PoliBa-DeepLearning/NanoSocrates
2
0
Fork 0
Code Issues 3 Pull Requests 1 Actions Packages Projects Releases Wiki Activity

dev.train #8

Merged
gape_01 merged 50 commits from dev.train into dev 2025-10-17 22:20:14 +02:00
Conversation 0 Commits 50 Files Changed 49 +2 -2
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files
Showing only changes of commit f98f5a2611 - Show all commits

4
Playgrounds/nanosocrates-train-experiment-2.py
View File

@ -46,7 +46,7 @@ NUMBER_OF_BLOCKS = 4
MAX_EPOCHS = int(3e3)
PRETRAIN_EPOCHS = int(300)
WARMUP_EPOCHS = int(1e3)
MINI_BATCH_SIZE = 300
MINI_BATCH_SIZE = 80
VALIDATION_STEPS = 5
CHECKPOINT_STEPS = VALIDATION_STEPS * 4
PATIENCE = 4
@ -185,7 +185,7 @@ while current_epoch < MAX_EPOCHS:
pred_logits = ENCODER_ONLY((enc_x, enc_x_pad))
pred_logits = pred_logits.permute(0, 2, 1)
print(torch.max(tgt))
# print(torch.max(tgt))
loss: torch.Tensor = encoder_ce(pred_logits, tgt)
loss.backward()