AstrAI/astrai
ViperEkura 2c2697390d feat: 新增 GradientCheckpointingCallback
- TrainConfig.gradient_checkpointing_modules 指定模块类型
- apply 递归遍历,兼容 DDP,不硬编码模型结构
- modules=None 时静默跳过,零开销
2026-05-17 18:21:05 +08:00
..
config feat: 新增 GradientCheckpointingCallback 2026-05-17 18:21:05 +08:00
dataset feat: 数据集加载时校验必填字段 2026-05-17 11:50:38 +08:00
inference fix: 移除多余 request 参数并增强 tokenizer 健壮性 2026-05-17 12:52:18 +08:00
model refactor: Transformer更名为AutoRegressiveLM并新增EmbeddingEncoder 2026-05-17 15:29:20 +08:00
parallel refactor: 优化并行训练配置与启动管理 2026-05-17 12:33:10 +08:00
tokenize fix: 移除多余 request 参数并增强 tokenizer 健壮性 2026-05-17 12:52:18 +08:00
trainer feat: 新增 GradientCheckpointingCallback 2026-05-17 18:21:05 +08:00
__init__.py refactor: Transformer更名为AutoRegressiveLM并新增EmbeddingEncoder 2026-05-17 15:29:20 +08:00
factory.py refactor: 工厂 kwargs 过滤及组件参数清理 2026-05-16 16:47:41 +08:00
serialization.py refactor: Config序列化统一BaseConfig基类 2026-05-16 22:06:39 +08:00