| fix problem in chapter1 code | 7 个月前 |
| fix: address reviewed issues 642 through 745 | 1 个月前 |
| docs(chapter11): 为 GRPO 学习率补充策略坍塌提示（注释与说明） 章节 11 的 GRPO demo 与文档建议学习率 1e-5~5e-5，在小参数模型 （如 Qwen3-0.6B）+ GSM8K 上，5e-5 会导致 GRPO 策略坍塌 （准确率 57.0% → 2.4%），1e-6 可稳定收敛。 本次仅做注释与文档层面的补充，不改动任何代码行为： - code/chapter11/00_quick_test.py、05_grpo_training.py：在 GRPO 配置处 添加注释，提示学习率过大可能导致策略坍塌，必要时可调至 1e-6； - docs/chapter11（中英两版）：GRPO 学习率建议改为 1e-6~1e-5，并提示风险。 | 1 个月前 |
| add chapter 12 | 11 个月前 |
| fix: address reviewed issues 642 through 745 | 1 个月前 |
| Update pyproject.toml | 6 个月前 |
| fix(chapter15): load backend .env so LLM_API_KEY is read at startup | 2 个月前 |
| update chapter 16 | 6 个月前 |
| update code of chapter2&3 and improve chapter4 | 1 年前 |
| 增加HF_ENDPOINT，避免Connection aborted. | 9 个月前 |
| fix(chapter4): skip empty choices chunks in streaming response Some OpenAI-compatible APIs (e.g. AIHubMix) append a final tail packet with choices=[] containing only usage statistics. Accessing chunk.choices[0] on this packet raises IndexError, which is caught by the outer try/except and returns None — silently discarding all previously streamed content. Adding if not chunk.choices: continue skips these packets while preserving the full response. Fixes #522 | 4 个月前 |
| docs(chapter5): update low-code platform content | 3 个月前 |
| 修复了原代码中忘记写关于创建API模型客户端导致代码无法跑通的问题。 | 8 个月前 |
| fix(ch7): resolve duplicate stream print in MySimpleAgent and docs | 8 天前 |
| fix: address reviewed issues 642 through 745 | 1 个月前 |
| fix: 修复 chapter9 示例中的 LLMResponse 类型问题 | 3 个月前 |