JiajunLI/roboimi
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(train): 添加验证集

Browse Source
This commit is contained in:
gouhanke
2026-02-06 18:00:09 +08:00
parent ea49e63eb7
commit 3d0c2ec5b1
1 changed files with 10 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
roboimi/demos/vla_scripts/train_vla.py
View File

@@ -270,40 +270,39 @@ def main(cfg: DictConfig):
log.info(f"Step {step}/{cfg.train.max_steps} | Loss: {loss.item():.4f}") log.info(f"Step {step}/{cfg.train.max_steps} | Loss: {loss.item():.4f}")
# ===================================================================== # =====================================================================
# Validation # Checkpoint saving & Validation
# ===================================================================== # =====================================================================
val_freq = int(cfg.train.get('val_freq', cfg.train.log_freq)) if step > 0 and step % cfg.train.save_freq == 0:
if val_loader is not None and val_freq > 0 and step % val_freq == 0: # Run validation
val_loss = run_validation() val_loss = run_validation()
if val_loss is not None: if val_loss is not None:
log.info(f"Step {step}/{cfg.train.max_steps} | Val Loss: {val_loss:.4f}") log.info(f"Step {step}/{cfg.train.max_steps} | Val Loss: {val_loss:.4f}")
# =====================================================================
# Checkpoint saving
# =====================================================================
if step > 0 and step % cfg.train.save_freq == 0:
checkpoint_path = checkpoint_dir / f"vla_model_step_{step}.pt" checkpoint_path = checkpoint_dir / f"vla_model_step_{step}.pt"
torch.save({ torch.save({
'step': step, 'step': step,
'model_state_dict': agent.state_dict(), 'model_state_dict': agent.state_dict(),
'optimizer_state_dict': optimizer.state_dict(), 'optimizer_state_dict': optimizer.state_dict(),
'loss': loss.item(), 'loss': loss.item(),
'val_loss': val_loss,
'dataset_stats': dataset_stats, 'dataset_stats': dataset_stats,
}, checkpoint_path) }, checkpoint_path)
log.info(f"💾 Checkpoint saved: {checkpoint_path}") log.info(f"💾 Checkpoint saved: {checkpoint_path}")
# Save best model # Save best model based on validation loss
if loss.item() < best_loss: eval_loss = val_loss if val_loss is not None else loss.item()
best_loss = loss.item() if eval_loss < best_loss:
best_loss = eval_loss
best_model_path = checkpoint_dir / "vla_model_best.pt" best_model_path = checkpoint_dir / "vla_model_best.pt"
torch.save({ torch.save({
'step': step, 'step': step,
'model_state_dict': agent.state_dict(), 'model_state_dict': agent.state_dict(),
'optimizer_state_dict': optimizer.state_dict(), 'optimizer_state_dict': optimizer.state_dict(),
'loss': loss.item(), 'loss': loss.item(),
'val_loss': val_loss,
'dataset_stats': dataset_stats, 'dataset_stats': dataset_stats,
}, best_model_path) }, best_model_path)
log.info(f"🌟 Best model updated: {best_model_path} (loss: {best_loss:.4f})") log.info(f"🌟 Best model updated: {best_model_path} (val_loss: {best_loss:.4f})")
# ========================================================================= # =========================================================================
# 5. Save Final Model # 5. Save Final Model