frozenleaves/MindSpeed-RL
1
0
Fork 0
mirror of https://gitee.com/ascend/MindSpeed-RL.git synced 2025-10-20 16:23:45 +08:00
Code Issues Packages Projects Releases Wiki Activity

!597 路径校验 安全问题单

Browse Source 
Merge pull request !597 from yzb/master
This commit is contained in:
yzb
2025-09-12 02:15:54 +00:00
committed by i-robot
parent c722ee7795
commit 296f10a4fc
7 changed files with 8 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
SECURITYNOTE.md
View File

@ -64,8 +64,7 @@ MindSpeed-RL 暂时未发布wheel包，无正式对外公开接口，所有功
| 类型 | 文件名 | 公网IP地址/公网URL地址/域名/邮箱地址 | 用途说明 |
| :------------: |:----------------------------------------------------------:| :----------------------------------------------------------: |:-----------------------------------------:|
| 开源引入 | MindSpeed-RL/tests/test_tools/dist_test.py | https://github.com/microsoft/DeepSpeed/blob/master/tests/unit/common.py | 开源引入 common 参考代码 |
| 开源引入 | MindSpeed-RL/mindspeed_rl/models/rollout/vllm_adapter/vllm_parallel_state.py | 8.8.8.8 | 引入公网IP地址 |
| 开源引入 | MindSpeed-RL/mindspeed_rl/workers/base_worker.py | 8.8.8.8 | 引入公网IP地址 |
| 开源引入 | MindSpeed-RL/mindspeed_rl/utils/utils.py | 8.8.8.8 | 引入公网IP地址 |
## 通信安全加固

2
docs/algorithms/dapo.md
View File

@ -15,7 +15,7 @@ MindSpeed RL 仓库复现 [Decoupled Clip and Dynamic sAmpling Policy Optimizati
# 读取 Math 17K 数据集
mkdir dataset
cd dataset/
wget https://huggingface.co/datasets/BytedTsinghua-SIA/DAPO-Math-17k/resolve/main/data/dapo-math-17k.parquet --no-check
wget https://huggingface.co/datasets/BytedTsinghua-SIA/DAPO-Math-17k/resolve/main/data/dapo-math-17k.parquet
cd ..
```

2
docs/algorithms/dpo.md
View File

@ -15,7 +15,7 @@ MindSpeed RL 仓库现已支持 [ Direct Preference Optimization (DPO) ](https:/
# 读取orca_dpo_pairs数据集
mkdir dataset
cd dataset/
wget https://huggingface.co/datasets/Intel/orca_dpo_pairs/resolve/main/orca_rlhf.jsonl --no-check
wget https://huggingface.co/datasets/Intel/orca_dpo_pairs/resolve/main/orca_rlhf.jsonl
cd ..
```

4
docs/algorithms/grpo.md
View File

@ -15,7 +15,7 @@
```bash
# 读取math_17k数据集
cd dataset/
wget https://huggingface.co/datasets/BytedTsinghua-SIA/DAPO-Math-17k/resolve/main/data/dapo-math-17k.parquet --no-check
wget https://huggingface.co/datasets/BytedTsinghua-SIA/DAPO-Math-17k/resolve/main/data/dapo-math-17k.parquet
cd ..
```
@ -32,7 +32,7 @@ bash examples/data/preprocess_data.sh math_17k
# 读取deepscaler数据集
mkdir dataset
cd dataset/
wget https://huggingface.co/datasets/agentica-org/DeepScaleR-Preview-Dataset/resolve/main/deepscaler.json --no-check
wget https://huggingface.co/datasets/agentica-org/DeepScaleR-Preview-Dataset/resolve/main/deepscaler.json
cd ..
```

2
docs/algorithms/ppo.md
View File

@ -18,7 +18,7 @@
# 读取deepscaler数据集
mkdir dataset
cd dataset/
wget https://huggingface.co/datasets/agentica-org/DeepScaleR-Preview-Dataset/resolve/main/deepscaler.json --no-check
wget https://huggingface.co/datasets/agentica-org/DeepScaleR-Preview-Dataset/resolve/main/deepscaler.json
cd ..
```

2
examples/dapo/dapo_trainer_qwen3_32b.sh
View File

@ -26,7 +26,7 @@ CURRENT_IP=$(ip -4 addr show $(ip -o -4 route show to default | awk '{print $5}'
if [ "$MASTER_ADDR" = "$CURRENT_IP" ]; then
# 主节点启动
ray start --head --port 6766 --dashboard-host=0.0.0.0 --node-ip-address=$CURRENT_IP --dashboard-port=8260 --resources='{"NPU": '$NPUS_PER_NODE'}'
ray start --head --port 6766 --dashboard-host=$MASTER_ADDR --node-ip-address=$CURRENT_IP --dashboard-port=8260 --resources='{"NPU": '$NPUS_PER_NODE'}'
while true; do
ray_status_output=$(ray status)

2
examples/grpo/grpo_trainer_qwen3_235b_a22b.sh
View File

@ -29,7 +29,7 @@ CURRENT_IP=$(ifconfig $SOCKET_IFNAME | grep -Eo 'inet (addr:)?([0-9]{1,3}\.){3}[
if [ "$MASTER_ADDR" = "$CURRENT_IP" ]; then
# 主节点启动
ray start --head --port 6766 --dashboard-host=0.0.0.0 --node-ip-address=$CURRENT_IP --dashboard-port=8260 --resources='{"NPU": '$NPUS_PER_NODE'}'
ray start --head --port 6766 --dashboard-host=$MASTER_ADDR --node-ip-address=$CURRENT_IP --dashboard-port=8260 --resources='{"NPU": '$NPUS_PER_NODE'}'
while true; do
ray_status_output=$(ray status)