Skip to content

feat(klx): deepeyes v2 agentic run through and raw_reward match - #309

Open
grootZ wants to merge 2 commits into
redai-studio:mainfrom
grootZ:mr_master
Open

feat(klx): deepeyes v2 agentic run through and raw_reward match#309
grootZ wants to merge 2 commits into
redai-studio:mainfrom
grootZ:mr_master

Conversation

@grootZ

@grootZ grootZ commented Sep 10, 2026

Copy link
Copy Markdown
Contributor

What

镜像:iregistry.baidu-int.com/xpu/xrelax_torch29_ubuntu2204_xsgl0510_dev:20260831_3
启动命令:bash examples/deepeyes_v2_agentic/run_deepeyes_v2_agentic_klx.sh
目前能够跑通,已完成200步并且raw_reward对齐

Why

How

Testing

  • pre-commit run --all-files passes
  • Tests pass (pytest tests/)
  • New tests added (if applicable)
  • Documentation updated (if applicable)

Type of Change

  • Bug fix (non-breaking change that fixes an issue)
  • New feature (non-breaking change that adds functionality)
  • Breaking change (fix or feature that would cause existing functionality to change)
  • Documentation update
  • Refactoring (no functional changes)
  • Performance improvement
  • CI/CD or build changes

Screenshots / Logs

@grootZ
grootZ requested a review from yxyOo as a code owner September 10, 2026 08:39

@rai-studio-bot rai-studio-bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

发现需要在合入前修复的 agent 启动问题,另有并发参数未生效的问题;细节和修复建议见行级评论。Bash 语法检查通过;CI 的 pre-commit 已通过,Python 测试仍在运行。当前无可用昆仑芯训练集群,未执行硬件集成测试。

  • P3 优先级:P3 非行级:PR 描述仍为空模板。请补充昆仑芯硬件与软件镜像版本、实际启动命令、已完成的 rollout/训练步数,以及标题中 “match” 的对齐指标和结果,便于复现和判断验证范围。
Powered by Nyanpasu with gpt-6-astra medium, please check the suggestions carefully.

Comment thread examples/deepeyes_v2_agentic/run_deepeyes_v2_agentic_klx.sh
Comment thread examples/deepeyes_v2_agentic/run_deepeyes_v2_agentic_klx.sh Outdated

@rai-studio-bot rai-studio-bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

描述已补充镜像及 200 步运行、raw_reward 对齐说明,上次的描述建议已部分落实。当前提交未变化,原行级评论中的启动和并发参数问题仍待修复;本次没有新增行级评论。CI 已全部通过。

  • P3 优先级:P3 非行级:新补充的验证命令指向 run_deepeyes_v2_agentic.sh,而本 PR 新增的是 run_deepeyes_v2_agentic_klx.sh,尚无法确认 200 步结果是否覆盖新增脚本。若实际验证的是昆仑芯脚本,请将命令更正为 bash examples/deepeyes_v2_agentic/run_deepeyes_v2_agentic_klx.sh,并注明运行所用提交和必要的环境覆盖;若运行的是原脚本,请明确该验证范围,并补充新增脚本的运行结果。
Powered by Nyanpasu with gpt-6-astra medium, please check the suggestions carefully.

@grootZ grootZ changed the title feat(klx): deepeyes v2 agentic run through and match feat(klx): deepeyes v2 agentic run through and raw_reward match Sep 11, 2026
@Michael-Salon

Copy link
Copy Markdown
Contributor

建议修改:
P1 — 缺少 Agent 必需的 Python 解释器配置,默认启动会失败。
run_deepeyes_v2_agentic_klx.sh (line 134)
KLX 版本删除了原脚本对 DEEPEYES_V2_APP_PYTHON 的初始化、存在性检查和 runtime env 传递,但 run_agent_app.sh 仍在 set -u 下直接使用该变量。外部环境未预设时,每个 Agent 都会退出。已通过最小调用复现:DEEPEYES_V2_APP_PYTHON: unbound variable。建议恢复默认路径、启动校验及 Ray worker 环境传递。

P1 — W&B API key 会明文进入启动日志。
run_deepeyes_v2_agentic_klx.sh (line 12)
脚本启用 set -x 后展开并导出真实 key,随后又通过 --wandb-key 传入 Ray 提交命令;两处都会被 xtrace 输出,命令参数也可能保留在任务元数据中。建议关闭密钥处理及提交阶段的 xtrace,并通过受控环境传递凭据,避免命令行明文参数。

@rai-studio-bot rai-studio-bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

已复查 a2b2094:原 agent 启动阻塞问题已修复,评估并发参数也已生效;未发现仍需阻塞合入的问题。训练侧旧参数的清理建议已补充到原线程,另在行级评论中指出了已有 Ray 集群下 W&B 凭据传递的回归。

Bash 语法及环境构造检查通过;CI 的 pre-commit 已通过,Python 测试仍在运行。无可用昆仑芯集群,未执行硬件集成测试。

  • P3 优先级:P3 非行级:描述已明确镜像、脚本及 200 步验证,建议已部分落实。请补充该次运行所用提交、必要的环境覆盖和 raw_reward 对照结果,以区分已验证配置与当前默认 2000 步配置。
Powered by Nyanpasu with gpt-6-astra medium, please check the suggestions carefully.

Comment thread examples/deepeyes_v2_agentic/run_deepeyes_v2_agentic_klx.sh
Comment thread examples/deepeyes_v2_agentic/run_deepeyes_v2_agentic_klx.sh Outdated
@rai-studio-bot

rai-studio-bot commented Sep 14, 2026

Copy link
Copy Markdown

Nyanpasu 审查看板

审查状态: 💬 已完成 · 有补充意见

审查版本: a3a52f7598fec58e815e405e6cf61a1473d8e574

Agent Python 和 W&B 凭据传递问题已解决。评估并发已修复,训练侧旧参数及验证信息仍待补充;没有新发现。Bash 语法及四种凭据合并隔离检查通过。pre-commit 已通过,Python CI 测试在审查时仍运行中。无可用昆仑芯集群,未执行硬件集成测试。

编号 问题 优先级 状态 规则来源
F1 Agent Python 路径配置和传递 P1 ✅ 已解决
F2 并发参数有效性:评估已修复,训练旧参数仍保留 P2 🛠️ 部分解决
F3 验证提交、环境覆盖及 raw_reward 对照结果 P3 🛠️ 部分解决
F4 已有 Ray 集群下传递 W&B 凭据 P2 ✅ 已解决
Powered by Nyanpasu with gpt-6-astra medium, please check the suggestions carefully.

@rai-studio-bot rai-studio-bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

已复查 a3a52f7:W&B 凭据已合并到 job runtime env,上次的凭据传递问题已解决。普通值、含转义字符的值、空值和占位值的隔离检查均通过,原环境字段得到保留,合并阶段未将测试凭据输出到 xtrace。未发现新的阻塞问题,本次没有新增行级评论;训练侧旧参数的清理建议仍见原讨论。

Bash 语法检查通过;pre-commit 已通过,Python CI 测试仍在运行。无可用昆仑芯集群,未执行硬件集成测试。

  • P3 优先级:P3 非行级:运行说明已部分补充;请继续补充验证提交、必要的环境覆盖及 raw_reward 对照结果,使描述中的 200 步结果可复现。
Powered by Nyanpasu with gpt-6-astra medium, please check the suggestions carefully.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants