wassname/AntiPaSTO
1
0
Fork 0
mirror of https://github.com/wassname/AntiPaSTO.git synced 2026-09-09 11:12:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

add val loss to info

Browse Source
This commit is contained in:
wassname
2026-01-15 12:28:24 +08:00
parent b05cf9aad2
commit 38546c03d1
1 changed files with 14 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
antipasto/train/train_adapter.py
+14 -3
View File
@@ -873,6 +873,10 @@ def compute_validation_loss(
avg_total = total_loss / n_batches if n_batches > 0 else float("inf")
avg_components = {k: np.mean(v) for k, v in loss_components.items() if not isinstance(v[0], str)}
# Mark all val infos with phase='val' for later filtering
for info in all_infos:
info['phase'] = 'val'
# Extract per-coefficient breakdown (log validation table inline)
df_coef, coef_metrics = extract_coef_metrics(
all_infos, log_table=log_tables,
@@ -880,7 +884,7 @@ def compute_validation_loss(
) if all_infos else (None, {})
val_summary = summarize_phase_for_compare(all_infos, phase="val")
return avg_total, avg_components, df_coef, coef_metrics, val_summary
return avg_total, avg_components, df_coef, coef_metrics, val_summary, all_infos
def train_epoch(
@@ -934,6 +938,9 @@ def train_epoch(
loss_subspace=loss_subspace,
scale_adapter_fn=scale_adapter_fn,
)
# Mark train infos with phase='train' for filtering in df_hist
for info in batch_infos:
info['phase'] = 'train'
infos.extend(batch_infos)
# Epoch-start snapshot: print per-coef table on the very first batch.
@@ -997,12 +1004,14 @@ def train_epoch(
if val_dataloader is not None and opt_step % val_n_steps == 0 and opt_step > 0 and step % config.grad_accum_steps == 0:
# Keep validation compute cadence (early stopping + wandb), but don't spam tables.
log_val = False
val_loss, val_components, val_df_coef, val_coef_metrics, val_summary = compute_validation_loss(
val_loss, val_components, val_df_coef, val_coef_metrics, val_summary, val_infos = compute_validation_loss(
model=model, val_dataloader=val_dataloader, loss_layers=loss_layers, loss_layer_indices=loss_layer_indices, config=config, step=opt_step,
loss_subspace=loss_subspace, flip_stats=flip_stats, scale_adapter_fn=scale_adapter_fn,
log_tables=log_val,
total_steps=total_steps,
)
# Add val infos to main infos list for saving to df_hist
infos.extend(val_infos)
last_val_summary = val_summary
last_val_loss = val_loss
@@ -1073,7 +1082,7 @@ def train_epoch(
# No validation ran during the epoch (e.g., tiny epoch). Compute once here.
# We do not print per-coef tables here to keep epoch-end output compact.
step_for_log = (epoch + 1) * len(train_dataloader) - 1
val_loss, _, _, _, last_val_summary = compute_validation_loss(
val_loss, _, _, _, last_val_summary, val_infos = compute_validation_loss(
model=model,
val_dataloader=val_dataloader,
loss_layers=loss_layers,
@@ -1086,6 +1095,8 @@ def train_epoch(
scale_adapter_fn=scale_adapter_fn,
log_tables=False,
)
# Add val infos to main infos list for saving to df_hist
infos.extend(val_infos)
last_val_loss = val_loss
last_val_step = step_for_log