AstrAI

Commit Graph

Author	SHA1	Message	Date
ViperEkura	88ec63121d	feat : GPT-2 residual scaling weight init - Linear: normal(0, init_std) replaces kaiming_uniform_(a=sqrt(5)) - o_proj / mlp.down: init_std = 0.02 / sqrt(2 * n_layers) - MoE: expert down scaled by 1/sqrt(1/n_shared + 1/K) - Embedding: normal(0, 0.02), unchanged	2026-06-25 15:08:31 +08:00
ViperEkura	1d54491809	refactor: 改用递归子模块 init 替代统一 normal_(0.006) - Embedding.reset_parameters: normal_(std=0.02) - Linear.reset_parameters: kaiming_uniform_ + uniform_ bias - Transformer._init_weights 通过 apply 递归调用子模块 reset_parameters - 移除全局 normal_(0.006) 覆盖，各模块使用更合适的分布	2026-05-17 10:44:18 +08:00
ViperEkura	ef25efffa2	refactor: 拆分 module.py 为 components 子包 - rope/linear/norm/embedding/mlp/attention/decoder_block 各自独立文件 - 依赖单向无循环 - 公开接口不变，外部无需修改	2026-05-15 20:08:36 +08:00

Author

SHA1

Message

Date

ViperEkura

88ec63121d

feat : GPT-2 residual scaling weight init

- Linear: normal(0, init_std) replaces kaiming_uniform_(a=sqrt(5))
- o_proj / mlp.down: init_std = 0.02 / sqrt(2 * n_layers)
- MoE: expert down scaled by 1/sqrt(1/n_shared + 1/K)
- Embedding: normal(0, 0.02), unchanged

2026-06-25 15:08:31 +08:00

ViperEkura

1d54491809

refactor: 改用递归子模块 init 替代统一 normal_(0.006)

- Embedding.reset_parameters: normal_(std=0.02)
- Linear.reset_parameters: kaiming_uniform_ + uniform_ bias
- Transformer._init_weights 通过 apply 递归调用子模块 reset_parameters
- 移除全局 normal_(0.006) 覆盖，各模块使用更合适的分布

2026-05-17 10:44:18 +08:00

ViperEkura

ef25efffa2

refactor: 拆分 module.py 为 components 子包

- rope/linear/norm/embedding/mlp/attention/decoder_block 各自独立文件
- 依赖单向无循环
- 公开接口不变，外部无需修改

2026-05-15 20:08:36 +08:00

3 Commits