feat(train): 跑通训练脚本

2026-02-05 14:08:43 +08:00
parent dd2749cb12
commit b0a944f7aa
17 changed files with 1002 additions and 464 deletions
@@ -1,12 +1,13 @@
 defaults:
  - _self_
-  - agent: base_siglip
-  - data: custom_hdf5  # 新增这一行，激活数据配置
+  - agent: resnet_diffusion
+  - data: resnet_dataset

 train:
-  batch_size: 4       # 减小 batch size 方便调试
-  lr: 1e-4
-  max_steps: 10
-  log_freq: 10
-  device: "cpu"
-  num_workers: 0      # 调试设为0，验证通过后改为 2 或 4
+  batch_size: 8       # Batch size for training
+  lr: 1e-4            # Learning rate
+  max_steps: 10000    # Maximum training steps
+  log_freq: 100       # Log frequency (steps)
+  save_freq: 1000     # Save checkpoint frequency (steps)
+  device: "cuda"      # Device: "cuda" or "cpu"
+  num_workers: 8      # DataLoader workers (set to 0 for debugging, 8 for production)