rollout_log_probs如果是None，会影响GRPO的训练吗
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 8.5k
- Forks
- 1.3k
- Avg merge
- 5h 36m
- Merged PRs (30d)
- 22
Description
- 如题。
- eval x: {'eval/aime': 0.0698958333333333, ....}，这个eval中的值是reward的平均值吗，用来作为评测指标用？为什么没用那种专门的官方benchmark的评测
- 日志里主要看哪些指标来判断模型训练有没有问题呢
- strands_agent的user question，应不应该加apply_chat_template。messages角度上，不应该加，但是sglang加载的这个训练模型的要吃的话，感觉又需要加，像search-r1的example里就有加。请教一下
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by tracing where rollout_log_probs, eval metrics, and the strands_agent user question are handled; the issue does not name files or tests. Clarify each question separately with the maintainers, then document the expected behavior and any reproducible training impact before defining a code change.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100