[Iluvatar] Fix FD launch error when specifing CUDA_VISBLE_DEVICE

Author: wuyujiji

.github/workflows/ci_iluvatar.yml

@@ -22,7 +22,7 @@ jobs:
 
       - name: Code Checkout
         env:
-          docker_image: ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:paddle-ocr-vl-1107
+          docker_image: ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:fd-ci
         run: |
           REPO="https://github.com/${{ github.repository }}.git"
           FULL_REPO="${{ github.repository }}"
@@ -55,7 +55,7 @@ jobs:
 
       - name: Run CI unittest
         env:
-          docker_image: ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:paddle-ocr-vl-1107
+          docker_image: ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:fd-ci
         run: |
           runner_name="${{ runner.name }}"
           last_char="${runner_name: -1}"

docs/get_started/installation/iluvatar_gpu.md

@@ -10,14 +10,14 @@
 Pull the Docker image
 
 ```bash
-docker pull ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:paddle-ocr-vl-1107
+docker pull ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:3.3.0
 ```
 
 ## 3. Container Preparation
 ### 3.1 Start Container
 
 ```bash
-docker run -itd --name paddle_infer --network host -v /usr/src:/usr/src -v /lib/modules:/lib/modules -v /dev:/dev -v /home/paddle:/home/paddle --privileged --cap-add=ALL --pid=host ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:paddle-ocr-vl-1107
+docker run -itd --name paddle_infer --network host -v /usr/src:/usr/src -v /lib/modules:/lib/modules -v /dev:/dev -v /home/paddle:/home/paddle --privileged --cap-add=ALL --pid=host ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:3.3.0
 docker exec -it paddle_infer bash
 ```
 
@@ -26,8 +26,8 @@ docker exec -it paddle_infer bash
 ### 3.2 Install paddle
 
 ```bash
-pip3 install paddlepaddle==3.3.0.dev20251103 -i https://www.paddlepaddle.org.cn/packages/nightly/cpu/
-pip3 install paddle-iluvatar-gpu==3.0.0.dev20251107 -i https://www.paddlepaddle.org.cn/packages/nightly/ixuca/
+pip3 install paddlepaddle==3.3.0.dev20251219 -i https://www.paddlepaddle.org.cn/packages/nightly/cpu/
+pip3 install paddle-iluvatar-gpu==3.0.0.dev20251223 -i https://www.paddlepaddle.org.cn/packages/nightly/ixuca/
 ```
 
 ### 3.3 Install or build FastDeploy
@@ -480,7 +480,7 @@ python3 -m fastdeploy.entrypoints.openai.api_server \
        --cache-queue-port 55660 \
        --max-model-len 16384 \
        --max-num-batched-tokens 16384 \
-       --max-num-seqs 32 \
+       --max-num-seqs 64 \
        --workers 2 \
        --block-size 16
 ```

docs/zh/get_started/installation/iluvatar_gpu.md

@@ -10,14 +10,14 @@
 Pull the Docker image
 
 ```bash
-docker pull ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:paddle-ocr-vl-1107
+docker pull ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:3.3.0
 ```
 
 ## 3. 准备容器
 ### 3.1 启动容器
 
 ```bash
-docker run -itd --name paddle_infer --network host -v /usr/src:/usr/src -v /lib/modules:/lib/modules -v /dev:/dev -v /home/paddle:/home/paddle --privileged --cap-add=ALL --pid=host ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:paddle-ocr-vl-1107
+docker run -itd --name paddle_infer --network host -v /usr/src:/usr/src -v /lib/modules:/lib/modules -v /dev:/dev -v /home/paddle:/home/paddle --privileged --cap-add=ALL --pid=host ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:3.3.0
 docker exec -it paddle_infer bash
 ```
 
@@ -26,8 +26,8 @@ docker exec -it paddle_infer bash
 ### 3.2 安装paddle
 
 ```bash
-pip3 install paddlepaddle==3.3.0.dev20251103 -i https://www.paddlepaddle.org.cn/packages/nightly/cpu/
-pip3 install paddle-iluvatar-gpu==3.0.0.dev20251107 -i https://www.paddlepaddle.org.cn/packages/nightly/ixuca/
+pip3 install paddlepaddle==3.3.0.dev20251219 -i https://www.paddlepaddle.org.cn/packages/nightly/cpu/
+pip3 install paddle-iluvatar-gpu==3.0.0.dev20251223 -i https://www.paddlepaddle.org.cn/packages/nightly/ixuca/
 ```
 
 ### 3.3 安装fastdeploy
@@ -480,7 +480,7 @@ python3 -m fastdeploy.entrypoints.openai.api_server \
        --cache-queue-port 55660 \
        --max-model-len 16384 \
        --max-num-batched-tokens 16384 \
-       --max-num-seqs 32 \
+       --max-num-seqs 64 \
        --workers 2 \
        --block-size 16
 ```

fastdeploy/engine/engine.py

@@ -577,6 +577,10 @@ def _start_worker_service(self):
         if self.cfg.structured_outputs_config.logits_processors is not None:
             arguments += f" --logits-processors {' '.join(self.cfg.structured_outputs_config.logits_processors)}"
 
+        # TODO (iluvatar): remove aftet paddle fix launch error
+        if current_platform.is_iluvatar() and "CUDA_VISIBLE_DEVICES" in os.environ:
+            arguments = arguments.replace(f"--devices {self.cfg.parallel_config.device_ids}", "")
+
         worker_store_true_flag = {
             "enable_expert_parallel": self.cfg.parallel_config.enable_expert_parallel,
             "enable_chunked_moe": self.cfg.parallel_config.enable_chunked_moe,

fastdeploy/model_executor/ops/iluvatar/moe_ops.py

@@ -17,7 +17,7 @@
 from typing import Optional
 
 import paddle
-from paddle.incubate.nn.functional import swiglu
+from paddle.nn.functional import swiglu
 from paddle.nn.quant import weight_only_linear
 
 try:

scripts/run_ci_iluvatar.sh

@@ -14,8 +14,8 @@ ln -sf /usr/local/bin/python3 /usr/local/bin/python
 echo "pip requirements"
 python -m pip install -r requirements_iluvatar.txt
 echo "install paddle cpu and custom device"
-python -m pip install paddlepaddle==3.3.0.dev20251103 -i https://www.paddlepaddle.org.cn/packages/nightly/cpu/
-python -m pip install paddle-iluvatar-gpu==3.0.0.dev20251107 -i https://www.paddlepaddle.org.cn/packages/nightly/ixuca/
+python -m pip install paddlepaddle==3.3.0.dev20251219 -i https://www.paddlepaddle.org.cn/packages/nightly/cpu/
+python -m pip install paddle-iluvatar-gpu==3.0.0.dev20251223 -i https://www.paddlepaddle.org.cn/packages/nightly/ixuca/
 echo "build whl"
 bash build.sh || exit 1