add Metax CICD workflow

.github/configs/metax.yml

@@ -0,0 +1,98 @@
+# CUDA Hardware Configuration for Megatron-LM-FL
+# This file defines CI/CD settings for CUDA-based testing
+# Test configurations are defined in tests/test_utils/config/platforms/cuda.yaml
+
+hardware_name: metax
+display_name: 'Metax Tests'
+
+# Docker image for this hardware
+ci_image: localhost:5000/cr.metax-tech.com/public-ai-release/maca/megatron-lm:0.12.0-maca.ai3.3.0.11-torch2.6-py312-ubuntu22.04-amd64
+
+# Runner labels for this hardware
+runner_labels:
+  - self-hosted
+  - Linux
+  - X64
+  - metax
+  - dev 
+
+# Container volumes (hardware-specific paths)
+container_volumes:
+  - /home/flagscale_cicd/flask/static:/workspace/report
+  - /home/flagscale_cicd/flask/config:/workspace/config
+  - /home/flagscale_cicd/docker/docker_build/docker_data:/home/gitlab-runner/data
+  - /home/flagscale_cicd/docker/docker_build/docker_tokenizers:/home/gitlab-runner/tokenizers
+  - /home/flagscale_cicd/docker/docker_build/docker_data/Megatron-LM/datasets:/opt/data/datasets
+  - /home/flagscale_cicd/docker/docker_build/docker_tokenizers/Megatron-LM/tokenizers:/opt/data/tokenizers
+
+  # ===== 需要修改 3：沐曦设备路径 =====
+  # 这些路径必须存在，否则容器启动会失败
+  - /dev/dri:/dev/dri                                  # 沐曦 GPU 设备文件
+  - /dev/mxcd:/dev/mxcd                                # 沐曦 GPU 设备文件
+  - /opt/maca:/opt/maca                                # MACA 软件栈
+  - /opt/mxdriver:/opt/mxdriver                        # 沐曦驱动路径
+
+# Container options (hardware-specific settings)
+# container_options: '--privileged --gpus all --shm-size=500g --hostname megatron_cicd --user root --ulimit nofile=65535:65535'
+# MetaX 容器选项 
+# ===== 需要修改 4：容器选项 =====
+container_options: >-
+  --privileged
+  --shm-size=500g
+  --hostname megatron_cicd
+  --user root
+  --ulimit nofile=65535:65535
+  --device=/dev/dri
+  --device=/dev/mxcd
+  --group-add video
+  -e PLATFORM=metax
+  -e TORCH_DISTRIBUTED_BACKEND=mccl
+  -e MUSA_VISIBLE_DEVICES=all
+device_types:
+  - C500
+
+# Test matrix configuration
+test_matrix:
+  unit:
+    devices:
+      - C500
+    # Ignored test files for unit tests
+    # These files will be skipped when running pytest
+    ignored_tests:
+      - tests/unit_tests/data/test_preprocess_data.py
+      - tests/unit_tests/dist_checkpointing/test_global_metadata_reuse.py
+      - tests/unit_tests/dist_checkpointing/test_optimizer.py
+      - tests/unit_tests/dist_checkpointing/test_nonpersistent.py
+      - tests/unit_tests/dist_checkpointing/test_optimizer.py
+      - tests/unit_tests/dist_checkpointing/test_safe_globals.py
+      - tests/unit_tests/dist_checkpointing/models/test_moe_experts.py
+      - tests/unit_tests/distributed/test_grad_sync_with_expert_parallel.py
+      - tests/unit_tests/distributed/test_mcore_fully_sharded_data_parallel.py
+      - tests/unit_tests/export/trtllm/test_distributed_fp8.py
+      - tests/unit_tests/export/trtllm/test_single_device_fp8.py
+      - tests/unit_tests/transformer/moe/test_a2a_token_dispatcher.py
+      - tests/unit_tests/test_inference.py
+      - tests/unit_tests/test_rl_utils.py
+      - tests/unit_tests/models/test_gpt_model.py
+      - tests/unit_tests/models/test_mamba_model.py
+      - tests/unit_tests/post_training/test_modelopt_module_spec.py
+      - tests/unit_tests/transformer/moe/test_aux_loss.py
+      - tests/unit_tests/transformer/moe/test_moe_layer_discrepancy.py
+      - tests/unit_tests/transformer/moe/test_routers.py
+      - tests/unit_tests/transformer/test_attention.py
+      - tests/unit_tests/transformer/test_attention_packed_seq.py
+      - tests/unit_tests/transformer/test_cuda_graphs.py
+      - tests/unit_tests/transformer/test_full_cuda_graph.py
+      - tests/unit_tests/transformer/test_multi_latent_attention.py
+      - tests/unit_tests/transformer/test_multi_token_prediction.py
+      - tests/unit_tests/transformer/test_retro_attention.py
+      - tests/unit_tests/transformer/test_transformer_block.py
+      - tests/unit_tests/transformer/test_transformer_block_custom_pgs.py
+      - tests/unit_tests/dist_checkpointing/test_local.py
+
+  # functional:
+  #   train:
+  #     - device: C500
+  #       task: train
+  #       model: gpt
+  #       case: all