|
|
5d95027894
|
refactor: 重写沙比 AI 写的训练代码
|
2026-05-13 18:20:31 +08:00 |
|
|
|
3676958781
|
feat: 分三阶段训练
Co-authored-by: Copilot <copilot@github.com>
|
2026-05-07 20:59:22 +08:00 |
|
|
|
b9032d94c8
|
refactor: 数据处理也将资源统一
Co-authored-by: Copilot <copilot@github.com>
|
2026-05-06 21:37:30 +08:00 |
|
|
|
3e273c5a9d
|
feat: 资源压缩为一种 tile
Co-authored-by: Copilot <copilot@github.com>
|
2026-05-05 21:46:38 +08:00 |
|
|
|
3874d4dd95
|
feat: VQ 编码器分为三部分
Co-authored-by: Copilot <copilot@github.com>
|
2026-04-28 18:41:57 +08:00 |
|
|
|
abbad781ab
|
feat: 添加少数结构性标签
Co-authored-by: Copilot <copilot@github.com>
|
2026-04-27 14:56:21 +08:00 |
|
|
|
21315a6cb0
|
chore: 调整超参数与细节
Co-authored-by: Copilot <copilot@github.com>
|
2026-04-27 13:46:48 +08:00 |
|
|
|
068940cae0
|
refactor: 采用 VQ + MaskGIT 方案
Co-authored-by: Copilot <copilot@github.com>
|
2026-04-26 23:45:56 +08:00 |
|
|
|
765cdcaeb0
|
feat: 冻结 MaskGIT 并联合训练
|
2026-04-22 22:15:05 +08:00 |
|
|
|
df23c891c6
|
fix: MaskGIT 训练调整
|
2026-04-09 12:47:18 +08:00 |
|
|
|
9814edd1b4
|
fix: 数据集报错
|
2026-04-08 19:42:18 +08:00 |
|
|
|
dbb0b9064c
|
feat: 降低 MaskGIT 对条件的依赖
|
2026-04-08 19:36:46 +08:00 |
|
|
|
b4f49e702d
|
feat: heatmap 生成模型
|
2026-04-05 22:13:58 +08:00 |
|
|
|
36a4faff4e
|
fix: 报错
|
2026-03-31 22:35:28 +08:00 |
|
|
|
8de66d87f1
|
feat: 随机数据增强
|
2026-03-31 22:24:25 +08:00 |
|
|
|
c64a783d5e
|
fix: 报错
|
2026-03-30 12:34:18 +08:00 |
|
|
|
020c2cf168
|
feat: 在训练时处理高斯模糊
|
2026-03-13 20:14:20 +08:00 |
|
|
|
d0f86018f1
|
chore: 调整部分依赖
|
2026-03-12 20:41:03 +08:00 |
|
|
|
513f27c7ac
|
refactor: 仅保留 SOTA
|
2026-03-12 20:36:49 +08:00 |
|
|
|
22a2db464f
|
feat: maskGIT 加入热力图条件限定
|
2026-03-11 16:27:27 +08:00 |
|
|
|
d312fb65b7
|
feat: maskGIT
|
2026-03-10 23:06:23 +08:00 |
|
|
|
771553b8b8
|
feat: rnn train
|
2025-12-13 17:38:57 +08:00 |
|
|
|
7fb1625e0b
|
feat: rnn
|
2025-12-12 14:57:06 +08:00 |
|
|
|
5586ea1039
|
chore: 微调模型
|
2025-06-14 15:06:09 +08:00 |
|
|
|
fb0323d874
|
perf: 改进生成流程与损失计算
|
2025-05-15 17:44:29 +08:00 |
|
|
|
fa48863946
|
perf: 模型微调
|
2025-05-11 23:50:08 +08:00 |
|
|
|
e3e496957c
|
feat: 损失值改进
|
2025-05-02 13:59:56 +08:00 |
|
|
|
53041ab754
|
perf: 改进网络结构
|
2025-05-01 22:08:39 +08:00 |
|
|
|
dc1eb8fa92
|
feat: 提高标量数量
|
2025-04-30 21:00:45 +08:00 |
|
|
|
44b90e7630
|
refactor: 修改目录结构 & feat: 条件注入
|
2025-04-29 18:24:01 +08:00 |
|
|
|
a94b07bda8
|
feat: 改进训练流程
|
2025-04-20 22:06:20 +08:00 |
|
|
|
87016c67e8
|
feat: 改进输出头和特征融合
|
2025-04-19 15:05:26 +08:00 |
|
|
|
9b5abf177c
|
feat: 添加 js 散度损失
|
2025-04-16 18:39:17 +08:00 |
|
|
|
447c28ff5e
|
perf: 修改损失值计算方式
|
2025-04-14 14:15:13 +08:00 |
|
|
|
f6b1ad6ebd
|
refactor: 课程训练
|
2025-04-13 21:06:07 +08:00 |
|
|
|
268b21e0b7
|
perf: 调优超参数和模型
|
2025-04-07 16:41:27 +08:00 |
|
|
|
29cfb4d029
|
refactor: 改为 Wasserstein GAN
|
2025-04-06 18:44:18 +08:00 |
|
|
|
6a1aeaa77e
|
fix: 浮点数被意外转成 LongTensor
|
2025-04-03 12:12:13 +08:00 |
|
|
|
325eb599c3
|
feat: 支持流式传输，改进数据集
|
2025-04-02 17:57:02 +08:00 |
|
|
|
8dea79a9f0
|
feat: 使用 IPC 通讯并实现单脚本对抗训练
|
2025-04-01 22:41:23 +08:00 |
|
|
|
49ee543732
|
fix: 数据集报错
|
2025-03-26 22:44:11 +08:00 |
|
|
|
8f892fc7f4
|
feat: 优化图块数量损失函数
|
2025-03-26 21:25:09 +08:00 |
|
|
|
f169167409
|
feat: 提高参数量
|
2025-03-24 16:59:53 +08:00 |
|
|
|
fd72b1e7f4
|
refactor: GINKA Model 改为 softmax 输出
|
2025-03-19 21:24:29 +08:00 |
|
|
|
09c63fedce
|
refactor: GINKA 生成器使用 Minamo 作为损失值的一部分
|
2025-03-18 23:51:14 +08:00 |
|
|
|
1566acf691
|
refactor: 重构 ginka 模型部分
|
2025-03-18 21:11:53 +08:00 |
|
|
|
9f62be3736
|
init: Minamo Model
|
2025-03-15 22:26:31 +08:00 |
|
