AstrAI/astrai/inference
ViperEkura 7ddebf2cd9 refactor: 统一采样路径为 Strategy + batch tensor,删除 apply_sampling_strategies
- TemperatureStrategy / TopKStrategy / TopPStrategy 支持 Union[float, Tensor]
- SamplingPipeline.sample() 一条调用完成 apply + softmax + multinomial
- 新增 sample() 独立函数作为 scheduler 入口
- scheduler decode 改为 batch tensor 参数传递,支持任意 batch size
- 删除 apply_sampling_strategies(被 sample() 取代)
2026-05-08 19:07:14 +08:00
..
__init__.py refactor: 统一采样路径为 Strategy + batch tensor,删除 apply_sampling_strategies 2026-05-08 19:07:14 +08:00
cache.py refactor: 设计模式优化 inference 模块导入结构 2026-05-08 16:57:57 +08:00
engine.py refactor: 设计模式优化 inference 模块导入结构 2026-05-08 16:57:57 +08:00
sampling.py refactor: 统一采样路径为 Strategy + batch tensor,删除 apply_sampling_strategies 2026-05-08 19:07:14 +08:00
scheduler.py refactor: 统一采样路径为 Strategy + batch tensor,删除 apply_sampling_strategies 2026-05-08 19:07:14 +08:00
server.py fix: 修复 remove_task 未释放 KV cache slot 导致第二轮对话死锁 2026-05-08 14:53:04 +08:00