GitLab

Issue/170 - Add HYGON support and improve device type handling.

issue/210 support random sampling with random number

issue/199 - 支持qwen3模型

issue/186 support longrope

Issue/168 InfiniLM接入paged attention接口

* 将 `cpp.LlamaForCausalLM` 提出，变为 `infinilm.infer_engine.InferEngine`

* 将 `Config` 构造逻辑拆分至 `AutoConfig` 中

* 在 `examples` 脚本中直接构造 `InferEngine`

* 将 `random_sample` 计算放入模型中

* 为 `InferEngine` 单独实现 `generate`

* 允许通过 `GenerationConfig` 传递 `temperature`、`top_k`、`top_p`

* 将 `random_sample` 处理从 `LlamaForCausalLM` 中转移到 `RankWorker` 里

* 在 `InferEngine.generate` 中直接 `append(output_id)`

* 修复 commit `13aa90c57de369f9985593c0066b6b06a7508b24` 引入的分布式卡死问题

* 将 `InferEngine.forward` 的接口与 C++ 层的 `InferEngine.Input` 对齐

* 提供了 `_measure_and_log_time` 参数来开启之前的 `generate` 内部计时功能

issue/125 add pybind for paged attention cache config

issue/125 添加Paged KV Cache接口

issue/157 - add/adjust cambricon support in scripts

issue/125 统一Cache接口

Issue/147：增加QY机器编译选项

* issue/122 :更新benchmark脚本和README.md
Signed-off-by: Ceng23333 <441651826@qq.com>

* .
Signed-off-by: Ceng23333 <441651826@qq.com>

* fix input_ids
Signed-off-by: Ceng23333 <441651826@qq.com>

* explicitly split mmul all subject
Signed-off-by: Ceng23333 <441651826@qq.com>

---------
Signed-off-by: Ceng23333 <441651826@qq.com>

issue/115 完善bench.py文件

Issue/121 - cache managements

issue/116 - using batched rope

issue/115 - 添加 bench.py