AncTe/ginka-generator
1
0
Fork 0
mirror of https://github.com/unanmed/ginka-generator.git synced 2026-05-21 18:31:13 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

chore: 微调参数

Browse Source
This commit is contained in:
unanmed 2026-03-12 20:16:57 +08:00
parent 011d828f1d
commit 4608a601be
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
ginka/train_maskGIT.py
View File

@ -43,7 +43,7 @@ MASK_TOKEN = 15
GENERATE_STEP = 8 GENERATE_STEP = 8
MAP_SIZE = 13 * 13 MAP_SIZE = 13 * 13
HEATMAP_CHANNEL = 9 HEATMAP_CHANNEL = 9
LABEL_SMOOTHING = 0.1 LABEL_SMOOTHING = 0
RAND_RATIO = 0.1 RAND_RATIO = 0.1
MASK_PROBS = [0.5, 0.5] # 纯随机，分块随机 MASK_PROBS = [0.5, 0.5] # 纯随机，分块随机
@ -85,7 +85,7 @@ def train():
optimizer = optim.AdamW(model.parameters(), lr=1e-4, weight_decay=1e-2) optimizer = optim.AdamW(model.parameters(), lr=1e-4, weight_decay=1e-2)
# 自定义调度器允许在 self_prob 提高时重置调度器信息并提高学习率以适应学习 # 自定义调度器允许在 self_prob 提高时重置调度器信息并提高学习率以适应学习
scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=args.epochs, eta_min=1e-5) scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=args.epochs, eta_min=1e-6)
# 用于生成图片 # 用于生成图片
tile_dict = dict() tile_dict = dict()