB_OUTPUT  # Set output variable
 
   functional-tests-inference:
     needs:
@@ -38,7 +38,7 @@ jobs:
         task:
           - deepseek_r1_distill_qwen-metax
           # TODO: Need update for flaggems@v3.0
-          - deepseek_r1_distill_qwen-flaggems-metax
+          # - deepseek_r1_distill_qwen-flaggems-metax  # TODO: test need fix
           - opi_llama3_1_instruct-metax
           - opi_llama3_1_instruct-flaggems-metax
           - qwen3-metax

.github/workflows/all-tests-nvidia.yml

@@ -3,8 +3,26 @@ name: All Tests Nvidia
 on:
   push:
     branches: ["main"]
+    paths-ignore:
+      - 'hardware/BI_V150/**'
+      - 'hardware/Cambricon_MLU/**'
+      - 'hardware/Huawei_Atlas800TA3/**'
+      - 'hardware/Hygon_BW1000/**'
+      - 'hardware/Kunlunxin_R310p/**'
+      - 'hardware/MUSA_S5000/**'
+      - 'hardware/Metax_C550/**'
+      - 'hardware/Tsing_micro/**'
   pull_request:
     branches: ["main"]
+    paths-ignore:
+      - 'hardware/BI_V150/**'
+      - 'hardware/Cambricon_MLU/**'
+      - 'hardware/Huawei_Atlas800TA3/**'
+      - 'hardware/Hygon_BW1000/**'
+      - 'hardware/Kunlunxin_R310p/**'
+      - 'hardware/MUSA_S5000/**'
+      - 'hardware/Metax_C550/**'
+      - 'hardware/Tsing_micro/**'
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}-${{ github.actor }}
@@ -21,8 +39,8 @@ jobs:
         run: |
           echo "ci_image=localhost:5000/flagscale:cuda12.8.1-cudnn9.7.1-python3.12-torch2.7.0-time2507111538" >> $GITHUB_OUTPUT  # Set output variable
 
-  # Megatron Unit Tests with Matrix
-  megatron-unit-tests:
+  # Train Megatron Unit Tests with Matrix
+  unit_tests_train_megatron:
     needs:
       - set-env
     uses: ./.github/workflows/unit-tests-nvidia.yml
@@ -43,34 +61,51 @@ jobs:
           - transformer/moe
           - transformer
           - ./
-    name: "megatron-${{ matrix.subset == './' && 'root' || matrix.subset }}"
+    name: "train_megatron-${{ matrix.subset == './' && 'root' || matrix.subset }}"
     with:
-      backend: megatron
+      backend: train_megatron
       subset: ${{ matrix.subset }}
       image: ${{ needs.set-env.outputs.ci_image }}
 
-  # Flagscale Unit Tests with Matrix
-  flagscale-unit-tests:
+  # Train Flagscale Unit Tests with Matrix
+  unit_tests_train_flagscale:
     needs:
       - set-env
-      - megatron-unit-tests
+      - unit_tests_train_megatron
     uses: ./.github/workflows/unit-tests-nvidia.yml
     strategy:
       matrix:
         subset:
           - runner
           - ./
-    name: "flagscale-${{ matrix.subset == './' && 'root' || matrix.subset }}"
+    name: "train_flagscale-${{ matrix.subset == './' && 'root' || matrix.subset }}"
     with:
-      backend: flagscale
+      backend: train_flagscale
+      subset: ${{ matrix.subset }}
+      image: ${{ needs.set-env.outputs.ci_image }}
+
+  # Inference Flagscale Unit Tests with Matrix
+  unit_tests_inference_flagscale:
+    needs:
+      - set-env
+      - unit_tests_train_flagscale
+    uses: ./.github/workflows/unit-tests-nvidia.yml
+    strategy:
+      matrix:
+        subset:
+          - inference
+          - transforms
+    name: "inference_flagscale-${{ matrix.subset == './' && 'root' || matrix.subset }}"
+    with:
+      backend: inference_flagscale
       subset: ${{ matrix.subset }}
       image: ${{ needs.set-env.outputs.ci_image }}
 
   # Functional Tests with Mision and Type Matrix
-  functional-tests-train:
+  functional_tests_train:
     needs:
       - set-env
-      - flagscale-unit-tests
+      - unit_tests_inference_flagscale
     uses: ./.github/workflows/functional-tests-nvidia.yml
     strategy:
       matrix:
@@ -85,10 +120,10 @@ jobs:
       type: train
       image: ${{ needs.set-env.outputs.ci_image }}
 
-  functional-tests-hetero:
+  functional_tests_hetero:
     needs:
       - set-env
-      - functional-tests-train
+      - functional_tests_train
     # TODO: test need fix
     uses: ./.github/workflows/functional-tests-nvidia.yml
     strategy:
@@ -101,10 +136,10 @@ jobs:
       type: hetero_train
       image: ${{ needs.set-env.outputs.ci_image }}
 
-  functional-tests-inference:
+  functional_tests_inference:
     needs:
       - set-env
-      - functional-tests-hetero
+      - functional_tests_hetero
     uses: ./.github/workflows/functional-tests-nvidia.yml
     strategy:
       matrix:
@@ -119,10 +154,10 @@ jobs:
       type: inference
       image: ${{ needs.set-env.outputs.ci_image }}
 
-  functional-tests-serve:
+  functional_tests_serve:
     needs:
       - set-env
-      - functional-tests-inference
+      - functional_tests_inference
     uses: ./.github/workflows/functional-tests-nvidia.yml
     strategy:
       matrix:
@@ -135,10 +170,10 @@ jobs:
       type: serve
       image: ${{ needs.set-env.outputs.ci_image }}
 
-  functional-tests-rl:
+  functional_tests_rl:
     needs:
       - set-env
-      - functional-tests-serve
+      - functional_tests_serve
     uses: ./.github/workflows/functional-tests-nvidia.yml
     strategy:
       matrix:
@@ -153,13 +188,14 @@ jobs:
   # Check All Tests
   all-tests:
     needs:
-      - megatron-unit-tests
-      - flagscale-unit-tests
-      - functional-tests-train
-      - functional-tests-hetero
-      - functional-tests-inference
-      - functional-tests-serve
-      - functional-tests-rl
+      - unit_tests_train_megatron
+      - unit_tests_train_flagscale
+      - unit_tests_inference_flagscale
+      - functional_tests_train
+      - functional_tests_hetero
+      - functional_tests_inference
+      - functional_tests_serve
+      - functional_tests_rl
     runs-on: ubuntu-latest
     steps:
       - name: All Tests Completed

.github/workflows/functional-tests-metax.yml

@@ -83,7 +83,7 @@ jobs:
             export CXXFLAGS="-I/opt/maca/include -I/opt/maca/include/mcr -I/opt/maca/include/common -I/opt/maca/include/mcsparse -I/opt/maca/include/mcsolver"
             python setup.py bdist_wheel
             pip uninstall vllm -y
-            pip install dist/vllm-0.8.5+maca2.33.0.12torch2.6-cp310-cp310-linux_x86_64.whl
+            pip install dist/vllm-*-linux_x86_64.whl
             source /opt/conda/bin/activate base
           else
             echo "Unknown backend type: ${{ inputs.type }}"

.github/workflows/functional-tests-nvidia.yml

@@ -49,6 +49,13 @@ jobs:
           pip install gitpython # TODO: Remove after updating the image
 
           git config --global --add safe.directory /__w/FlagScale/FlagScale
+
+          TMP_DIR=$(mktemp -d)
+          echo "Copying source to temporary directory: $TMP_DIR"
+          cp -r /__w/FlagScale/FlagScale/. $TMP_DIR
+          cd $TMP_DIR
+          trap 'echo "Cleaning up temporary directory..."; rm -rf "$TMP_DIR"' EXIT
+
           if [ "${{ inputs.type }}" = "train" ] || [ "${{ inputs.type }}" = "hetero_train" ]; then
             PYTHONPATH=./:$PYTHONPATH pip install . --no-build-isolation --verbose --config-settings=device="gpu" --config-settings=backend="Megatron-LM"
             if [ "${{ inputs.task }}" = "llava_onevision" ]; then
@@ -99,4 +106,5 @@ jobs:
           tests/scripts/functional_tests/test_task.sh --type ${{ inputs.type }} --task ${{ inputs.task }}
           exit_code=$?
           echo "Exit code: $exit_code"
+          exit $exit_cod
         shell: bash

.github/workflows/unit-tests-nvidia.yml

@@ -53,10 +53,17 @@ jobs:
           pip install gitpython # TODO: Remove after updating the image
 
           git config --global --add safe.directory /__w/FlagScale/FlagScale
-          if [ "${{ inputs.backend }}" = "megatron" ] || [ "${{ inputs.backend }}" = "flagscale" ]; then
+          
+          TMP_DIR=$(mktemp -d)
+          echo "Copying source to temporary directory: $TMP_DIR"
+          cp -r /__w/FlagScale/FlagScale/. $TMP_DIR
+          cd $TMP_DIR
+          trap 'echo "Cleaning up temporary directory..."; rm -rf "$TMP_DIR"' EXIT
+
+          if [ "${{ inputs.backend }}" = "train_megatron" ] || [ "${{ inputs.backend }}" = "train_flagscale" ]; then
             echo ""
             # PYTHONPATH=./:$PYTHONPATH pip install . --config-settings=backend="Megatron-LM" --verbose --no-build-isolation
-          elif [ "${{ inputs.backend }}" = "vllm" ]; then
+          elif [ "${{ inputs.backend }}" = "vllm" ] || [ "${{ inputs.backend }}" = "inference_flagscale" ]; then
             source /root/miniconda3/bin/activate flagscale-inference
             pip install scikit-build scikit-build-core
             pip install --no-build-isolation git+https://github.com/FlagOpen/FlagGems.git@release_v1.0.0

docker/Dockerfile.metax

@@ -1,4 +1,7 @@
-FROM cr.metax-tech.com/public-ai-release/maca/modelzoo.llm.vllm:maca.ai2.33.0.13-torch2.6-py310-ubuntu22.04-amd64
+ARG MACA_VERSION=ai3.0.0.5
+ARG TORCH_VERSION=2.6
+ARG PYTHON_VERSION=310
+FROM cr.metax-tech.com/public-ai-release/maca/modelzoo.llm.vllm:maca.${MACA_VERSION}-torch${TORCH_VERSION}-py${PYTHON_VERSION}-ubuntu22.04-amd64
 
 ##############################################################################
 # Temporary Installation Directory
@@ -21,9 +24,11 @@ RUN /opt/conda/bin/pip3 install cryptography gitpython cmake
 # used to generate cmakecache files
 # install tooklit Only, IGNORE driver or samples
 ##############################################################################
-RUN wget https://developer.download.nvidia.com/compute/cuda/11.6.0/local_installers/cuda_11.6.0_510.39.01_linux.run && \
-    sh cuda_11.6.0_510.39.01_linux.run --silent --toolkit && \
-    rm cuda_11.6.0_510.39.01_linux.run
+ARG CUDA_VERSION=11.6.0
+ARG TOOLKIT_VERSION=11.6.0_510.39.01
+RUN wget https://developer.download.nvidia.com/compute/cuda/${CUDA_VERSION}/local_installers/cuda_${TOOLKIT_VERSION}_linux.run && \
+    sh cuda_${TOOLKIT_VERSION}_linux.run --silent --toolkit && \
+    rm cuda_${TOOLKIT_VERSION}_linux.run
 
 ##############################################################################
 # Clone FlagScale

examples/aquila/conf/train_hetero.yaml

@@ -12,13 +12,34 @@ experiment:
   runner:
     hostfile: torchrun # Please replace with your actual hostfile path
     rdzv_backend: "static" # hetero training only supports static
-  envs:
-    CUDA_VISIBLE_DEVICES: 0,1,2,3,4,5,6,7
-    CUDA_DEVICE_MAX_CONNECTIONS: 1
   cmds:
     before_start: "ulimit -n 1048576"
     after_stop: ""
 
+  envs:
+    CUDA_VISIBLE_DEVICES: 0,1,2,3,4,5,6,7
+    CUDA_DEVICE_MAX_CONNECTIONS: 1
+
+ # specific envs
+ # the priority: node ip > device type > the global envs
+    # device_type_specific:
+    #   mlu590:
+    #     MLU_VISIBLE_DEVICES: 0,1,2,3,4,5,6,7
+    #     CNCL_IB_RELAXED_ORDERING_ENABLE: 1
+    #     CNCL_MLULINK_OVER_ROCE_DISABLE: 1
+    #   A800:
+    #     CUDA_DEVICE_MAX_CONNECTIONS: 1
+    # node_specific:
+    #   xxx.xxx.xx.xxx:
+    #     CUDA_VISIBLE_DEVICES: 0,1,2,3
+
+# the other specific config for different device type or node
+# device_type_specific:
+#   mlu590:
+#     build_dir: xxxx/build/Cambricon_MLU/FlagScale/
+# node_specific:
+#   xxxxx.xxx.xx.xxx:
+#     build_dir: xxxx/xxxx/build/Cambricon_MLU/FlagScale/
 action: run
 
 hydra:

examples/cloud/conf/serve.yaml

@@ -9,9 +9,9 @@ experiment:
     type: serve
   runner:
     type: cloud
-    hostfile: ${oc.env:HOSTFILE, /etc/hostfile}
-    master_addr: ${oc.env:MASTER_ADDR, 127.0.0.1}
-    master_port: 7396
+    nnodes: ${oc.env:LWS_GROUP_SIZE, 1}
+    master_addr: ${oc.env:LWS_LEADER_ADDRESS, 127.0.0.1}
+    master_port: ${oc.env:MASTER_PORT, 7396}
     device_type: ${oc.env:DEVICE_TYPE, gpu}
     nproc_per_node: ${oc.env:AIRS_ACCELERATOR_NUM, 1}
     deploy:

examples/cloud/conf/serve/cloud_model.yaml

@@ -1,7 +1,7 @@
 - serve_id: vllm_model
   engine: vllm
   engine_args:
-    model: null
+    model: /models/Qwen2.5-0.5B-Instruct
     port: 8000
     host: 0.0.0.0
     served_model_name: cloud

examples/openjourney/conf/inference.yaml

@@ -0,0 +1,26 @@
+defaults:
+  - _self_
+  - inference: t2i
+
+experiment:
+  exp_name: openjourney
+  exp_dir: outputs/${experiment.exp_name}
+  model: prompthero/openjourney
+  task:
+    type: inference
+    backend: vllm # TODO(yupu): Remove this restriction
+    entrypoint: flagscale/inference/diffusion_entrypoint.py
+  runner:
+    hostfile: null
+  cmds:
+    before_start: source /root/miniconda3/bin/activate flagscale-inference
+  envs:
+    CUDA_VISIBLE_DEVICES: 0
+    CUDA_DEVICE_MAX_CONNECTIONS: 1
+    # Optionally, set HF_HOME and HF_ENDPOINT
+
+action: run
+
+hydra:
+  run:
+    dir: ${experiment.exp_dir}/hydra